cs.RO / 1 / 2607.24860
Egocentric Station Holding of Robotic Fish in Unknown Turbulent Background Flow
在未知湍流背景流中机器人鱼的自我中心驻留控制
Abstract
Approaching a target position and holding station in flowing water is a fundamental and critical capability for robotic fish operating in natural aquatic environments. Despite decades of advances in enhancing swimming efficiency and maneuverability, this capability remains underdeveloped, largely owing to the insufficiently characterized, highly nonlinear fluid-structure interactions inherent to freely swimming robotic fish in flows. To bridge this gap, we propose the SWiFT framework, a Swimming With Flow Toolbox that enables the efficient exploration of an egocentric station-holding policy for a body and/or caudal fin (BCF) robotic fish in unknown and turbulent background flows via reinforcement learning (RL). Our SWiFT integrates a free-swimming flow-tank experimental setup with a highly efficient, physically consistent computational fluid dynamics (CFD)-based simulator and a systematic sim-to-real transfer pipeline. The resulting policy achieves substantial improvements over state-of-the-art methods across all metrics, most notably root-mean-square error (RMSE) of distance. Furthermore, we validated that egocentric feedback alone, without any explicit flow sensing, enables station-holding in unknown turbulent flows, closely mirroring the biological phenomenon of rheotaxis. Accordingly, the success of this egocentric station-holding policy not only advances robotic fish control toward real-world deployment, but also highlights SWiFT's promise as a foundation for tackling complex swimming tasks for underwater robots.
Chinese Translation
在流动水域中接近目标位置并保持驻留是机器人鱼在自然水域环境中操作的基本且关键的能力。尽管在提高游泳效率和机动性方面取得了数十年的进展，这一能力仍然发展不足，主要是由于自由游动的机器人鱼在流动中固有的流体-结构相互作用的非线性特性尚未得到充分表征。为了解决这一问题，我们提出了SWiFT框架（Swimming With Flow Toolbox），该框架通过强化学习（RL）使得在未知和湍流背景流中对身体和/或尾鳍（BCF）机器人鱼的自我中心驻留控制策略进行高效探索。我们的SWiFT将自由游动的水槽实验设置与高效、物理一致的基于计算流体动力学（CFD）的模拟器以及系统的仿真到现实转移管道相结合。所得到的策略在所有指标上都显著优于现有最先进的方法，特别是在距离的均方根误差（RMSE）方面。此外，我们验证了仅凭自我中心反馈，无需任何显式流动感知，便能在未知湍流中实现驻留控制，这与生物现象的流动趋向性（rheotaxis）密切相似。因此，这一自我中心驻留控制策略的成功不仅推动了机器人鱼控制向现实世界的应用发展，也凸显了SWiFT作为解决水下机器人复杂游泳任务基础的潜力。
cs.RO / 2 / 2607.24863
Steeringless Drifting: Differential-Torque Control of a Four-Wheel Independently Driven Vehicle
无转向漂移：四轮独立驱动车辆的差分扭矩控制
Abstract
Control methods for emerging vehicle chassis architectures are important for autonomous driving near handling limits. Unlike conventional drift control, which relies on mechanical steering and rear-tire saturation, a steering-free four-wheel independently driven (4WID) vehicle can generate direct yaw moment through differential wheel torques. This paper proposes a differential-torque drift control method for such a vehicle. A double-track vehicle model incorporating four-wheel differential actuation is established, based on which a drift-equilibrium calculation method and a closed-loop drift controller are developed. The proposed approach is validated through simulations and experiments on a 1:10-scale vehicle. The results show that the vehicle can achieve steady circular drifting with a sideslip angle of approximately 20$^\circ$ and perform figure-eight drift tracking. This study demonstrates the feasibility of drift control using only differential wheel torques and provides a new perspective on near-limit control for steering-free vehicle architectures.
Chinese Translation
针对新兴车辆底盘架构的控制方法在接近操控极限的自动驾驶中至关重要。与依赖机械转向和后轮饱和的传统漂移控制不同，无转向的四轮独立驱动（4WID）车辆可以通过差分轮胎扭矩直接产生偏航力矩。本文提出了一种针对该类车辆的差分扭矩漂移控制方法。基于四轮差分驱动的双轨车辆模型，建立了漂移平衡计算方法和闭环漂移控制器。通过对1:10比例车辆的仿真和实验验证了所提方法。结果表明，该车辆能够以约20$^ ext{°}$的侧滑角实现稳定的圆形漂移，并进行八字漂移追踪。本研究展示了仅使用差分轮胎扭矩进行漂移控制的可行性，并为无转向车辆架构的近极限控制提供了新的视角。
cs.RO / 3 / 2607.24959
Amortising Trajectory Optimisation for Residual MPC via Implicit Contact Differentiation
通过隐式接触微分的剩余模型预测控制的摊销轨迹优化
Abstract
Differentiable simulation can accelerate contact-rich trajectory optimisation by exposing local sensitivities of task outcomes to controls. Existing approaches either use finite differences, which are expensive and step-size sensitive; differentiate iterative contact solvers by unrolling automatic differentiation (AD), which stores a growing computation trace; or require intricate, solver-specific KKT sensitivity derivations. We introduce an AD-assisted implicit derivative for regularised smooth contacts and apply it to Mujoco MJX, based on the Implicit Function Theorem (IFT). The method differentiates the stationarity residual at the tolerance-converged solution, avoiding both solver unrolling and hand-assembled KKT systems. IFT keeps compiled temporary memory nearly constant with solver effort, changing by less than 4$\%$ from one to ten iterations versus 10.6$\times$ growth for unrolled AD. IFT memory grows slower with active contacts and model dimension, using 20$\times$ less memory at 256 contacts and 6$\times$ less at 16 contacts and 96 DoF. We further introduce optimiser distillation for residual MPC, amortising batched full-horizon iLQR into a policy that guides short-horizon residual iLQR. Across Finger, Franka, and Unitree, this raises six-step success by 28-98 percentage points over standard iLQR.
Chinese Translation
可微仿真可以通过暴露任务结果对控制的局部敏感性来加速接触丰富的轨迹优化。现有方法要么使用有限差分，这种方法成本高且对步长敏感；要么通过展开自动微分（AD）来对迭代接触求解器进行微分，这会存储不断增长的计算轨迹；或者需要复杂的特定求解器的KKT敏感性推导。我们引入了一种用于正则化平滑接触的AD辅助隐式导数，并将其应用于基于隐式函数定理（IFT）的Mujoco MJX。该方法在容忍收敛解处对平稳残差进行微分，避免了求解器展开和手动组装KKT系统。IFT使得编译的临时内存在求解器努力下几乎保持不变，在从一次迭代到十次迭代时变化小于4$ ext{ extperthousand}$，而展开AD则增长了10.6$ imes$。在活跃接触和模型维度下，IFT内存增长较慢，在256个接触时使用的内存减少了20$ imes$，在16个接触和96个自由度时减少了6$ imes$。我们进一步引入了针对剩余模型预测控制的优化器蒸馏，将批量全时域iLQR摊销为指导短时域剩余iLQR的策略。在Finger、Franka和Unitree上，这使得六步成功率比标准iLQR提高了28-98个百分点。
cs.RO / 4 / 2607.25049
FIRMGrasp: A Friction-Informed Risk Margin for Robust Grasp Synthesis
FIRMGrasp：一种基于摩擦信息的鲁棒抓取合成风险边际
Abstract
Classical grasp quality metrics assume a single deterministic friction coefficient, so they cannot predict whether a grasp retains force closure across the range of friction values the contacting surfaces may exhibit. To predict these failures, we present FIRMGrasp, a family of friction-volatility-aware grasp quality metrics grounded in the Conditional Value-at-Risk (CVaR) risk measure. Unlike standard grasp quality assessors that assume a single friction realization, our metric evaluates the force-closure margin at the CVaR-discounted mean of the adverse friction tail, yielding a risk-adjusted margin $\varepsilon^{(\beta)}$, the inscribed-ball radius of the risk-adjusted wrench space. We establish its monotonicity in the confidence level $\beta$, its differentiability in the grasp parameters, and a probabilistic closure certificate that guarantees force closure with probability at least $\beta$ whenever $\varepsilon^{(\beta)}$ is positive. Under a calibrated friction distribution, analytic evaluation shows our $\varepsilon^{(\beta)}$ metric identifies friction-sensitive grasps that the nominal Ferrari-Canny epsilon rates as high-quality, and we compare against the nominal epsilon and recent differentiable baselines. Across 1,599 LEAP Hand and Allegro Hand grasps, 53% of the grasps the nominal Ferrari-Canny margin certifies lose force closure in the adverse friction tail. On the same set, the nominal margin separates realized shake and pick success with probabilities of only 0.53 and 0.67, near chance on shake success, whereas $\varepsilon^{(\beta)}$ orders the pair correctly with probabilities of 0.63 and 0.78, respectively. In simulated lift trials with gravity enabled at an adverse friction coefficient of 0.2, grasps $\varepsilon^{(\beta)}$ certifies reach a 70% success rate under lateral pull, against 25% for grasps the nominal margin certifies but $\varepsilon^{(\beta)}$ rejects.
Chinese Translation
经典的抓取质量指标假设单一的确定性摩擦系数，因此无法预测抓取在接触表面可能表现出的摩擦值范围内是否保持力闭合。为了预测这些失败，我们提出了FIRMGrasp，这是一系列基于条件风险价值（Conditional Value-at-Risk, CVaR）风险度量的摩擦波动感知抓取质量指标。与假设单一摩擦实现的标准抓取质量评估方法不同，我们的指标在不利摩擦尾部的CVaR折扣均值下评估力闭合边际，得出风险调整边际$eta^{(eta)}$，即风险调整扭矩空间的内切球半径。我们建立了其在置信水平$eta$上的单调性、在抓取参数上的可微性，以及一个概率闭合证明，确保每当$eta^{(eta)}$为正时，力闭合的概率至少为$eta$。在经过校准的摩擦分布下，解析评估显示我们的$eta^{(eta)}$指标能够识别出摩擦敏感的抓取，而名义上的Ferrari-Canny边际则将其评定为高质量抓取，我们还与名义边际和最近的可微基线进行了比较。在1,599个LEAP手和Allegro手的抓取中，名义Ferrari-Canny边际认证的53%的抓取在不利摩擦尾部失去了力闭合。在同一组中，名义边际对实现的摇晃和抓取成功的概率仅为0.53和0.67，摇晃成功的概率接近随机，而$eta^{(eta)}$则正确地将这对的概率排序为0.63和0.78。在模拟的提升试验中，在不利摩擦系数为0.2的情况下，$eta^{(eta)}$认证的抓取在侧向拉力下达到了70%的成功率，而名义边际认证的抓取成功率仅为25%，但被$eta^{(eta)}$拒绝。
cs.RO / 5 / 2607.25053
Motion Generation With Environmental Constraints
具有环境约束的运动生成
Abstract
Robot motion planning faces challenges in high-dimensional spaces and uncertain environments, often constrained by the need for collision-free motions. We advocate an alternative approach, Environmental Constraint Exploitation (ECE), where deliberate contact with the environment simplifies planning by reducing dimensionality and computational complexity. By integrating ECE into motion planning algorithms, we bias exploration to task-relevant regions and leverage contact for uncertainty reduction to improve robustness during execution. We evaluate ECE benefits with RRT-based planners and demonstrate their practical benefits in a real-world application. This work consolidates and extends prior research, showcasing how ECE simplifies motion planning while enhancing adaptability and performance in complex environments.
Chinese Translation
机器人运动规划在高维空间和不确定环境中面临挑战，通常受到无碰撞运动需求的限制。我们倡导一种替代方法，即环境约束利用（Environmental Constraint Exploitation, ECE），通过与环境的故意接触来简化规划，从而降低维度和计算复杂性。通过将ECE整合到运动规划算法中，我们将探索偏向于与任务相关的区域，并利用接触来减少不确定性，从而提高执行过程中的鲁棒性。我们使用基于RRT的规划器评估ECE的优势，并在实际应用中展示其实际益处。本研究整合并扩展了先前的研究，展示了ECE如何简化运动规划，同时增强在复杂环境中的适应性和性能。
cs.RO / 6 / 2607.25056
Hybrid Artificial Potential Fields and Spatio-Temporal Transformers for Real-Time AUV Path Planning
混合人工势场与时空变换器用于实时自主水下航行器路径规划
Abstract
Autonomous Underwater Vehicles (AUVs) operate in complex, unstructured environments where efficient and safe path planning is critical for mission success and energy conservation. This paper presents a comprehensive comparative evaluation of thirteen path planning algorithms, ranging from classical graph-search methods (A*, Dijkstra) and sampling-based approaches (RRT*) to metaheuristics (PSO, GA, ACO, BCO) and learning-based architectures. Special emphasis is placed on a proposed hybrid approach combining Artificial Potential Fields (APF) with a Spatio-Temporal (ST) Transformer. Evaluated across five navigation scenarios on high-resolution underwater terrain maps, all algorithms achieved 100\% task completion; however, significant trade-offs emerged in path optimality, collision avoidance, and computational load. The Hybrid APF + ST-Transformer demonstrated superior balanced performance, achieving the shortest average path length (943.15 units), a low collision rate (0.031), and efficient computation time (0.96 s), outperforming standalone learning models, which required fallback mechanisms and classical methods that incurred higher latency. While classical algorithms guaranteed collision-free paths, their excessive path lengths and processing times render them less suitable for dynamic underwater operations. Conversely, metaheuristic approaches introduced trajectory complexity unsuitable for strict energy constraints. Based on these findings, the Hybrid APF + ST framework is recommended as a principal approach for real-time AUV navigation, offering a robust solution that harmonizes reactive obstacle avoidance with global path optimality in resource-constrained underwater systems.
Chinese Translation
自主水下航行器（AUV）在复杂且无结构的环境中操作，高效和安全的路径规划对于任务成功和能量节约至关重要。本文对十三种路径规划算法进行了全面的比较评估，这些算法包括经典的图搜索方法（A*、Dijkstra）、基于采样的方法（RRT*）、元启发式算法（PSO、GA、ACO、BCO）以及基于学习的架构。特别强调了一种将人工势场（APF）与时空（ST）变换器相结合的混合方法。在高分辨率水下地形图上的五种导航场景中进行评估，所有算法均实现了100%的任务完成率；然而，在路径最优性、碰撞避免和计算负载方面出现了显著的权衡。混合APF + ST-变换器表现出优越的平衡性能，达到了最短的平均路径长度（943.15单位）、低碰撞率（0.031）和高效的计算时间（0.96秒），优于需要后备机制的独立学习模型，以及导致更高延迟的经典方法。虽然经典算法保证了无碰撞路径，但其过长的路径长度和处理时间使其不太适合动态水下操作。相反，元启发式方法引入了不适合严格能量限制的轨迹复杂性。基于这些发现，推荐将混合APF + ST框架作为实时AUV导航的主要方法，提供了一种强健的解决方案，能够在资源受限的水下系统中协调反应性障碍避免与全局路径最优性。
cs.RO / 7 / 2607.25071
Input Shaping for Point-to-Point Motion with a Continuum Robot Arm
用于连续机器人臂点对点运动的输入整形
Abstract
A cable-driven continuum robot arm is an underactuated mechanism and may suffer residual vibration at the end of a rest-to-rest maneuver. In this work, a time-delay filter is applied as an input shaper to the system to eliminate the excitation of vibratory modes. A non-robust and a robust time-delay filter are designed based on a linear system model and demonstrate improved response compared to a velocity-driven pulse input. Experimental results using the continuum robot validate the application of the input shaper, with reduced overshoot and settling time exemplifying the reduction in oscillation at the end of the maneuver. It is also shown that utilizing the robust shaper further improves the response of the arm in comparison to applying the non-robust shaper. These results are significant towards the precise and robust implementation of continuum robots in applications involving arbitrary end-effector trajectories.
Chinese Translation
一种由电缆驱动的连续机器人臂是一种欠驱动机制，在静止到静止的操作过程中可能会出现残余振动。在本研究中，应用时间延迟滤波器作为输入整形器，以消除振动模式的激励。基于线性系统模型设计了非鲁棒和鲁棒时间延迟滤波器，并展示了与速度驱动脉冲输入相比的改进响应。使用连续机器人进行的实验结果验证了输入整形器的应用，减少的超调和稳定时间表明在操作结束时振荡的减少。还显示，使用鲁棒整形器相比于应用非鲁棒整形器进一步改善了机器臂的响应。这些结果对于在涉及任意末端执行器轨迹的应用中精确和鲁棒地实施连续机器人具有重要意义。
cs.RO / 8 / 2607.25138
Reactive 3D Motion Planning for a Franka Arm via Star-World Workspace Reshaping
通过星世界工作空间重塑实现Franka臂的反应式三维运动规划
Abstract
Safety inflation can cause nearby obstacles to overlap, violating the disjoint-obstacle assumptions used by many modulation-based reactive planners. We investigate Star-World workspace reshaping for three-dimensional reactive control of a Franka Emika Panda manipulator. At each update, intersecting inflated obstacles are clustered and replaced by star-shaped proxies before a dynamical-system-based end-effector controller is evaluated. A null-space artificial-potential-field term provides complementary arm-body avoidance. We compare reshaped and unreshaped obstacle representations in six PyBullet scenarios using goal attainment, path-length ratio, and computation time. In this preliminary 12-trial evaluation, reshaping reaches the goal in five of six scenarios, compared with four of six for the unreshaped baseline. It resolves the canonical overlapping-wall case and requires 0.68--8.70\,ms per workspace update for scenes containing one to seven obstacles. However, it also increases path length, produces near-equilibria in two cases, and closes a navigable corridor through over-aggressive merging. These results show both the promise and the practical limitations of transferring Star-World guarantees from workspace geometry to a redundant manipulator controlled through inverse kinematics.
Chinese Translation
安全膨胀可能导致附近障碍物重叠，从而违反许多基于调制的反应式规划器所使用的障碍物不重叠假设。我们研究了星世界工作空间重塑在Franka Emika Panda机械臂的三维反应式控制中的应用。在每次更新时，交叉的膨胀障碍物被聚类并由星形代理替代，然后评估基于动态系统的末端执行器控制器。一个零空间人工势场项提供了补充的臂体避让。我们在六个PyBullet场景中比较了重塑和未重塑的障碍物表示，使用目标达成率、路径长度比和计算时间。在这项初步的12次试验评估中，重塑在六个场景中有五个达成目标，而未重塑基线则为四个。它解决了经典的重叠墙问题，并且在包含一个到七个障碍物的场景中，每次工作空间更新需要0.68到8.70毫秒。然而，它也增加了路径长度，在两个案例中产生了近平衡状态，并通过过于激进的合并关闭了可导航的走廊。这些结果展示了将星世界保证从工作空间几何转移到通过逆向运动学控制的冗余机械臂的潜力和实际限制。
cs.RO / 9 / 2607.25195
Decentralized Scalable Exploration via Emergent Adaptive L\'evy Walks on Minimal-Sensing Platforms
基于最小感知平台的自适应涌现莱维游走的去中心化可扩展探索
Abstract
Efficient autonomous exploration with palm-sized nano-UAVs remains challenging due to severe limitations in sensing, computation, and flight endurance. We present a lightweight sensor-driven L\'evy walk (SDLW) controller for aerial robots weighing under 50 grams and equipped with sparse local sensing. The method combines discrete L\'evy step-length sampling with a sensor-reactive heading policy using directional range measurements. Each robot independently samples its L\'evy exponent from a uniform prior to diversify exploration without inter-robot communication for exploration control. Each robot then selects headings using a von Mises distribution that biases motion toward open directions while preserving superdiffusive exploration properties. The controller operates at constant computational cost, enabling scalable multi-UAV exploration. Simulation results show coverage improvements of 79.6% in open arenas, 43.1% in rooms-and-corridors layouts, and 13.6% in cluttered environments, with collision reductions of 13.0%, 7.1%, and 1.4%, respectively, relative to a uniform-heading L\'evy walk baseline. This work provides a practical framework for scalable multi-robot exploration on minimal-sensing, resource-constrained nano-UAVs.
Chinese Translation
由于感知、计算和飞行耐力的严重限制，使用掌中宝大小的纳米无人机进行高效自主探索仍然具有挑战性。我们提出了一种轻量级传感器驱动的莱维游走（SDLW）控制器，适用于重量在50克以下并配备稀疏本地感知的空中机器人。该方法结合了离散的莱维步长采样和基于方向范围测量的传感器反应航向策略。每个机器人独立从均匀先验中采样其莱维指数，以多样化探索，而无需进行机器人间的通信来控制探索。然后，每个机器人使用冯·米塞斯分布选择航向，该分布偏向于开放方向的运动，同时保持超扩散探索特性。该控制器以恒定的计算成本运行，使得多无人机探索具备可扩展性。仿真结果显示，在开放场地中的覆盖率提高了79.6%，在房间和走廊布局中提高了43.1%，在杂乱环境中提高了13.6%；与均匀航向的莱维游走基线相比，碰撞减少了13.0%、7.1%和1.4%。本研究为在最小感知、资源受限的纳米无人机上进行可扩展的多机器人探索提供了一个实用框架。
cs.RO / 10 / 2607.25215
Leveraging Semantic Maps for City-Scale Cross-View Localization
利用语义地图进行城市规模的跨视角定位
Abstract
We want robots to localize in previously untraversed environments against commonly available prior data. Rich semantic data available from OpenStreetMap can be useful in this task. However, existing methods either ignore this semantic information, directly matching panoramas and overhead imagery, or dramatically compress the semantic information, working with a small set of fixed classes. To leverage this rich semantic information, two challenges need to be overcome. First, useful semantic information needs to be extracted from the robot's egocentric observations. Second, the observed information must be quickly associated with the large prior semantic map (e.g., up to 628 km^2). We show that VLMs are effective at both extracting relevant landmarks from panoramas, and identifying feasible correspondences between these landmarks and prior overhead landmarks. However, using VLMs to propose all correspondences scales poorly as the number of mapped landmarks increases. Instead, we propose distilling a lightweight matcher from a VLM which computes correspondences for all entities in a map. We use this output to form an observation likelihood which is fused over time with a Bayes filter to create a time series of pose estimates. To support further investigation into generalizable cross-view methods that leverage semantic information, we release a dataset of extracted semantics and evaluation trajectories spanning eleven environments, including panoramas we collected in a snowstorm and at night in Boston. We demonstrate our method, trained on a single city's fair-weather data, generalizes across location, lighting, weather, and other challenges. Code and datasets are available at https://efahnestock.github.io/loci/.
Chinese Translation
我们希望机器人能够在以前未经过的环境中，根据常见的先验数据进行定位。来自OpenStreetMap的丰富语义数据在这一任务中可能非常有用。然而，现有的方法要么忽略这些语义信息，直接匹配全景图和航拍图像，要么大幅压缩语义信息，仅处理一小组固定类别。为了利用这些丰富的语义信息，需要克服两个挑战。首先，需要从机器人的自我中心观察中提取有用的语义信息。其次，观察到的信息必须迅速与大型先验语义地图（例如，面积可达628平方公里）进行关联。我们展示了VLM（视觉语言模型）在从全景图中提取相关地标以及识别这些地标与先前航拍地标之间的可行对应关系方面的有效性。然而，随着映射地标数量的增加，使用VLM来提出所有对应关系的效率较低。因此，我们提出从VLM中提炼出一个轻量级匹配器，该匹配器为地图中的所有实体计算对应关系。我们使用该输出形成一个观察似然性，并通过贝叶斯滤波器随时间融合，以创建一系列姿态估计。为了支持对利用语义信息的可推广跨视角方法的进一步研究，我们发布了一个包含提取语义和评估轨迹的数据集，涵盖了包括我们在波士顿的暴风雪和夜间收集的全景图在内的十一种环境。我们展示了我们的方法在单一城市的良好天气数据上训练后，能够跨越位置、光照、天气和其他挑战进行推广。代码和数据集可在 https://efahnestock.github.io/loci/ 获取。
cs.RO / 11 / 2607.25219
SONG: A Photorealistic 3D Gaussian Simulation Platform for Benchmarking Social Navigation
SONG：用于社交导航基准测试的逼真3D高斯模拟平台
Abstract
Social navigation has progressed from simplified 2D environments toward a more general vision-based setting, in which a robot needs to achieve socially compliant behavior purely from onboard visual observations. Yet supporting simulation platforms have not kept pace: existing options either lack visual observations, lack moving human avatars, or fall short of real-world fidelity in appearance and pedestrian behavior, offering limited support for advancing vision-based social navigation. We introduce SONG, a SOcial Navigation platform powered by 3D Gaussian splatting (3DGS). It leverages 3DGS for both scene and avatar representations, drives pedestrians using semantically grounded trajectories generated by a large language model, and synthesizes their full-body motion with a trajectory-conditioned generator to produce continuous, natural movement. On top of the platform, we curate SONG-Bench, a set of evaluation episodes stratified by difficulty, and propose a multi-dimensional metric suite covering effectiveness, safety, and social compliance. A systematic evaluation of representative navigation baselines reveals three findings: (a) vision-based social navigation is far from solved; (b) a critical safety deficit precedes social etiquette; (c) real-world data matters more than model scale. Crucially, we demonstrate that fine-tuning on our curated data effectively improves the success rate in real-world environments. We hope our platform provides a faithful and rigorous testbed for the next generation of vision-based social navigation research.
Chinese Translation
社交导航已经从简化的2D环境发展到更为通用的基于视觉的设置，在这种设置中，机器人需要仅通过机载视觉观测实现符合社会规范的行为。然而，支持的模拟平台并未跟上这一进展：现有选项要么缺乏视觉观测，要么缺乏移动的人类化身，或者在外观和行人行为的真实感上不足，限制了基于视觉的社交导航的进步。我们介绍了SONG，一个由3D高斯点云（3D Gaussian splatting, 3DGS）驱动的社交导航平台。它利用3DGS进行场景和化身的表示，通过大型语言模型生成的语义基础轨迹驱动行人，并使用轨迹条件生成器合成他们的全身运动，以产生连续、自然的移动。在此平台之上，我们策划了SONG-Bench，一个按难度分层的评估集，并提出了一套涵盖有效性、安全性和社会合规性的多维度指标。对代表性导航基线的系统评估揭示了三个发现：（a）基于视觉的社交导航远未解决；（b）社会礼仪之前存在严重的安全缺失；（c）真实世界数据比模型规模更为重要。关键是，我们证明在我们策划的数据上进行微调能有效提高在真实环境中的成功率。我们希望我们的平台为下一代基于视觉的社交导航研究提供一个真实而严谨的测试平台。
cs.RO / 12 / 2607.25327
Belief-Aware Influence and Trust (BAIT): Shaping Human Belief During Repeated Human-Robot Interaction
信念感知影响与信任（BAIT）：在重复的人机交互中塑造人类信念
Abstract
Repeated human-robot interaction (HRI) requires proactively accounting for humans who continually adapt to evolving beliefs about the robot. Prior frameworks often treat encounters as isolated events, suffering cumulative task performance decay as human perception drifts, or maintain long-term influence through erratic, unpredictable behavior that erodes perceived human trust and relies on computationally unscalable formulations. To address these gaps, we introduce the Belief- Aware Influence and Trust (BAIT) controller. BAIT integrates a hierarchical particle filter, which infers both fast human strategic shifts and slow perceptual belief updates, with a belief-aware Model Predictive Path Integral planner. BAIT explicitly optimizes the trade-off between long-horizon influence and human trust, while enforcing immediate task performance as a strict constraint. Across simulations, a human-subject study, and a real-world GEM vehicle deployments in repeated lane-merging scenarios, BAIT achieves task performance comparable to baselines that optimize long-term influence through unpredictability while yielding significantly higher user trust. The video demonstrating our experiments is available at https://youtu.be/GsPfHRujzVs.
Chinese Translation
重复的人机交互（HRI）需要主动考虑人类如何不断适应对机器人的演变信念。以往的框架通常将交互视为孤立事件，导致人类感知漂移时任务表现逐渐下降，或者通过不稳定、不可预测的行为维持长期影响，这种行为侵蚀了人类的信任感，并依赖于计算上不可扩展的公式。为了解决这些问题，我们提出了信念感知影响与信任（BAIT）控制器。BAIT集成了一个层次粒子滤波器，该滤波器同时推断快速的人类战略转变和缓慢的感知信念更新，并结合了信念感知的模型预测路径积分规划器。BAIT明确优化了长期影响与人类信任之间的权衡，同时将即时任务表现作为严格约束。在模拟实验、人类受试者研究以及在重复车道合并场景中的实际GEM车辆部署中，BAIT实现了与通过不可预测性优化长期影响的基线相当的任务表现，同时显著提高了用户信任度。展示我们实验的视频可在 https://youtu.be/GsPfHRujzVs 获取。
cs.RO / 13 / 2607.25388
SGTP: Sampling-based Game-Theoretic Planning for Real-Time Multi-Vehicle Autonomous Racing
SGTP：基于采样的博弈论规划用于实时多车辆自主赛车
Abstract
Autonomous multi-vehicle racing requires real-time planning of diverse competitive behaviors in intense interactions. Existing planners often struggle to balance strategic diversity and computational efficiency. To address this challenge, we propose Sampling-based Game-Theoretic Planning (SGTP), a real-time framework that combines game-theoretic reasoning with GPU-accelerated sampling of control sequences and dynamics rollouts. Sampled trajectories are ranked using a game-aware cost to capture competitive interactions and generate diverse racing behaviors. Our planner then performs feasibility selection by explicitly enforcing track-boundary and dynamic collision-avoidance constraints, ensuring safe and reliable transitions between racing strategies. Extensive simulations on challenging tracks show that SGTP achieves a 95.24% win rate and a 99.35% task-completion ratio in highly interactive races, with a mean computational time of 0.095 s over multiple iterative solving steps. We also demonstrate the successful application of SGTP in large-scale scenarios with up to 10 agents. We release our code and provide an open-source benchmark of multi-agent autonomous racing algorithms to facilitate future research. Project page: https://sgtp-racing.github.io/.
Chinese Translation
自主多车辆赛车需要在激烈的互动中实时规划多样化的竞争行为。现有的规划器往往难以平衡战略多样性和计算效率。为了解决这一挑战，我们提出了基于采样的博弈论规划（SGTP），这是一个实时框架，结合了博弈论推理与GPU加速的控制序列和动态展开的采样。通过使用博弈感知成本对采样轨迹进行排名，以捕捉竞争互动并生成多样化的赛车行为。我们的规划器随后通过明确施加赛道边界和动态避碰约束来执行可行性选择，确保赛车策略之间的安全可靠过渡。在具有挑战性的赛道上进行的大量仿真表明，SGTP在高度互动的比赛中实现了95.24%的胜率和99.35%的任务完成率，多个迭代求解步骤的平均计算时间为0.095秒。我们还展示了SGTP在多达10个智能体的大规模场景中的成功应用。我们发布了代码，并提供了一个开源的多智能体自主赛车算法基准，以促进未来的研究。项目页面：https://sgtp-racing.github.io/
cs.RO / 14 / 2607.25397
Decompose and Reorganize: Planning with Primitives and Visuomotor Policies Learned from Demonstrations
分解与重组：基于示范学习的原语与视觉运动策略规划
Abstract
Successfully automating dexterous, long-horizon robotic manipulation requires frameworks capable of both high-level reasoning and fine-grained execution. Traditional task and motion planning (TAMP), while excellent at symbolic planning, is often brittle in contact-rich operations. Simultaneously, imitation learning (IL), while effective in manipulation tasks with visual feedback, is limited by its low capability in spatial generalization and multi-stage operation. To reconcile their complementary strengths and limitations, we propose DR-LfD (Decomposed and Reorganized Skills Learned from Demonstrations), a framework that seamlessly integrates visuomotor policies into a TAMP-gated decision-making system. Based on contact relationships, DR-LfD decomposes human demonstrations into atomic skills, which are reproduced as visuomotor policies or object-centric primitives. The initiation, termination, and constraints of the visuomotor policies are carefully modeled and implemented in a TAMP-compatible form, enabling reorganization of skills learned from different sources. DR-LfD transforms the learning problem from one requiring exponential demonstration data over possible skill sequences to one whose demonstration burden scales with the number of distinct skill types, with limited data for each skill. Through comprehensive real-world and simulation benchmarking across diverse scenarios, we demonstrate the strong performance of DR-LfD on tasks involving multiple steps, unseen setups, and physical constraints. Project website: https://dr-lfd.github.io/DR-LfD-website.
Chinese Translation
成功实现灵巧的长时间机器人操作自动化需要能够进行高层次推理和细粒度执行的框架。传统的任务与运动规划（TAMP）在符号规划方面表现出色，但在接触丰富的操作中往往显得脆弱。同时，模仿学习（IL）在具有视觉反馈的操作任务中有效，但在空间泛化和多阶段操作方面能力有限。为了调和它们的互补优势和局限性，我们提出了DR-LfD（从示范中学习的分解与重组技能）框架，该框架将视觉运动策略无缝集成到一个TAMP门控的决策系统中。基于接触关系，DR-LfD将人类示范分解为原子技能，这些技能被再现为视觉运动策略或以物体为中心的原语。视觉运动策略的启动、终止和约束被仔细建模并以TAMP兼容的形式实现，从而能够重组来自不同来源的学习技能。DR-LfD将学习问题从需要对可能技能序列进行指数级示范数据的要求，转变为示范负担与不同技能类型的数量成比例的形式，每种技能的数据量有限。通过在多种场景下进行全面的现实世界和仿真基准测试，我们展示了DR-LfD在涉及多个步骤、未见设置和物理约束的任务中的强大性能。项目网站：https://dr-lfd.github.io/DR-LfD-website。
cs.RO / 15 / 2607.25448
Room-Mediated Co-occurrence for Zero-Shot Object-Centric Semantic Navigation via Frontier Scoring
通过边界评分实现零样本对象中心语义导航的房间介导共现
Abstract
Zero-shot ObjectNav methods increasingly use vision-language priors, but direct object-object similarity in the latent space is often a weak proxy for spatial co-occurrence. We present an analytical, training-free semantic navigation pipeline that mediates object relationships through a compact room lexicon. Each object label is mapped to a CLIP-derived Room Probability Vector (RPV), and object-target co-occurrence is computed from RPV distribution overlap. These scores are projected onto a value map using geodesic flood-fill propagation (Fast Marching Method), with adaptive signal decay, and are used to rank frontiers by semantic score for navigation. Together, these components form an integrated, training-free, object-centric pipeline for open-vocabulary zero-shot navigation. Results show that our object-centric approach improves Success Rate (SR) and Success by weighted inverse Path Length (SPL) by a relative 3% and 1.3%, respectively, compared to image-holistic baselines on the HM3D dataset validation split, while preserving interpretability and open-vocabulary flexibility. Code is available at: uts-ri.github.io/RPV-SemNav.
Chinese Translation
零样本对象导航（Zero-shot ObjectNav）方法越来越多地使用视觉-语言先验，但潜在空间中对象之间的直接相似性通常是空间共现的一个较弱代理。我们提出了一种分析性的、无训练的语义导航管道，通过紧凑的房间词汇中介对象关系。每个对象标签被映射到一个基于CLIP的房间概率向量（Room Probability Vector, RPV），并通过RPV分布重叠计算对象与目标的共现。这些评分通过测地线洪泛传播（Fast Marching Method）投影到值图上，并采用自适应信号衰减，用于根据语义评分对边界进行排名以实现导航。这些组件共同构成了一个集成的、无训练的、以对象为中心的开放词汇零样本导航管道。结果表明，与HM3D数据集验证集上的图像整体基线相比，我们的对象中心方法在成功率（Success Rate, SR）和加权逆路径长度成功率（Success by weighted inverse Path Length, SPL）上分别提高了3%和1.3%，同时保持了可解释性和开放词汇的灵活性。代码可在：uts-ri.github.io/RPV-SemNav获取。
cs.RO / 16 / 2607.25516
A Causality-aware Infer-diagnose-refine Framework for Test-time Modality Adaptation in VLA Models
一种因果感知的推理-诊断-精炼框架用于VLA模型的测试时模态适应
Abstract
Vision-language-action (VLA) models predict sequential actions to execute tasks specified by language instructions, conditioned on visual observations and proprioceptive states. However, how to fuse modalities in VLA models remains an open problem, since robot manipulation involves dynamic phases, such as long-distance movements and close-range interactions, in which the importance of visual observations may vary over time. In this paper, we propose an infer-diagnose-refine (IDR) framework, a model-agnostic framework that can be integrated with diverse VLA architectures for refining action predictions at test time. IDR first infers actions under factual and counterfactual scenarios of visual observations, and then diagnoses the causal effects of visual observations as the estimated dynamic importance, which is finally used to refine the action predictions in a training-free manner. We further design a causality-aware action refiner to realize the IDR framework, including zero-padding interventions for inferring counterfactual actions, norm-based quantification for diagnosing causal effects, and gated residual fusion for refining actions. Extensive experiments on both simulation benchmarks and real-world tasks show improvements in overall performance across multiple VLA backbones, demonstrating the efficacy of dynamically adjusting visual importance at test time.
Chinese Translation
视觉-语言-行动（VLA）模型预测顺序动作以执行由语言指令指定的任务，基于视觉观察和本体状态。然而，如何在VLA模型中融合模态仍然是一个未解决的问题，因为机器人操控涉及动态阶段，例如远距离移动和近距离交互，在这些阶段中视觉观察的重要性可能随时间变化。在本文中，我们提出了一种推理-诊断-精炼（IDR）框架，这是一种与模型无关的框架，可以与多种VLA架构集成，以在测试时精炼动作预测。IDR首先在视觉观察的事实和反事实场景下推断动作，然后诊断视觉观察的因果效应作为估计的动态重要性，最后以无训练的方式用于精炼动作预测。我们进一步设计了一种因果感知的动作精炼器，以实现IDR框架，包括用于推断反事实动作的零填充干预、用于诊断因果效应的基于范数的量化，以及用于精炼动作的门控残差融合。在模拟基准和真实任务上的大量实验表明，在多个VLA骨干网络中整体性能有所提升，证明了在测试时动态调整视觉重要性的有效性。
cs.RO / 17 / 2607.25541
P3: Probabilistic Policy Propagation for Stable VAE-Based Robot Learning
P3：基于概率策略传播的稳定变分自编码器机器人学习
Abstract
Variational Autoencoders are widely used to encode high-dimensional and noisy observations in robotics. However, their stochastic latent creates a mismatch with Proximal Policy Optimization (PPO): an effective policy marginalizes over the latent distribution, whereas former implementations estimate its probability ratio and KL divergence using only one latent sample. We identify a fundamental but overlooked theoretical cause: naive single-sample approximations in stochastic latent space induce significant variance and bias in the surrogate loss. To address this, we introduce P^3 (Probabilistic Policy Propagation), a distribution-aware optimization framework for VAE-based policies. $P^3$ couples moment-based probabilistic method for stable and efficient learning with sampling-based calibration for robust policy behavior under latent uncertainty. In our experiments, P^3 boosts data efficiency from 64.6% to >96%, reduces convergence steps by >20%. Furthermore, P^3 is evaluated on challenging humanoid parkour tasks and shows an effective foundation for VAE-based PPO. Code is available at https://github.com/ylyem9x/P3_Open.
Chinese Translation
变分自编码器广泛用于编码机器人中的高维和噪声观察。然而，它们的随机潜变量与近端策略优化（Proximal Policy Optimization, PPO）之间存在不匹配：有效的策略在潜在分布上进行边际化，而早期实现仅使用一个潜在样本来估计其概率比和KL散度。我们识别出一个基本但被忽视的理论原因：随机潜在空间中的天真单样本近似会在代理损失中引入显著的方差和偏差。为了解决这个问题，我们引入了P^3（概率策略传播），这是一个针对基于变分自编码器政策的分布感知优化框架。$P^3$将基于矩的概率方法与基于采样的校准相结合，以实现稳定和高效的学习，并在潜在不确定性下实现稳健的策略行为。在我们的实验中，P^3将数据效率从64.6%提升至>96%，并将收敛步骤减少了>20%。此外，P^3在具有挑战性的人形跑酷任务上进行了评估，显示出基于变分自编码器的PPO的有效基础。代码可在 https://github.com/ylyem9x/P3_Open 获取。
cs.RO / 18 / 2607.25593
When Does Legacy Data Start to Help? Emergent Transfer in Cross-Configuration Robot Learning
遗留数据何时开始发挥作用？跨配置机器人学习中的新兴转移
Abstract
Robotic hardware evolves over time, but demonstration data is often tied to a specific sensor and actuator configuration. This raises a practical and underexplored question: when does legacy data begin to benefit an upgraded robot? We study this question on a wheeled humanoid platform across two hardware generations, where both the camera and gripper are changed while the overall morphology remains fixed. Contrary to the common assumption that more cross-configuration data is always helpful, we observe a grokking-like transition: legacy data remains ineffective until the upgraded configuration acquires a minimum level of task competence, after which co-training gains rise sharply before diminishing near saturation. We hypothesize that this task-dependent transition is governed by a transfer threshold and characterize the resulting three-phase pattern. Across real-robot manipulation tasks, we observe all three phases: no measurable benefit at low competence ($10.0\% \rightarrow 10.0\%$), a sharp gain after crossing the threshold ($23.3\% \rightarrow 86.7\%$ on flower insertion), and diminishing returns at high competence ($85.0\% \rightarrow 93.3\%$ on pen insertion). We provide a theoretical account based on gradient alignment and residual policy uncertainty, and derive a phase-aware rule for deciding when to collect more new-hardware data and when to reuse legacy demonstrations. We further validate this three-phase pattern on a mobile dual-arm watering task, with results consistent with our predictions.
Chinese Translation
机器人硬件随着时间的推移而演变，但演示数据通常与特定的传感器和执行器配置相关。这引发了一个实际且未被充分探索的问题：遗留数据何时开始惠及升级后的机器人？我们在一个轮式类人平台上研究这个问题，涉及两个硬件代际，其中相机和夹具均发生变化，而整体形态保持不变。与普遍认为的跨配置数据总是有帮助的假设相反，我们观察到一种类似于理解的过渡：遗留数据在升级配置获得最低任务能力之前仍然无效，此后共同训练的收益急剧上升，然后在接近饱和时减小。我们假设这种依赖任务的过渡受转移阈值的控制，并描述了由此产生的三阶段模式。在真实机器人操作任务中，我们观察到所有三个阶段：在低能力下没有可测量的收益（$10.0
ightarrow 10.0$），在跨越阈值后收益急剧增加（花卉插入任务中从$23.3
ightarrow 86.7$），以及在高能力下收益递减（笔插入任务中从$85.0
ightarrow 93.3$）。我们提供了一个基于梯度对齐和残差策略不确定性的理论解释，并推导出一个阶段感知规则，用于决定何时收集更多的新硬件数据以及何时重用遗留演示。我们进一步在一个移动双臂浇水任务中验证了这一三阶段模式，结果与我们的预测一致。
cs.RO / 19 / 2607.25728
Shared Voxel-Map-Based Cooperative Indoor UAV Guidance with a Multi-Agent Soft Actor-Critic Controller
基于共享体素地图的多智能体软演员-评论者控制的室内无人机协作引导
Abstract
This paper presents a cooperative indoor UAV guidance framework that combines a shared voxel-map world model with a multi-agent Soft Actor-Critic (MASAC) controller. Multiple drones fuse 360 LiDAR observations into a common world-frame occupancy map, which is converted into a compact bird's-eye-view (BEV) representation and provided to each agent as an ego-aligned local crop. This integrate-in-world, act-in- ego design enables consistent multi-UAV spatial fusion whilst retaining decentralised continuous control. The policy combines BEV map features, near-field obstacle observations, and compact goal and peer-state information within a centralised-training, decentralised-execution framework. In simulation, the learned controller achieves a 90.3% success rate in corridor navigation, outperforming Astar planning, an artificial potential field controller, and a prior guidance method. To address residual sim-to-real mismatch, the simulation-trained policy is further adapted using offline imitation fine-tuning from real-world data. Real-world experiments in GNSS-denied indoor environments demonstrate stable two-UAV cooperative operation across increasingly chal- lenging obstacle layouts. The results show that shared voxel-map representations provide an effective and scalable spatial substrate for learned cooperative indoor UAV guidance.
Chinese Translation
本文提出了一种协作室内无人机引导框架，该框架结合了共享体素地图世界模型和多智能体软演员-评论者（MASAC）控制器。多个无人机将360度激光雷达观测融合为一个共同的世界框架占用图，该图被转换为紧凑的鸟瞰图（BEV）表示，并作为自我对齐的局部裁剪提供给每个智能体。这种在世界中集成、在自我中行动的设计使得在保持去中心化连续控制的同时，实现一致的多无人机空间融合。该策略结合了BEV地图特征、近场障碍物观测以及紧凑的目标和同伴状态信息，采用集中训练、去中心化执行的框架。在仿真中，学习到的控制器在走廊导航中达到了90.3%的成功率，优于A*规划、人工势场控制器和先前的引导方法。为了解决残余的仿真与现实不匹配问题，仿真训练的策略进一步通过真实世界数据的离线模仿微调进行适应。在GNSS信号缺失的室内环境中的实际实验表明，在日益复杂的障碍布局中，两个无人机的协作操作稳定。结果表明，共享体素地图表示为学习的协作室内无人机引导提供了有效且可扩展的空间基础。
cs.RO / 20 / 2607.25731
Tri-Manual Visuomotor Imitation Learning of Robot Policies
三手视觉运动模仿学习的机器人策略
Abstract
Bimanual teleoperation provides an effective way to collect robot demonstrations, but it assumes that the operator and robot have matching numbers of simultaneous control channels. This assumption breaks for tri-manual systems: the robot can coordinate three arms concurrently, whereas a single operator can continuously control only two. Pairwise mode switching may therefore record otherwise independent motions sequentially, causing behaviour cloning to reproduce delays imposed by the interface rather than required by the task. We present TriManPolicy, a tri-manual imitation learning system that allows one operator to demonstrate behaviours for three arms. Its central component is Dependency-Aware Tri-Arm Scheduling (DATS). The key idea is to preserve the demonstrated arm motions while reconsidering when they occur. DATS retimes demonstrations offline by preserving local sensorimotor segments of fixed duration and repositioning them according to constraints on task order and arm usage that are reviewed by a human. The resulting data train a single synchronous policy for all three arms, while deployment requires neither the dependency graph nor the scheduler. Across six challenging tasks performed in the real world, policies trained on demonstrations retimed by DATS exhibit more efficient coordination while maintaining comparable observed task success. Offline analysis further shows that DATS changes the supervision across arms rather than merely removing idle periods.
Chinese Translation
双手远程操作提供了一种有效的方式来收集机器人演示，但它假设操作员和机器人具有匹配数量的同时控制通道。这一假设在三手系统中失效：机器人可以同时协调三只手臂，而单个操作员只能持续控制两只手臂。因此，成对的模式切换可能会将原本独立的动作顺序记录，从而导致行为克隆重现由接口施加的延迟，而非任务所需的延迟。我们提出了TriManPolicy，一个三手模仿学习系统，允许一个操作员为三只手臂演示行为。其核心组件是依赖感知三臂调度（Dependency-Aware Tri-Arm Scheduling, DATS）。其关键思想是在重新考虑动作发生的时间的同时，保留演示的手臂动作。DATS通过保留固定时长的局部传感器运动段并根据人类审查的任务顺序和手臂使用约束重新定位这些段来离线调整演示。生成的数据训练出一个适用于所有三只手臂的同步策略，而部署时则不需要依赖图或调度器。在现实世界中进行的六项具有挑战性的任务中，基于DATS重新调整的演示训练的策略展现出更高效的协调能力，同时保持了可比的任务成功率。离线分析进一步表明，DATS改变了手臂之间的监督，而不仅仅是去除闲置时间。
cs.RO / 21 / 2607.25754
Cooperative Multi-UAV Navigation in Complex Environments via Systematic Multi-Agent Deep Reinforcement Learning
通过系统化多智能体深度强化学习实现复杂环境中的多无人机协作导航
Abstract
Cooperative navigation of multi-agent UAVs in complex environments faces key challenges including local optima traps, sparse rewards, learning imbalance among agents, and insufficient cross-scenario generalisation. This paper proposes a multi-agent deep reinforcement learning framework that addresses these issues through coordinated exploration, demonstration exploitation, safe curriculum scheduling, and structure-aware generalisation. First, a perception mechanism combining memory of visited states, directional novelty estimates, and penalty backpropagation enables agents to proactively detect and escape local optima. Second, a hierarchical collaborative demonstration buffer with tiered behaviour cloning manages trajectories by degree of team collaboration and applies differential supervision to the actor network, improving demonstration utilisation under sparse collaborative signals. Third, a safety-aware dual-condition curriculum scheduling mechanism reviews mastered scenarios through back-testing and experience pre-filling during training, suppressing catastrophic forgetting while ensuring both task performance and flight safety. For generalisation, local geometric features computed from sensor readings are abstracted into a domain parameter, through which a structure-aware gating network and mixture-of-experts mechanism condition the policy on local structural patterns rather than scenario-specific coordinates, enabling cross-scenario transfer without exposure to the target environment. The framework is further validated under mixed static-dynamic obstacle settings, showing robust adaptability to dynamic disturbances. Simulation results confirm strong performance in collaboration success rate, navigation robustness, zero-shot cross-scenario generalisation, and dynamic environment adaptability.
Chinese Translation
多智能体无人机在复杂环境中的协作导航面临诸多关键挑战，包括局部最优陷阱、稀疏奖励、智能体之间的学习不平衡以及跨场景泛化不足。本文提出了一种多智能体深度强化学习框架，通过协调探索、示范利用、安全课程调度和结构感知泛化来解决这些问题。首先，结合访问状态记忆、方向新颖性估计和惩罚反向传播的感知机制，使智能体能够主动检测并逃离局部最优。其次，具有分层行为克隆的层级协作示范缓冲区根据团队协作程度管理轨迹，并对演员网络应用差异化监督，提高在稀疏协作信号下的示范利用率。第三，安全感知的双条件课程调度机制通过回测和训练期间的经验预填充审查掌握的场景，抑制灾难性遗忘，同时确保任务性能和飞行安全。为了实现泛化，从传感器读数计算的局部几何特征被抽象为领域参数，通过此参数，结构感知的门控网络和专家混合机制使策略基于局部结构模式而非特定场景坐标进行条件化，从而实现跨场景转移而无需接触目标环境。该框架在混合静态-动态障碍设置下进一步验证，显示出对动态干扰的强大适应性。仿真结果确认了在协作成功率、导航鲁棒性、零样本跨场景泛化和动态环境适应性方面的强劲表现。
cs.RO / 22 / 2607.25781
Tripody: An Overconstrained 3-SPR-like Parallel Robot for High-Reach Construction Tasks
Tripody：一种超约束的类3-SPR平行机器人用于高空施工任务
Abstract
Many ceiling construction tasks still rely on heavy serial manipulators that are difficult to deploy in cluttered interiors, motivating lightweight, field-ready alternatives that reach ceiling height while maintaining millimeter-level accuracy and the stiffness demanded by overhead tool loads. We introduce Tripody, a wheeled 3-DoF parallel robot for high-reach tasks that replaces the base spherical joints of a classical 3-SPR (3 legs; S: base spherical joint; P: actuated prismatic joint; R: end-effector revolute joint) morphology with universal joints, intentionally overconstraining the mechanism; small, distributed elastic deflections absorb the resulting incompatibilities, preserving predominantly translational motion. The 33kg system extends from 1.7m to 3.4m in height, supports a continuous 32kg payload, and offers a modular end-effector interface for ceiling operations. We detail the mechanical design - including custom linear actuators and a kinematic-compatibility analysis - and a control stack for accurate positioning that combines SE(3) state estimation, forward kinematics, and task-space control. In experiments, Tripody exhibits similar in-plane stiffness to a spherical-base variant but substantially higher torsional stiffness - an increase of 67% at 1.7m, 196% at 2.6m, and 454% at 3.4m - while maintaining negligible cross-axis coupling. Closed-loop positioning with a total station converges below 0.6mm across the entire workspace; pure model extrapolation achieves a 95th-percentile error of 2.7mm (max 3.6mm). Finally, we demonstrate task-level ceiling-drilling feasibility in an open-loop study by drilling a 15-hole pattern with 4.5mm maximum relative hole-position error after rigid alignment. These results support overconstrained, compliance-absorbing 3-SPR-like architectures as a practical path to lightweight, high- reach, millimeter-accurate construction robots.
Chinese Translation
许多天花板施工任务仍然依赖于难以在杂乱室内部署的重型串联机械手，这促使我们寻找轻量化、适合现场使用的替代方案，以在保持毫米级精度和满足悬挂工具负载所需刚度的同时达到天花板高度。我们介绍了Tripody，这是一种用于高空任务的轮式3自由度平行机器人，它用万向节替代了经典3-SPR（3条腿；S：基座球形关节；P：驱动的伸缩关节；R：末端执行器旋转关节）形态的基座球形关节，故意对机制进行超约束；小而分布的弹性变形吸收了由此产生的不兼容性，保持了主要的平移运动。该系统重33公斤，能够在1.7米到3.4米的高度范围内伸展，支持连续32公斤的有效载荷，并提供用于天花板操作的模块化末端执行器接口。我们详细介绍了机械设计，包括定制线性驱动器和运动学兼容性分析，以及用于精确定位的控制栈，结合了SE(3)状态估计、正向运动学和任务空间控制。在实验中，Tripody展现出与球形基座变体相似的平面内刚度，但扭转刚度显著更高——在1.7米时提高67%，在2.6米时提高196%，在3.4米时提高454%，同时保持微不足道的交叉轴耦合。使用全站仪的闭环定位在整个工作空间内收敛至0.6毫米以下；纯模型外推的95百分位误差为2.7毫米（最大3.6毫米）。最后，我们通过在刚性对齐后钻孔15个孔的模式，展示了任务级天花板钻孔的可行性，最大相对孔位误差为4.5毫米。这些结果支持超约束、吸收顺应性的类3-SPR架构作为轻量化、高空、毫米级精度施工机器人的实际路径。
cs.RO / 23 / 2607.25784
Motion-Acceleration Calibration and Compensation in IMUs without External Equipment for Attitude Estimation Filters
无需外部设备的惯性测量单元中的运动加速度校准与补偿用于姿态估计滤波器
Abstract
Attitude estimation based on inertial sensing requires measurements of local angular velocities and local gravity via gyroscopes and accelerometers. However, during the motion of a mobile system the inertial measurement unit (IMU) will be subject to additional accelerations which skews the measurement of local gravity. This effect gets amplified the further away the IMU is from the base of the system. Many attitude estimation filters, such as "Madgwick" or "Mahony", account for this by relying more on gyroscope integration for periods of high angular velocity. However, this approach is prone to accumulate long term error especially around the gravity vector. In this work we utilize the gyroscope measurements to compensate the additional accelerations induced by the motion of the system, i.e., centripetal- and tangential-accelerations. Additionally, we introduce a calibration method that estimates intrinsic IMU parameters such as axes misalignment, bias, scale, as well as the extrinsic base-to-IMU vector without the necessity for additional external equipment. Our evaluation in simulation as well as in the real-world shows that this method improves any attitude filter that relies on the direction of gravity. Furthermore we demonstrate the effectivenes on highly dynamic systems, and systems that are unable to put the IMU at the center of rotation, using our real-world spherical mobile mapping system.
Chinese Translation
基于惯性传感的姿态估计需要通过陀螺仪和加速度计测量局部角速度和局部重力。然而，在移动系统运动过程中，惯性测量单元（IMU）将受到额外加速度的影响，这会扭曲局部重力的测量。随着IMU离系统基座的距离增加，这种效应会被放大。许多姿态估计滤波器，如“Madgwick”或“Mahony”，通过在高角速度期间更多依赖陀螺仪积分来考虑这一点。然而，这种方法容易在重力向量附近积累长期误差。在本研究中，我们利用陀螺仪测量来补偿系统运动引起的额外加速度，即向心加速度和切向加速度。此外，我们提出了一种校准方法，可以在无需额外外部设备的情况下估计IMU的内在参数，如轴对齐、偏差、比例，以及外部基座到IMU的向量。我们在仿真和实际应用中的评估表明，该方法改善了任何依赖重力方向的姿态滤波器。此外，我们还展示了该方法在高度动态系统和无法将IMU放置在旋转中心的系统中的有效性，使用我们的现实世界球形移动测绘系统进行验证。
cs.RO / 24 / 2607.25798
Transformer Transformer: A Unified Model for Motion-Conditioned Robot Co-design
变换器变换器：一种统一的运动条件机器人协同设计模型
Abstract
An often overlooked factor of robot manipulation performance is the embodiment of the robot itself. Motivated by this problem, we study motion-conditioned robot co-design, where the goal is to generate complete robot designs that track target end-effector trajectories (from human demonstrations) while optimizing user-defined rewards. We introduce Transformer Transformer, a diffusion transformer trained on RoboTokens, a unified tokenization of robot embodiments, states, and actions. The same architecture can be used across embodiment spaces (e.g., wheeled bimanual, quadrupeds, humanoids) and use cases (embodiment generation, cross embodiment controller). Rather than overfitting to one reward function, Transformer Transformer is a dynamics model, whose reward-agnostic state and action predictions can be converted into reward-specific value predictions. These value predictions are used to steer embodiment diffusion towards high value robot designs, through a procedure we call Dynamics Self-Guidance. Experiments across multiple design spaces show zero-shot optimization of unseen rewards and trajectories, improving performance and runtime over the evolutionary baseline. Finally, we fabricated an optimized ALOHA design, which reduced tracking error by over 70% compared to the original design.
Chinese Translation
机器人操作性能中一个常被忽视的因素是机器人的具体体现。基于这一问题，我们研究了运动条件下的机器人协同设计，其目标是生成完整的机器人设计，以跟踪目标末端执行器轨迹（来自人类示范），同时优化用户定义的奖励。我们引入了变换器变换器（Transformer Transformer），这是一种在RoboTokens上训练的扩散变换器，RoboTokens是机器人体现、状态和动作的统一标记化。相同的架构可用于不同的体现空间（例如，轮式双手、四足、类人机器人）和使用案例（体现生成、跨体现控制器）。变换器变换器不是针对单一奖励函数进行过拟合，而是一个动力学模型，其与奖励无关的状态和动作预测可以转化为特定奖励的价值预测。这些价值预测用于通过我们称之为动力学自我引导（Dynamics Self-Guidance）的过程，引导体现扩散朝向高价值的机器人设计。在多个设计空间的实验中，展示了对未见奖励和轨迹的零-shot优化，相较于进化基线，性能和运行时间均有所改善。最后，我们制造了一个优化的ALOHA设计，与原始设计相比，跟踪误差减少了超过70%。
cs.RO / 25 / 2607.25807
Modular Robotic Catheters for Endovascular Aneurysm Repair
用于血管内动脉瘤修复的模块化机器人导管
Abstract
Fenestrated/Branched endovascular aneurysm repair (FEVAR/BEVAR) require surgeons to navigate catheters and guidewires into various branches of the abdominal aorta, before deploying stent grafts to alleviate pressure on the aneurysm. Previous clinical studies suggests that surgeons continue to struggle with vessel access using standard commercial instruments, prolonging the procedural time and inducing further complications. In this work, we present two contributions to solving this problem: 1) A bespoke 2-segment steerable catheter, consisting of 4 degrees of freedom to enhance dexterity. 2) An expandable, modular tendon-driven actuation platform that can accommodate for the redundancies introduced in our system. To fabricate the catheter, we capitalized on thermal fiber drawing, a technique that creates high-aspect ratio devices at scale, and processed the catheter with laser micro-machining to soften its tip. We evaluated the system using simulations, where we investigated the catheter's bending stiffness, then its steerability with in-vitro experiments in vascular phantoms. This handheld, robotic steerable catheter system has the potential to shorten the length of future endovascular surgeries, and give clinicians the tools to resolve challenging clinical cases.
Chinese Translation
带孔/分支的血管内动脉瘤修复（FEVAR/BEVAR）要求外科医生将导管和引导线导航到腹主动脉的各个分支，然后部署支架移植物以减轻动脉瘤的压力。以往的临床研究表明，外科医生在使用标准商业仪器进行血管通路时仍然面临困难，导致手术时间延长并引发更多并发症。在本研究中，我们提出了两项解决此问题的贡献：1）一种定制的两段可操控导管，具有4个自由度以增强灵活性；2）一种可扩展的模块化腱驱动执行平台，以适应我们系统中引入的冗余。为了制造导管，我们利用了热纤维拉伸技术，这是一种能够大规模生产高纵横比设备的技术，并通过激光微加工处理导管以软化其尖端。我们通过模拟评估了该系统，研究了导管的弯曲刚度，然后在血管模型中进行了体外实验以测试其可操控性。这种手持式机器人可操控导管系统有潜力缩短未来血管内手术的时间，并为临床医生提供解决复杂临床案例的工具。
cs.RO / 26 / 2607.25895
HiFi-UMI: Learning Deployable Manipulation Policies from High-Fidelity UMI Data Alone
HiFi-UMI：仅从高保真UMI数据中学习可部署的操作策略
Abstract
Learning deployable manipulation policies is bottlenecked by the scarcity of data that is both high-fidelity and scalable. Real-robot teleoperation is accurate but costly to scale; robot-free UMI capture scales readily, and current practice uses the resulting data mainly for pre-training, adding a small real-robot "anchor" at post-training. We ask whether raising the fidelity of robot-free UMI data, rather than shrinking the real-robot fraction, can remove that anchor. We present HiFi-UMI, a portable UMI data-production system co-designed for trajectory accuracy, inter-gripper relative pose, synchronization, and field of view: head-mounted offline stereo-inertial SLAM, native rather than reconstructed relative pose, a shared microsecond GPIO trigger, and two wide-angle cameras per hand covering ~200 degrees. It reaches 3 mm workspace-local end-effector accuracy without external tracking infrastructure. Using this corpus, we demonstrate zero-robot post-training: a policy post-trained solely on HiFi-UMI demonstrations deploys directly on a real robot and matches in-domain teleoperation across three backbones spanning the vision-language-action and world-action-model families, with success-rate differences of -2.5, +3.1, and -0.6 percentage points on StarVLA-QwenPI, OpenPI-pi_0.5, and LingBot-VA; the strongest policy reaches 85% on a precision insertion task, even though the teleoperation baseline is collected in the evaluation scene and no HiFi-UMI trajectory is. Pre-training on 4,000 hours from the same corpus lowers action error on ten unseen tasks by 41% and, on StarVLA-QwenPI, raises real-robot success by a further 18.1 percentage points. We open-source HiFi-UMI-2K, 2,000 hours of microsecond-synchronized, ultra-wide-FoV demonstrations, each automatically reconstructed and validated through simulation replay, as a large-scale, high-fidelity resource for the robot-learning community.
Chinese Translation
学习可部署的操作策略受到高保真且可扩展数据稀缺的瓶颈。真实机器人遥操作虽然准确，但扩展成本高；无机器人UMI捕获易于扩展，目前的实践主要将生成的数据用于预训练，在后期训练中添加一个小的真实机器人“锚点”。我们探讨提高无机器人UMI数据的保真度，而不是缩小真实机器人比例，是否能够去除该锚点。我们提出了HiFi-UMI，一个便携的UMI数据生产系统，旨在实现轨迹精度、夹具间相对姿态、同步和视场的共同设计：头戴式离线立体惯性SLAM、原生而非重建的相对姿态、共享微秒GPIO触发器，以及每只手配备的两个覆盖约200度的广角相机。它在没有外部跟踪基础设施的情况下，实现了3毫米工作空间局部末端执行器精度。利用这一数据集，我们展示了零机器人后训练：仅基于HiFi-UMI演示进行后训练的策略可以直接在真实机器人上部署，并在三个涵盖视觉-语言-动作和世界-动作-模型家族的骨干网络中与领域内遥操作相匹配，成功率差异分别为-2.5、+3.1和-0.6个百分点，针对StarVLA-QwenPI、OpenPI-pi_0.5和LingBot-VA；最强策略在精确插入任务中达到85%的成功率，尽管遥操作基线是在评估场景中收集的，而没有HiFi-UMI轨迹。对来自同一数据集的4,000小时数据进行预训练，使十个未见任务的动作误差降低了41%，并在StarVLA-QwenPI上进一步提高了真实机器人成功率18.1个百分点。我们开源了HiFi-UMI-2K，提供2,000小时微秒同步的超广视场演示，每个演示通过仿真重放自动重建和验证，作为机器人学习社区的大规模高保真资源。
cs.RO / 27 / 2607.25912
SAM3D-Guided Object-Centric Representation Alignment for Vision-Language-Action Models
基于SAM3D的面向对象的三维表示对齐用于视觉-语言-动作模型
Abstract
Vision-Language-Action (VLA) models have shown strong potential for general robot manipulation, but most existing models rely on 2D visual-language backbones and lack fine-grained 3D understanding of target objects, especially under occlusion, pose variation, scale changes, and precise spatial interaction. We propose an object-centric 3D representation alignment framework built upon $\pi_0$, using SAM3D as a frozen 3D teacher to provide target-object 3D priors during training. Specifically, we localize task-relevant objects with object recognition models, generate corresponding object masks, and use SAM3D to extract dense object-level 3D representations, which are aligned with intermediate visual features of $\pi_0$. This enables the policy to internalize target-object 3D information while preserving the original RGB-language-to-action inference pipeline without requiring depth, point clouds, masks, SAM3D, or additional 3D modules at test time. Simulation experiments show consistent improvements, achieving 99.1\% on LIBERO and an average length of 4.11 on CALVIN. Real-world experiments further demonstrate that our method is particularly effective in long-horizon manipulation scenarios where the robot must focus on different target objects across multiple subtasks.
Chinese Translation
视觉-语言-动作（VLA）模型在通用机器人操作中展现出强大的潜力，但大多数现有模型依赖于二维视觉-语言骨干，缺乏对目标对象的细粒度三维理解，尤其是在遮挡、姿态变化、尺度变化和精确空间交互的情况下。我们提出了一种基于$ ext{π}_0$的面向对象的三维表示对齐框架，使用SAM3D作为冻结的三维教师，在训练过程中提供目标对象的三维先验信息。具体而言，我们利用对象识别模型定位任务相关对象，生成相应的对象掩码，并使用SAM3D提取密集的对象级三维表示，这些表示与$ ext{π}_0$的中间视觉特征对齐。这使得策略能够内化目标对象的三维信息，同时在测试时无需深度、点云、掩码、SAM3D或额外的三维模块，从而保留原始的RGB-语言-动作推理流程。仿真实验显示出一致的改进，在LIBERO上达到99.1\%的准确率，在CALVIN上平均长度为4.11。现实世界的实验进一步证明，我们的方法在长时间操作场景中特别有效，在这些场景中，机器人必须在多个子任务中关注不同的目标对象。
cs.RO / 28 / 2607.25918
DC-WAM: Dynamic-Centric Visual Supervision and Reasoning for World-Action Models
DC-WAM：面向世界-动作模型的动态中心视觉监督与推理
Abstract
World-Action Models (WAMs) augment robot policies with future visual prediction, but it remains unclear what the visual modality should learn for control. While photorealistic future prediction provides dense supervision, it also incurs substantial computation and can allocate capacity to texture, illumination, and background variations that are only weakly related to action selection. Recent efficient WAM variants suggest that the main benefit of the video branch may not lie in the rendered future itself, but in the control-relevant visual representations induced during training. In this work, we revisit future video prediction from a dynamic-centric perspective and ask whether an existing RGB-based WAM can be redirected from appearance-dominated reconstruction toward interaction-induced visual dynamics without introducing additional modality-specific predictions or online inputs at deployment. We propose DC-WAM, a dynamic-centric WAM framework that redistributes supervision and computation in the RGB video branch. At the supervision level, DC-WAM combines temporal-difference flow matching with trajectory-guided weighting, emphasizing dense temporal changes and localized regions where the gripper, manipulated objects, and contact areas move. At the reasoning level, DynaRoute predicts token-wise dynamic relevance and converts it into an attention bias, guiding the model toward control-relevant future tokens. Experiments in simulation and on real-world manipulation tasks show that DC-WAM consistently improves policy performance, especially under out-of-distribution perturbations in lighting, object appearance, and background texture.
Chinese Translation
世界-动作模型（WAMs）通过未来视觉预测增强机器人策略，但尚不清楚视觉模态应学习什么以实现控制。尽管逼真的未来预测提供了密集的监督，但也带来了巨大的计算开销，并且可能将能力分配给与动作选择关系较弱的纹理、光照和背景变化。最近的高效WAM变体表明，视频分支的主要好处可能不在于渲染的未来本身，而在于训练过程中诱导的与控制相关的视觉表征。在本研究中，我们从动态中心的角度重新审视未来视频预测，并询问现有的基于RGB的WAM是否可以在不引入额外模态特定预测或在线输入的情况下，从以外观为主导的重建转向交互诱导的视觉动态。我们提出了DC-WAM，一个动态中心的WAM框架，重新分配RGB视频分支中的监督和计算。在监督层面，DC-WAM结合了时间差流匹配与轨迹引导加权，强调密集的时间变化和夹持器、操控物体及接触区域移动的局部区域。在推理层面，DynaRoute预测逐个标记的动态相关性，并将其转换为注意力偏置，引导模型朝向与控制相关的未来标记。仿真和真实世界操控任务中的实验表明，DC-WAM始终提高了策略性能，尤其是在光照、物体外观和背景纹理的分布外扰动下。
cs.RO / 29 / 2607.25985
Physics-Aware End-to-End Deep Reinforcement Learning for Quadcopter Control with Actuator Dynamics
考虑物理的端到端深度强化学习在四旋翼控制中的应用与执行器动态
Abstract
Unmanned aerial vehicles (UAVs), particularly quadcopters, present unique challenges for autonomous control due to their underactuated dynamics: only four available control inputs must govern six degrees of freedom. This paper investigates a physics-aware, end-to-end deep reinforcement learning (DRL) approach that acts directly on low-level body inputs, total thrust and body torques $(T, \tau_x, \tau_y, \tau_z)$, and closes the loop through a high-fidelity Simulink environment. Our simulator integrates a 12-state rigid-body model (MATLAB Level-2 S-Function) with (i) an Action2RPM allocation based on the Moore-Penrose pseudo-inverse of a coefficient matrix derived from thrust and drag terms, and (ii) first-order actuator dynamics for each motor (time constant $T_m = 0.076$ s), including rotor gyroscopic coupling. A shaped reward balances goal-reaching and stability using an exponential position well, attitude penalties, and quadratic velocity costs. Four DRL algorithms, DDPG, TD3, PPO, and SAC, are evaluated in two stages: (S1) thrust-only hover and (S2) hover with pitch torque and a translated goal. Results show that SAC and TD3 achieve superior stability and exploration efficiency, while PPO is less sample-efficient. The study highlights the significance of modeling actuator lags and aerodynamic moments for stable low-level control and provides a reproducible benchmark for quadcopter DRL.
Chinese Translation
无人机（UAV），尤其是四旋翼，因其欠驱动动态而在自主控制方面面临独特挑战：仅有四个可用控制输入需控制六个自由度。本文研究了一种考虑物理的端到端深度强化学习（DRL）方法，该方法直接作用于低级别的身体输入，即总推力和身体扭矩 $(T, au_x, au_y, au_z)$，并通过高保真Simulink环境闭环控制。我们的模拟器整合了一个12状态刚体模型（MATLAB Level-2 S-Function），并采用(i) 基于从推力和阻力项导出的系数矩阵的Moore-Penrose伪逆的Action2RPM分配，以及(ii) 每个电机的一级执行器动态（时间常数 $T_m = 0.076$ s），包括转子陀螺耦合。设计的奖励函数通过指数位置井、姿态惩罚和二次速度成本平衡目标达成与稳定性。评估了四种DRL算法：DDPG、TD3、PPO和SAC，分为两个阶段进行测试：(S1) 仅推力悬停和(S2) 带俯仰扭矩的悬停及平移目标。结果表明，SAC和TD3在稳定性和探索效率上表现优越，而PPO的样本效率较低。本研究强调了建模执行器滞后和气动力矩在稳定低级控制中的重要性，并提供了一个可重复的四旋翼DRL基准。
cs.RO / 30 / 2607.26047
S2A2: Audio-Visual Imitation Learning for Manipulation Tasks Using Acoustic Spatial Information
S2A2：基于声学空间信息的音视频模仿学习在操控任务中的应用
Abstract
Acoustic information provides rich cues about object location, material properties, and changes caused by contact or motion. This paper introduces a new set of acoustic-aware manipulation tasks for imitation learning, in which robots must use auditory cues to determine manipulation targets. These tasks require sound source localization and identification for active exploration in robotic manipulation. Also, we propose a multimodal imitation learning framework, Spatial-Spectral Audio Action (S2A2), that integrates visual features with acoustic spatial and acoustic signal information for the acoustic-aware manipulation tasks. We implemented S2A2 models that integrates policies such as ACT, Diffusion Policy, VQ-BeT, and $\pi_0$, into our framework. Simulation experiments showed that the proposed method is the most effective for tasks requiring both position and timbre. Furthermore, real-robot experiments confirm the applicability of the proposed tasks and framework to real-world manipulation.
Chinese Translation
声学信息提供了关于物体位置、材料特性以及接触或运动引起的变化的丰富线索。本文介绍了一组新的声学感知操控任务，用于模仿学习，其中机器人必须利用听觉线索来确定操控目标。这些任务要求进行声源定位和识别，以便在机器人操控中进行主动探索。此外，我们提出了一种多模态模仿学习框架，空间-谱音频动作（S2A2），该框架将视觉特征与声学空间和声学信号信息结合，以应对声学感知操控任务。我们实现了将策略如ACT、扩散策略（Diffusion Policy）、VQ-BeT和$ ext{π}_0$整合到我们框架中的S2A2模型。仿真实验表明，所提方法在需要位置和音色的任务中最为有效。此外，真实机器人实验验证了所提任务和框架在实际操控中的适用性。
cs.RO / 31 / 2607.26055
$\pi\mathbf{R}^2$: Reactive Real-time Flow Policies
$oldsymbol{ ext{πR}}^2$: 反应式实时流政策
Abstract
Generalist manipulation policies increasingly take the form of action-chunking flow policies built on large pretrained backbones. Such chunks run open-loop, so the policy cannot react to sensory input arriving mid-execution, sacrificing \emph{reactivity}. Replanning more often would restore it, but the perception-to-action pipeline (a large backbone plus multiple denoising steps) is too slow: this \emph{latency} forbids frequent replanning and leaves committed actions stale, making such policies ill-suited for dynamic, closed-loop control. We present $\pi\mathbf{R}^2$, which makes these policies reactive and real-time while retaining large backbones, expressive multi-modal policies, and multi-action prediction. Built on the per-position noise schedule of diffusion forcing, $\pi\mathbf{R}^2$ contributes two ideas. First, it splits conditioning into a fast channel (proprioception, fresh every tick) and an asynchronously updated slow channel (vision-language features), so the policy reacts to proprioception within a chunk while tolerating stale vision. Second, a latency-adaptive flow schedule treats in-flight actions as inpainting conditioning and emits actions in one denoising step per call, letting one trained model adapt to varying hardware latency. Requiring minimal modification to existing architectures, $\pi\mathbf{R}^2$ can be finetuned from a pretrained policy: applied to GR00T-N1.7 on a real xArm6+XHand platform, it replans closed-loop roughly $4\times$ faster than the base policy (~$25$Hz on an A5000 GPU), acting on a fresh observation every $40$ms. Across simulation and real-world manipulation tasks, $\pi\mathbf{R}^2$ improves the success rate by up to $23\%$ in simulation and $30\%$ in the real world over the strongest baseline. Project page: https://pi-r2-flow.github.io/
Chinese Translation
通用操作策略越来越多地采用基于大型预训练骨干网络的动作分块流政策。这些分块以开放循环的方式运行，因此策略无法在执行过程中对到达的感官输入做出反应，从而牺牲了 extit{反应性}。更频繁的重新规划可以恢复反应性，但感知到行动的管道（一个大型骨干网络加上多个去噪步骤）过于缓慢：这种 extit{延迟}禁止了频繁的重新规划，并使得已承诺的动作变得过时，使得这些政策不适合动态闭环控制。我们提出了$oldsymbol{ ext{πR}}^2$，使这些政策在保留大型骨干网络、表达性多模态政策和多动作预测的同时，变得反应式和实时。$oldsymbol{ ext{πR}}^2$基于扩散强制的每个位置噪声调度，贡献了两个思想。首先，它将条件分为一个快速通道（本体感知，每个时刻都是新的）和一个异步更新的慢通道（视觉-语言特征），因此策略在一个分块内对本体感知做出反应，同时容忍过时的视觉信息。其次，延迟自适应流调度将正在进行的动作视为修补条件，并在每次调用中以一个去噪步骤发出动作，让一个训练好的模型适应不同的硬件延迟。$oldsymbol{ ext{πR}}^2$对现有架构的修改要求最小，可以从预训练政策进行微调：在真实的xArm6+XHand平台上应用于GR00T-N1.7时，它的闭环重新规划速度大约是基础政策的$4 imes$（在A5000 GPU上约为$25$Hz），每$40$毫秒对新的观测做出反应。在模拟和现实世界的操作任务中，$oldsymbol{ ext{πR}}^2$在模拟中将成功率提高了高达$23\%$，在现实世界中提高了$30\\%$，超越了最强基线。项目页面：https://pi-r2-flow.github.io/
cs.RO / 32 / 2607.26056
INTACT: Isomorphic Intent-to-Action Learning for Search-Free World Models
INTACT：无搜索世界模型的同构意图到行动学习
Abstract
Forward latent world models predict how actions change a scene, but recover actions for a desired change only through expensive test-time search. We introduce INTACT (INtent-To-ACTion), an end-to-end JEPA that turns action-labeled, reward-free trajectories into a deployable intent-to-action interface. Each transition supplies physical intent $z_{t+1}-z_t$, while a future goal supplies deployment intent $\operatorname{sg}(z_g)-z_t$. The architecture is isomorphic between the local and goal motion-intent backbone-input graphs through an identical four-slot grammar and shared parameters, and between supported local and goal motion-intent families through action-law semantics induced by the same predictor rather than pointwise latent equality. INTACT also provides intact transfer from RGB evidence to action-effective latent intent coordinates and from intent families to their corresponding action-law families. Asymmetric endpoint gradients ground physical successors and fix future goals as anchors, joining representation learning and control without pointwise latent matching or globally linear dynamics. The resulting coordinates support a robust distributional action law: its conditional mean serves directly as a search-free policy, while sampling remains available for diversity or optional verification. On the four official LeWM tasks, one-epoch, zero-search models reach 85.78\%, 100.00\%, 97.67\%, and 97.89\% success. Optional local CEM centered on the Direct plan reaches 96.86\% macro success using 384 instead of 9,000 candidate sequences, reducing sampling by $23.44\times$ while improving pure CEM by 16.00 points. One shared four-task encoder reaches 89.39\% E5 Direct macro and improves every task over jointly trained LeWM, while predicted--expert action-family kNN tracks Direct success at $r=0.954$. Direct inference takes 2.9--5.5 ms.
Chinese Translation
前向潜在世界模型预测动作如何改变场景，但仅通过昂贵的测试时搜索来恢复所需变化的动作。我们引入了INTACT（意图到行动），一种端到端的JEPA，将标记动作的无奖励轨迹转化为可部署的意图到行动接口。每个过渡提供物理意图 $z_{t+1}-z_t$，而未来目标提供部署意图 $ ext{sg}(z_g)-z_t$。该架构在局部和目标运动意图主干输入图之间是同构的，通过相同的四槽语法和共享参数，以及在支持的局部和目标运动意图家族之间通过同一预测器诱导的动作法则语义，而不是逐点潜在相等。INTACT还提供从RGB证据到动作有效潜在意图坐标的完整转移，以及从意图家族到其对应的动作法则家族的转移。非对称端点梯度将物理后继固定为锚点，并将未来目标作为锚点，结合表示学习和控制，而无需逐点潜在匹配或全局线性动态。生成的坐标支持一个稳健的分布式动作法则：其条件均值直接作为无搜索策略，而采样仍然可用于多样性或可选验证。在四个官方LeWM任务中，一轮、零搜索模型的成功率分别达到85.78%、100.00%、97.67%和97.89%。以Direct计划为中心的可选局部CEM使用384而不是9,000个候选序列，达到了96.86%的宏观成功，减少了采样$23.44 imes$，同时将纯CEM提高了16.00分。一个共享的四任务编码器达到了89.39%的E5 Direct宏观成功，并在每个任务上优于联合训练的LeWM，而预测的专家动作家族kNN在$r=0.954$时跟踪Direct成功。直接推理的时间为2.9至5.5毫秒。
cs.CV / 1 / 2607.24835
Gradient-Based Latent Decomposition Reveals Mechanisms of Feature Degradation in Weakly Supervised Mammography
基于梯度的潜在分解揭示弱监督乳腺摄影中特征退化的机制
Abstract
Weakly supervised hierarchical models exhibit a persistent asymmetry: coarse lesion-type features are preserved under reconstruction while fine-grained malignancy cues degrade---a pattern with direct consequences for the clinical reliability of breast cancer screening pipelines. We introduce gradient-based orthogonal latent decomposition for hierarchical Variational Autoencoders~(H-VAEs) to mechanistically explain this asymmetry. The latent space is partitioned into a task-aligned component~($z_1$), shaped by coarse supervisory gradients, and an orthogonal residual~($z_{\text{res}}$) capturing remaining representational capacity. On~3,550 mammographic Regions of Interest~(ROIs) from CBIS-DDSM, only~$\sim$4.4\% of latent magnitude aligns with supervisory gradients, leaving~$\sim$95.6\% in the orthogonal residual upon which fine-grained pathology prediction primarily depends. The model achieves Stage-1~AUC~0.866 and Stage 2~AUC~0.552, with a reconstruction stability gap of $\Delta_{\text{diag}}=5\%$ ($p=0.005$) and a classification gap of $\Delta_{\text{AUC}}=0.314$ ($p{<}0.001$). Latent ablation confirms that features for both tasks reside heavily in~$z_{\text{res}}$, structurally explaining why reconstruction degrades pathology stability disproportionately. Comparisons with Multi-Instance Learning~(MIL) and Multi-Task Learning~(MTL) confirm generalization across architectures and modalities. These findings reveal that in high-dimensional spaces, a single coarse supervisory signal isolates only a sparse 1D latent direction, forcing critical fine-grained features into the vulnerable residual subspace.
Chinese Translation
弱监督层次模型表现出一种持续的不对称性：粗略的病变类型特征在重建过程中得以保留，而细粒度的恶性线索却退化——这一模式对乳腺癌筛查管道的临床可靠性产生直接影响。我们引入基于梯度的正交潜在分解方法，应用于层次变分自编码器（H-VAEs），以机制性地解释这种不对称性。潜在空间被划分为一个与任务对齐的成分（$z_1$），由粗略的监督梯度塑造，以及一个捕获剩余表征能力的正交残差（$z_{ ext{res}}$）。在来自CBIS-DDSM的3,550个乳腺摄影感兴趣区域（ROIs）中，只有约4.4%的潜在幅度与监督梯度对齐，约95.6%则留在正交残差中，细粒度病理预测主要依赖于此。该模型在阶段1的AUC为0.866，阶段2的AUC为0.552，重建稳定性差距为$ ext{diag} riangle = 5\%$（$p=0.005$），分类差距为$ ext{AUC} riangle = 0.314$（$p{<}0.001$）。潜在消融实验确认，两个任务的特征主要集中在$z_{ ext{res}}$中，结构性地解释了为何重建过程不成比例地降低了病理稳定性。与多实例学习（MIL）和多任务学习（MTL）的比较确认了跨架构和模态的泛化能力。这些发现揭示，在高维空间中，单一的粗略监督信号仅能孤立出稀疏的1D潜在方向，迫使关键的细粒度特征进入脆弱的残差子空间。
cs.CV / 2 / 2607.24852
Track-Leakage-Free Hold-Out Self-Validation for Photogrammetric Reconstruction: Protocol, Sensitivity, and Limits
无轨迹泄漏的自我验证保留协议用于摄影测量重建：协议、灵敏度与局限性
Abstract
Automated photogrammetric inspection emits metric measurements from a 3D reconstruction whose own correctness is normally unknown without an external survey. Can a reconstruction estimate its own reliability with no ground truth? We formalise a track-leakage-free hold-out protocol: a deterministic subset of images is withheld and each re-localised against only 3D points seen by at least two retained images -- a track-level barrier so a view is never tested against structure it helped create -- aggregated into an mAA confidence. Held-out observations still entered the bundle adjustment that built the trusted structure, so we call it track-leakage-free (a stricter re-mapping variant confirms this at 3/5 deg). Across operational GNSS-referenced captures plus ETH3D, EuRoC and the IMC 2025 benchmark we report four findings. (i) The protocol is computationally well-posed: good reconstructions score millidegree self-consistency. (ii) Thresholded self-consistency saturates and does not track absolute accuracy: confidence stays near 1.00 while true RTK error swings up to 14x within a capture, and the per-capture correlation is sign-unstable across five captures (-0.57 to +0.98; pooled 95% CI spanning zero), so it cannot gate accuracy. (iii) It flags gross failure only when the failure destroys internal consistency: a fragmenting model drops confidence, but a single self-consistent, globally-distorted model evades it -- three of four captures gave one model wrong by 55-106 m at confidence 1.00. (iv) The same dichotomy holds on independent ETH3D and IMC 2025 ground truth. Track-leakage-free hold-out measures internal geometric consistency, not absolute accuracy: neither a substitute for control-point assessment nor a general gross-failure gate. We release the protocol and degradation harness.
Chinese Translation
自动化摄影测量检查从3D重建中发出度量测量，而其自身的正确性通常在没有外部调查的情况下是未知的。重建是否能够在没有真实值的情况下估计其自身的可靠性？我们形式化了一种无轨迹泄漏的保留协议：一个确定性的图像子集被保留，并且每个重新定位仅针对至少由两个保留图像看到的3D点进行测试——这是一个轨迹级别的障碍，因此某个视图永远不会与其帮助创建的结构进行测试——聚合成一个mAA置信度。被保留的观测仍然进入构建可信结构的束调整，因此我们称之为无轨迹泄漏（一个更严格的重新映射变体在3/5度时确认了这一点）。在操作性GNSS参考捕获以及ETH3D、EuRoC和IMC 2025基准测试中，我们报告了四个发现。(i) 该协议在计算上是良好设定的：良好的重建得分为毫度自一致性。(ii) 阈值自一致性饱和且不跟踪绝对精度：置信度保持在接近1.00，而真实的RTK误差在一次捕获中波动高达14倍，并且每次捕获的相关性在五次捕获中是不稳定的（-0.57到+0.98；合并的95%置信区间跨越零），因此它不能作为精度的门限。(iii) 只有当失败破坏内部一致性时，它才会标记严重失败：一个碎片化的模型降低了置信度，但一个单一自一致的、全球失真的模型则逃避了这一点——四次捕获中的三次给出了一个置信度为1.00、错误值为55-106米的模型。(iv) 在独立的ETH3D和IMC 2025真实值上同样存在这种二分法。无轨迹泄漏的保留测量内部几何一致性，而非绝对精度：既不是控制点评估的替代品，也不是一般的严重失败门限。我们发布了该协议和降级工具。
cs.CV / 3 / 2607.24856
DisasterTD: Disaster Toponym Disambiguation Using Multimodal LLMs and Cross-View Geolocalization
DisasterTD：基于多模态大语言模型和跨视角地理定位的灾害地名消歧义
Abstract
Social media imagery (SMI) provides timely and fine-grained ground perspectives that are valuable for situational awareness and emergency response. Unlike satellite or aerial imagery, SMI can capture disaster impacts and ground-level conditions in a timely manner. However, geographic references in SMI are often vague or ambiguous, making accurate geolocalization challenging. To address this issue, we propose DisasterTD, a disaster toponym disambiguation framework that integrates multimodal large language model (MLLMs)-based semantic reasoning with cross-view geolocalization. First, MLLMs extract toponyms and generate candidate geolocations from noisy textual inputs. Then, cross-view matching between SMI, remote sensing imagery (RSI), and optionally street-view imagery (SVI) is used to verify and refine these candidate results. We evaluate DisasterTD on the Hurricane Harvey dataset, where SMI is augmented with collected RSI and SVI to construct a cross-view benchmark for disaster geolocalization. The dataset is divided into four categories based on toponym clarity and ambiguity, allowing a fine-grained performance analysis across scenarios. Results show that DisasterTD consistently outperforms MLLM-only and cross-view-only baselines without disambiguation, achieving geolocalization accuracies of 71.62% within 1000 m, 62.36% within 500 m, 57.99% within 250 m, 52.09% within 100 m, and 47.01% within 50 m, while reducing the mean and median errors to 11.33 km and 0.68 km, respectively. The largest improvements appear in ambiguous toponyms, where semantic reasoning with cross-view evidence reduces candidate dispersion and errors. These findings demonstrate the effectiveness of integrating MLLM-based candidate generation with cross-view verification for fine-grained disaster geolocalization.
Chinese Translation
社交媒体图像（SMI）提供了及时且细致的地面视角，这对情境意识和应急响应具有重要价值。与卫星或航空图像不同，SMI能够及时捕捉灾害影响和地面条件。然而，SMI中的地理参考往往模糊或不明确，这使得准确的地理定位变得具有挑战性。为了解决这一问题，我们提出了DisasterTD，一个整合了基于多模态大语言模型（MLLMs）的语义推理和跨视角地理定位的灾害地名消歧义框架。首先，MLLMs从嘈杂的文本输入中提取地名并生成候选地理位置。然后，通过SMI、遥感图像（RSI）和可选的街景图像（SVI）之间的跨视角匹配来验证和优化这些候选结果。我们在飓风哈维数据集上评估DisasterTD，其中SMI与收集的RSI和SVI相结合，构建了一个用于灾害地理定位的跨视角基准。该数据集根据地名的清晰度和模糊性分为四个类别，从而允许在不同场景中进行细致的性能分析。结果表明，DisasterTD在没有消歧义的情况下，始终优于仅使用MLLM和仅使用跨视角的基线，达到了71.62%的1000米内地理定位准确率，62.36%的500米内准确率，57.99%的250米内准确率，52.09%的100米内准确率，以及47.01%的50米内准确率，同时将平均和中位数误差分别降低到11.33公里和0.68公里。在模糊地名中，最大的改进表现得尤为明显，语义推理结合跨视角证据减少了候选分散和错误。这些发现证明了将基于MLLM的候选生成与跨视角验证相结合用于细致的灾害地理定位的有效性。
cs.CV / 4 / 2607.24898
Harm is not Universal: Community-Specific Toxicity Detection is Urgently Needed
伤害并非普遍存在：迫切需要社区特定的毒性检测
Abstract
State-of-the-art toxicity detectors for text-to-image generation adopt a one-size-fits-all approach: a single universal model applying fixed safety guidelines to all users. Our empirical evidence shows that these detectors fail to shield marginalized communities: approximately 35% of generated images labeled safe are considered harmful by disability communities. In this position paper, we argue for community-specific toxicity detection (CTD). To demonstrate its feasibility, we collaborate with disability experts to develop safety guidelines for two communities: dwarfism and blind/low vision. Using a dataset of 2,400 annotated T2I-generated images we demonstrate that both large vision-language models and existing general-purpose toxicity detectors catastrophically fail to recognize harmful content under these guidelines in zero-shot settings with F1 score lower than random guessing (F1 0.32 and 0.37). Promisingly, prompt-based adaptation methods (ICL, VQA) substantially improve harm detection performance (GPT-4o: F1 0.50 and 0.78), while parameter-efficient fine-tuning improves smaller models (0.5b-7b with best F1 0.48 and 0.59) with less than 100 demonstrations, but remains sensitive to evolving guidelines. Despite these gains, CTD performance remains far below F1 $\approx 0.9$ achieved for general-purpose toxicity detection, highlighting the challenge and the need for sustained research effort.
Chinese Translation
当前最先进的文本到图像生成毒性检测器采用一刀切的方法：一个通用模型对所有用户应用固定的安全指南。我们的实证证据表明，这些检测器未能保护边缘化社区：大约35%的被标记为安全的生成图像在残疾社区中被认为是有害的。在这篇立场论文中，我们主张进行社区特定的毒性检测（CTD）。为了证明其可行性，我们与残疾专家合作，为两个社区（侏儒症和盲/低视力）制定安全指南。通过使用2400个标注的T2I生成图像的数据集，我们展示了大型视觉-语言模型和现有的通用毒性检测器在这些指南下在零样本设置中识别有害内容的能力严重不足，F1分数低于随机猜测（F1 0.32和0.37）。令人鼓舞的是，基于提示的适应方法（ICL, VQA）显著提高了伤害检测的性能（GPT-4o: F1 0.50和0.78），而参数高效的微调则改善了较小模型（0.5b-7b，最佳F1 0.48和0.59），在少于100个示例的情况下，但仍对不断变化的指南敏感。尽管取得了这些进展，CTD的性能仍远低于通用毒性检测所达到的F1 $ ext{approx} 0.9$，突显了这一挑战和持续研究努力的必要性。
cs.CV / 5 / 2607.24904
Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model
Mage-VL：一种高效的编解码器原生流媒体多模态基础模型
Yang, Senqiao, Zhang, Kaichen, Jia, Zhaoyang, Guo, Jinghao, Shen, Yifei, Zhang, Xinjie, Zhang, Xiaoyi, Wang, Haoqing, Li, Xiao, Zhang, Peng, An, Xiang, Xie, Yin, Liu, Zhening, Guo, Xun, Li, Jiahao, Zheng, Shicheng, Wang, Jinglu, Guo, Zongyu, Xie, Wenxuan, Zheng, Zihan, Luo, Yuxuan, Li, Bin, Lu, Yan
Abstract
Standard vision-language models (VLMs) suffer from Moravec's paradox: they excel at complex offline visual reasoning but struggle with simple streaming perception tasks and process them inefficiently. We present Mage-VL, an efficient codec-native streaming foundation model for real-time multimodal understanding and interaction. At its core, our custom tokenizer, Mage-ViT, replaces uniform frame sampling by selectively encoding dynamic, entropy-rich regions using motion vectors and residual energy across sparse anchor (I) and predicted (P) frames. Operating at a 16 x 16 patch level, this reduces visual token consumption by over 75% while preserving spatiotemporal context. Trained from scratch on approximately 560M unlabeled images and 100M unlabeled video frames, Mage-ViT matches or outperforms flagship encoders trained on billions of image-text pairs. We establish AI4AI data pipelines encompassing prompt-code joint optimization for multimodal captioning and AI-driven performance diagnosis to guide training recipes. Furthermore, through a bio-inspired dual-system architecture - a lightweight System 1 event gate and a causal System 2 decoder - Mage-VL enables proactive streaming perception. Extensive evaluations show that Mage-VL-4B matches Qwen3-VL-4B on static tasks while achieving strong gains in video understanding and 2D/3D spatial reasoning, with up to a 3.5x wall-clock inference speedup, and comprehensively surpasses the 15B Phi-4-reasoning-vision baseline. Beyond model artifacts, we deliver seven key empirical findings covering pre-training data efficiency, variable-resolution scaling, codec system acceleration, VideoQA SFT redundancy, motion-spatial synergy, AI4AI data pipelines, and Zero-Vision SFT for multimodal RL.
Chinese Translation
标准的视觉-语言模型（VLMs）面临莫拉维克悖论：它们在复杂的离线视觉推理方面表现出色，但在简单的流媒体感知任务中却表现不佳且效率低下。我们提出了Mage-VL，一种高效的编解码器原生流媒体基础模型，用于实时多模态理解和交互。其核心是我们的定制分词器Mage-ViT，它通过使用运动向量和稀疏锚点（I帧）及预测帧（P帧）之间的残余能量，选择性地编码动态的、富有熵的区域，从而替代均匀帧采样。在16 x 16的补丁级别上操作，这使得视觉标记的消耗减少了超过75%，同时保留了时空上下文。在约5.6亿张未标记图像和1亿帧未标记视频上从头开始训练，Mage-ViT的性能与在数十亿图像-文本对上训练的旗舰编码器相匹配或超越。我们建立了AI4AI数据管道，涵盖多模态字幕的提示-代码联合优化和AI驱动的性能诊断，以指导训练方案。此外，通过一种生物启发的双系统架构——轻量级的系统1事件门和因果系统2解码器——Mage-VL实现了主动的流媒体感知。广泛的评估显示，Mage-VL-4B在静态任务上与Qwen3-VL-4B相匹配，同时在视频理解和2D/3D空间推理方面取得显著进展，推理速度提升高达3.5倍，并全面超越了15B Phi-4推理-视觉基线。除了模型工件外，我们还提供了七个关键的实证发现，涵盖预训练数据效率、可变分辨率缩放、编解码系统加速、VideoQA SFT冗余、运动-空间协同、AI4AI数据管道和多模态强化学习的零视觉SFT。
cs.CV / 6 / 2607.24957
PerceptionBench: Evaluating Atomic Visual Perception in Multimodal Large Language Models
PerceptionBench：评估多模态大型语言模型中的原子视觉感知能力
Lin, Zichao, Xie, Yifeng, Qu, Bowen, Wang, Haiming, Li, Jia, Wu, Haoning, Dong, Yuhao, Yang, Zuhao, Zhu, Jinguo, Lu, Haoyu, Zhao, Zijia, Yue, Tongtian, Qi, Zhangyang, Yang, Junwei, Dong, Mengfan, Cao, Peizhou, Du, Chenzhuang, Zhou, Zaida, Yao, Haotian, Yang, Hao, Gao, Hongcheng, Sui, Lin, Li, Weihong, Zu, Xinxing, Chen, Jia, Wang, Yao, Wu, Xiaoxue, Wang, Yalin, Charles, Y., Bao, Yiping, Liu, Yangyang, Huang, Zhiqi, Zhou, Xinyu
Abstract
We introduce PerceptionBench, a benchmark specifically designed to evaluate the atomic visual perception capabilities of Multimodal Large Language Models (MLLMs). Existing benchmarks often fail to isolate perception: holistic evaluations conflate perceptual errors with failures in reasoning or domain knowledge, while application-driven benchmarks only cover narrow, fragmented domains shaped by heuristic designs. To address these limitations, PerceptionBench adopts a bottom-up approach: by diagnosing the earliest failure points in the responses of frontier MLLMs across 42 existing benchmarks, we construct an error taxonomy whose perception branch defines ten atomic perceptual capabilities. Guided by this taxonomy, we construct 3,000 verified questions with short, unambiguous answers, each isolating a single capability, with difficulty stemming from perception rather than reasoning or knowledge. Benchmark results across sixteen frontier MLLMs reveal that atomic perception remains largely unsolved---no model reaches 60\% accuracy, perception-related hallucination is the weakest capability on average, and similar overall scores conceal sharply divergent capability profiles. PerceptionBench thus provides a capability-level standard for measuring and diagnosing the visual perception boundaries of MLLMs.
Chinese Translation
我们介绍了PerceptionBench，这是一个专门设计用于评估多模态大型语言模型（MLLMs）原子视觉感知能力的基准测试。现有的基准测试往往无法孤立感知：整体评估将感知错误与推理或领域知识的失败混为一谈，而应用驱动的基准测试仅涵盖由启发式设计塑造的狭窄、碎片化的领域。为了解决这些局限性，PerceptionBench采用自下而上的方法：通过诊断42个现有基准测试中前沿MLLMs响应的最早失败点，我们构建了一个错误分类法，其感知分支定义了十种原子感知能力。在这一分类法的指导下，我们构建了3000个经过验证的问题，每个问题都有简短、明确的答案，且每个问题孤立出单一能力，难度源于感知而非推理或知识。对十六个前沿MLLMs的基准测试结果显示，原子感知能力仍然在很大程度上未得到解决——没有模型达到60%的准确率，感知相关的幻觉是平均最弱的能力，且相似的整体得分掩盖了截然不同的能力特征。因此，PerceptionBench为测量和诊断MLLMs的视觉感知边界提供了能力级别的标准。
cs.CV / 7 / 2607.24981
Enabling Fully Integer-Only Inference for Lightweight Detection Transformers
实现轻量级检测变换器的完全整数推理
Abstract
Vision Transformer detectors now approach the accuracy of CNNs but remain difficult to deploy on NPUs and microcontrollers because key components, including deformable attention, feature fusion, and nonlinear activation functions, are not natively compatible with integer arithmetic. Existing quantized detectors either retain operators such as Softmax, GELU, and LayerNorm or focus on heavyweight backbones, leaving lightweight detection transformers without an end-to-end integer implementation. We address this gap with I-LW-DETR, the first fully integer-only lightweight DETR, in which every operation in the forward pass, including transformer nonlinearities, is executed in integer arithmetic. I-LW-DETR is built upon three key components: a scale-preserving split convolution that assigns independent activation scale to each branch of the multi-scale projector; SD-ShiftGELU, a sign-dependent GELU approximation that preserves element-wise behavior while avoiding the accuracy degradation; and a constrained Shiftmax that maintains stable Softmax normalization. Experimental results demonstrate that the proposed quantization pipeline consistently produces efficient fully integer-only models across different model scales. Across all model scales, the proposed pipeline incurs only a moderate accuracy degradation while reducing the model size by approximately $3.6\times$ and the computational cost by more than one order of magnitude.
Chinese Translation
视觉变换器检测器的准确性现已接近卷积神经网络（CNN），但由于关键组件（包括可变形注意力、特征融合和非线性激活函数）与整数算术不兼容，仍然难以在神经处理单元（NPU）和微控制器上部署。现有的量化检测器要么保留如 Softmax、GELU 和 LayerNorm 等运算符，要么专注于重量级骨干网络，导致轻量级检测变换器缺乏端到端的整数实现。我们通过 I-LW-DETR 来填补这一空白，这是首个完全基于整数的轻量级 DETR，其中前向传播中的每个操作（包括变换器的非线性部分）均以整数算术执行。I-LW-DETR 建立在三个关键组件之上：一个保持尺度的分裂卷积，为多尺度投影器的每个分支分配独立的激活尺度；SD-ShiftGELU，一种依赖符号的 GELU 近似，保持逐元素行为，同时避免精度下降；以及一个约束的 Shiftmax，保持稳定的 Softmax 归一化。实验结果表明，所提出的量化管道在不同模型规模下始终产生高效的完全整数模型。在所有模型规模中，所提出的管道仅导致适度的精度下降，同时将模型大小减少约 $3.6 imes$，计算成本降低超过一个数量级。
cs.CV / 8 / 2607.25014
Unifying Active Learning and Semi-Supervised Learning for Medical Image Segmentation
统一主动学习与半监督学习用于医学图像分割
Abstract
In practical settings, medical image segmentation models are often developed with limited annotated data rather than fully labeled datasets. Training frequently begins in ultra-low labeled regimes where only a small number of volumes are annotated. In such scenarios, practitioners must simultaneously decide which cases to annotate and how to best use the remaining unlabeled data. Although active learning (AL) and semi-supervised learning (SSL) both target annotation scarcity, they are typically designed and optimized independently, resulting in objective mismatch and unstable training during early-stage "cold start" conditions. We propose RegAL, a unified active semi-supervised framework governed by a shared topology-aware Pareto optimization that couples sample acquisition with unlabeled data utilization. RegAL evaluates images along three complementary axes, voxel-wise uncertainty, feature diversity, and a novel topological consistency metric, to select anatomically informative edge cases for annotation. On the other hand, the same criteria are used to identify geometrically stable atlas candidates for diffeomorphic registration-guided augmentation to train a self-supervised Mean Teacher segmentation network. Across BraTS 2021, dHCP, and ProstateX, RegAL remains stable with few labeled volumes and consistently outperforms state-of-the-art AL, SSL, and active semi-supervised baselines across Dice and boundary-distance (ASD, HD95) metrics under extreme annotation scarcity.
Chinese Translation
在实际应用中，医学图像分割模型通常是在有限的标注数据下开发的，而不是在完全标记的数据集上进行训练。训练通常始于超低标注状态，仅有少量体积被标注。在这种情况下，实践者必须同时决定哪些案例需要标注，以及如何最好地利用剩余的未标注数据。尽管主动学习（Active Learning, AL）和半监督学习（Semi-Supervised Learning, SSL）都针对标注稀缺问题，但它们通常是独立设计和优化的，导致目标不匹配和在早期“冷启动”条件下训练不稳定。我们提出了RegAL，一个统一的主动半监督框架，由共享的拓扑感知帕累托优化（Pareto Optimization）主导，将样本获取与未标注数据利用相结合。RegAL沿着三个互补轴评估图像：体素级不确定性、特征多样性和一种新颖的拓扑一致性度量，以选择解剖学上信息丰富的边缘案例进行标注。另一方面，使用相同的标准识别几何稳定的图谱候选，以进行基于微分同胚注册引导的增强，训练自监督的均值教师分割网络。在BraTS 2021、dHCP和ProstateX数据集上，RegAL在少量标注体积下保持稳定，并在极端标注稀缺条件下，在Dice和边界距离（ASD, HD95）指标上始终优于最先进的AL、SSL和主动半监督基线。
cs.CV / 9 / 2607.25078
Diff-ID: Identity Consistent Facial Image Generation and Morphing via Diffusion Models
Diff-ID：基于扩散模型的一致身份面部图像生成与变形
Abstract
Generative diffusion models have revolutionized facial image synthesis, yet robust identity preservation in high resolution outputs remains a critical challenge. This issue is especially vital for security systems, biometric authentication, and privacy sensitive applications, where any drift in identity integrity can undermine trust and functionality. We introduce Diff-ID, a diffusion based framework that enforces identity consistency while delivering photorealistic quality. Central to our approach is a custom 210K image dataset synthesized from CelebA-HQ, FFHQ, and LAION-Face and captioned via a fine tuned BLIP model to bolster identity awareness during training. Diff-ID integrates ArcFace and CLIP embeddings through a dual cross attention adapter within a fine tuned Stable Diffusion UNet. To further reinforce identity fidelity, we propose a pseudo discriminator loss based on ArcFace cosine similarity with exponential timestep weighting. Experiments on held out and unseen faces show that Diff-ID does not exceed InstantID in raw ArcFace Face Similarity, but achieves substantially lower FID and the strongest FIQ based identity--realism trade off among the evaluated methods. We also present a unified DDIM based morphing pipeline that enables qualitative facial interpolation without per identity fine tuning. We further argue that identity preservation and photorealism should be evaluated jointly rather than in isolation, as high identity similarity alone does not guarantee realistic outputs. To make this trade off explicit, we report Face Image Quality (FIQ) as a complementary ratio based score that combines identity similarity and perceptual realism while keeping FS and FID as the primary metrics.
Chinese Translation
生成性扩散模型已彻底改变了面部图像合成，但在高分辨率输出中保持稳健的身份一致性仍然是一个关键挑战。这个问题对于安全系统、生物识别认证和隐私敏感应用尤为重要，因为身份完整性的任何偏差都可能破坏信任和功能性。我们提出了Diff-ID，一个基于扩散的框架，旨在在提供照片级真实感的同时强制执行身份一致性。我们的方法的核心是一个从CelebA-HQ、FFHQ和LAION-Face合成的210K图像数据集，并通过微调的BLIP模型进行标注，以增强训练过程中的身份意识。Diff-ID通过在微调的Stable Diffusion UNet中集成ArcFace和CLIP嵌入，采用双重交叉注意力适配器。为了进一步增强身份保真度，我们提出了一种基于ArcFace余弦相似度的伪鉴别器损失，并采用指数时间步加权。对保留和未见面孔的实验表明，Diff-ID在原始ArcFace面部相似度上不超过InstantID，但在FID上显著降低，并在评估方法中实现了最强的FIQ基础身份-真实感权衡。我们还提出了一个统一的基于DDIM的变形管道，能够在不进行每个身份微调的情况下实现定性面部插值。我们进一步认为，身份保留和照片真实感应当共同评估，而不是孤立评估，因为单纯的高身份相似性并不能保证真实的输出。为了使这种权衡明确，我们报告了面部图像质量（FIQ）作为一种补充的基于比率的评分，结合了身份相似性和感知真实感，同时将FS和FID作为主要指标。
cs.CV / 10 / 2607.25092
MorphUNet: Alpha-Controlled Biometric Transport for Diffusion-Based Face Morphing Attacks
MorphUNet：基于扩散的人脸变形攻击的α控制生物特征传输
Abstract
Face morphing attacks create synthetic images verifiable against multiple identities, threatening border control and identity verification systems. We introduce MorphUNet, a diffusion morphing framework formulating two-parent generation as alpha-controlled biometric transport: each parent is decomposed into CLIP appearance and ArcFace identity evidence, aligned into a CLIP-compatible token space, with the two contributors preserved as separate identity-aware token banks. To our knowledge, MorphUNet is the first diffusion-based morphing framework using trainable parent-separated dual cross-attention inside the denoising U-Net: a Biometric Transport Layer carrying parent-specific identity evidence through denoising, attending to each parent separately before combining residuals via the morphing parameter alpha. DDIM-inverted latent interpolation gives a coherent denoising start, while weaker-parent-guided selection favours morphs maximising the lower parent-similarity score, reducing collapse toward one contributor. We evaluate MorphUNet against three state-of-the-art baselines (StableMorph, MIPGAN-II, and MorDIFF) on FEI and FRLL using six recognition systems, and propose CFD-based unseen-identity stress testing across gender and ethnicity pairing, demographic shifts, and parent-similarity extremes. MorphUNet achieves the best Morphing Attack Potential (MAP) when at least three of six systems are fooled by one morph, reaching 0.919 on FEI and 0.886 on FRLL, and obtains the best FID on both datasets (35.19 FEI, 44.86 FRLL). It also gives the highest APCER at 5% BPCER in the same-dataset setting, and remains highly difficult to detect under cross-dataset transfer, with APCER 0.996 on FEI and 0.946 on FRLL. The full evaluation analyses MAP, MAD, per-system vulnerability, identity balance, image quality, top/bottom-similarity stress tests, and CFD unseen-identity robustness.
Chinese Translation
人脸变形攻击生成可验证多个身份的合成图像，威胁边境控制和身份验证系统。我们提出了MorphUNet，一种扩散变形框架，将双亲生成形式化为α控制的生物特征传输：每个父母被分解为CLIP外观和ArcFace身份证据，排列到CLIP兼容的令牌空间中，两个贡献者作为独立的身份感知令牌库保留。据我们所知，MorphUNet是第一个在去噪U-Net中使用可训练的父母分离双重交叉注意力的基于扩散的变形框架：生物特征传输层在去噪过程中携带特定于父母的身份证据，分别关注每个父母，然后通过变形参数α结合残差。DDIM反演潜在插值提供了一个连贯的去噪起始，而较弱父母引导的选择有利于最大化较低父母相似性评分的变形，减少向一个贡献者的崩溃。我们在FEI和FRLL上使用六个识别系统评估MorphUNet，并提出基于CFD的未见身份压力测试，涵盖性别和种族配对、人口变化和父母相似性极端情况。MorphUNet在至少六个系统中有三个被一个变形欺骗时，达到了最佳的变形攻击潜力（MAP），在FEI上达到0.919，在FRLL上达到0.886，并在两个数据集上获得最佳FID（FEI 35.19，FRLL 44.86）。在同一数据集设置下，它在5% BPCER时也给出了最高的APCER，并且在跨数据集转移下仍然非常难以检测，FEI上的APCER为0.996，FRLL上的APCER为0.946。完整评估分析了MAP、MAD、每个系统的脆弱性、身份平衡、图像质量、顶部/底部相似性压力测试和CFD未见身份的鲁棒性。
cs.CV / 11 / 2607.25106
IMPRINT: Image-Conditioned Query Enrichment for Long-Tail Object Goal Navigation
IMPRINT：基于图像的查询增强用于长尾物体目标导航
Abstract
Embodied AI increasingly relies on queryable semantic maps built from pre-trained vision-language models to enable zero-shot Object Goal Navigation (ObjectNav). However, existing approaches typically depend on text-only queries, which become less reliable as semantic specificity increases toward fine-grained object categories. We introduce IMPRINT, a zero-shot plug-and-play framework that enriches textual object queries with web-sourced images to improve grounding in queryable maps. Retrieved images are encoded using a vision-language model, matched against the semantic map to produce similarity maps, and aggregated to yield context-aware localization. Notably, this requires no training or modification of the underlying navigation policy. To explicitly evaluate long-tail behavior, we present HSSD-rare, a new ObjectNav benchmark built on Habitat Synthetic Scenes and featuring semantically specific subcategories. Across both OVON and HSSD-rare, image-conditioned queries consistently improve object grounding and yield end-to-end navigation gains. Further analysis reveals that translating localization gains to navigation performance depends critically on downstream detection quality, highlighting a key systems bottleneck in long-tail embodied navigation.
Chinese Translation
具身人工智能越来越依赖于从预训练视觉-语言模型构建的可查询语义地图，以实现零样本物体目标导航（ObjectNav）。然而，现有方法通常依赖于仅文本的查询，随着语义特异性的增加，尤其是针对细粒度物体类别时，这些查询的可靠性下降。我们提出了IMPRINT，一个零样本即插即用框架，通过网络获取的图像来丰富文本物体查询，从而改善在可查询地图中的基础定位。检索到的图像使用视觉-语言模型进行编码，并与语义地图匹配以生成相似性地图，最终聚合以实现上下文感知的定位。值得注意的是，这一过程无需对底层导航策略进行训练或修改。为了明确评估长尾行为，我们提出了HSSD-rare，一个基于Habitat合成场景的新ObjectNav基准，具有语义特定的子类别。在OVON和HSSD-rare中，基于图像的查询始终改善了物体定位，并带来了端到端的导航提升。进一步分析表明，将定位提升转化为导航性能在很大程度上依赖于下游检测质量，这突显了长尾具身导航中的关键系统瓶颈。
cs.CV / 12 / 2607.25108
OPERA: Offline Policy-guided Expert Routing and Adaptation for Universal Biomedical Image Analysis
OPERA：用于通用生物医学图像分析的离线策略引导专家路由与适应
Abstract
Biomedical image analysis spans diverse modalities and tasks, yet real-world deployment is hindered by severe distribution shifts across scanners, protocols, and patient populations. High-performing models consequently require repeated domain-specific fine-tuning, which is a costly cycle that becomes impractical when labels are scarce or privacy constraints limit data sharing. We propose OPERA (Offline Policy-guided Expert Routing and Adaptation), a multi-agent ensemble framework that addresses this deployment bottleneck by treating expert weight assignment as an offline policy learning problem: a routing policy is learned from a small validation set without gradient updates to any expert agent, then deployed with test-time adaptation to handle distribution shift. OPERA coordinates heterogeneous specialist agents through complementary mechanisms. The expert profiling module learns selection policies offline, enabling informed allocation of expertise. Each agent undergoes confidence calibration through temperature adjustment, ensuring more reliable probabilistic outputs. OPERA also incorporates distribution aware adaptation, where class weights are dynamically adjusted at the batch level using statistics derived from unlabeled test data. Instance level routing assigns each sample to the most suitable expert by leveraging inter model agreement and predictive entropy. We evaluate OPERA on 9 datasets covering fundus photography, chest X-ray, CT, MRI, and multimodal diagnostic benchmarks, comparing against 30+ baselines across classification, segmentation, and multimodal settings. OPERA consistently improves performance and calibration quality, demonstrating that offline policy-guided expert agents coordination is a practical path to deployable biomedical AI without retraining. Code is on \href{https://github.com/HUANGLIZI/OPERA}{GitHub}.
Chinese Translation
生物医学图像分析涵盖多种模态和任务，但在实际应用中，由于扫描仪、协议和患者群体之间存在严重的分布变化，部署受到阻碍。因此，高性能模型需要反复进行特定领域的微调，这是一种成本高昂的循环，当标签稀缺或隐私限制数据共享时，这种方法变得不切实际。我们提出了OPERA（Offline Policy-guided Expert Routing and Adaptation），一个多智能体集成框架，通过将专家权重分配视为离线策略学习问题来解决这一部署瓶颈：从小型验证集学习路由策略，而不对任何专家智能体进行梯度更新，然后在测试时进行适应以应对分布变化。OPERA通过互补机制协调异构专家智能体。专家画像模块离线学习选择策略，使得能够进行更为合理的专业知识分配。每个智能体通过温度调整进行置信度校准，确保更可靠的概率输出。OPERA还结合了分布感知适应，在此过程中，使用来自未标记测试数据的统计信息动态调整批次级别的类别权重。实例级路由通过利用模型间一致性和预测熵，将每个样本分配给最合适的专家。我们在涵盖眼底摄影、胸部X光、CT、MRI和多模态诊断基准的9个数据集上评估了OPERA，并与30多个基线进行比较，涉及分类、分割和多模态设置。OPERA始终提高了性能和校准质量，证明了离线策略引导的专家智能体协调是实现可部署生物医学人工智能的切实路径，而无需重新训练。代码可在 exttt{https://github.com/HUANGLIZI/OPERA} 找到。
cs.CV / 13 / 2607.25125
LENS: Adaptive Spatio-Temporal Zooming for Keyframe Sampling in Long-Form Videos
LENS：用于长视频关键帧采样的自适应时空缩放
Abstract
Despite rapid progress in Multi-modal Large Language Models (MLLMs), understanding long-form videos is still bottlenecked by limited context windows. While recent keyframe sampling methods attempt to mitigate this by distilling video inputs into a compact set of query-relevant frames, navigating the vast spatio-temporal search space remains challenging, as spatial detail and temporal coverage often conflict. To address this, we introduce LENS, a training-free keyframe sampling framework that dynamically decides when to zoom in for fine-grained details and when to zoom out for broader context based on the text query. Concretely, LENS adaptively allocates a limited frame budget between spatial zoom-ins, which highlight query-relevant regions within individual frames, and temporal zoom-outs, which expand the temporal scope through multi-frame aggregation, enabling the model to reason across multiple granularities while capturing both high-fidelity details and long-range context. Across diverse long-form video benchmarks, LENS consistently outperforms prior state-of-the-art keyframe sampling methods and delivers substantial gains over uniform sampling, improving Video-MME accuracy from 53.3% to 60.7% with Qwen2.5-VL.Code is available at https://github.com/zhangce01/LENS.
Chinese Translation
尽管多模态大型语言模型（MLLMs）取得了快速进展，但理解长视频仍然受到有限上下文窗口的瓶颈。尽管近期的关键帧采样方法试图通过将视频输入提炼为一组紧凑的与查询相关的帧来缓解这一问题，但在广阔的时空搜索空间中导航仍然具有挑战性，因为空间细节和时间覆盖往往存在冲突。为了解决这个问题，我们提出了LENS，这是一种无训练的关键帧采样框架，它根据文本查询动态决定何时放大以获取细粒度细节，何时缩小以获取更广泛的上下文。具体而言，LENS自适应地在空间放大（突出显示单个帧中与查询相关的区域）和时间缩小（通过多帧聚合扩展时间范围）之间分配有限的帧预算，使模型能够在捕捉高保真细节和长距离上下文的同时，在多个粒度上进行推理。在各种长视频基准测试中，LENS始终优于之前的最先进的关键帧采样方法，并在均匀采样上实现了显著提升，将Video-MME的准确率从53.3%提高到60.7%，使用的模型为Qwen2.5-VL。代码可在 https://github.com/zhangce01/LENS 获取。
cs.CV / 14 / 2607.25146
FIDAC: An Easy-to-use Pipeline to Extract and Interpret Interpersonal Distance From Video
FIDAC：一个易于使用的管道，用于从视频中提取和解释人际距离
Abstract
The distance between persons reveals significant information about their perception of each other. However, such information is not easily extractable and interpretable from video input. We developed an open-sourced library, Facial Interpersonal Distance Analysis and Coding (FIDAC) that transforms facial detection results into actionable data about location and interpersonal distance. This tool merges data from multiple open-source facial detection models, strategically compensating for gaps in any individual model. In addition, we include methods for more accurate tracking, such as a pipeline for human coding of the selection of faces and a benchmarking tool to reduce depth distortion. For next steps, we plan on building upon FIDAC by evaluating its effectiveness at measuring interpersonal distance at various depths and orientations while further integrating features of proxemic analysis such as synchrony into its software.
Chinese Translation
人与人之间的距离揭示了他们对彼此的感知的重要信息。然而，这些信息并不容易从视频输入中提取和解释。我们开发了一个开源库，面部人际距离分析与编码（Facial Interpersonal Distance Analysis and Coding，FIDAC），该库将面部检测结果转化为关于位置和人际距离的可操作数据。该工具整合了多个开源面部检测模型的数据，战略性地弥补了任何单一模型的不足。此外，我们还包括了更准确跟踪的方法，例如人类编码面部选择的管道和减少深度失真基准工具。接下来的步骤，我们计划在FIDAC的基础上进行扩展，评估其在不同深度和方向下测量人际距离的有效性，同时进一步将同步等亲密分析的特征集成到其软件中。
cs.CV / 15 / 2607.25153
OpenPVMapper: A Multi-source, Nationwide Database of Rooftop Photovoltaic Systems in France
OpenPVMapper：法国屋顶光伏系统的多源全国数据库
Abstract
Rooftop photovoltaic (PV) systems account for the vast majority of PV grid connections, yet no open, comprehensive, installation-level dataset of these systems exists: public registries aggregate data only above a capacity threshold, and remote sensing-based detection efforts, while extensive, are typically confined to a single method, a limited geographic scope, or a single point in time. We introduce OpenPVMapper, a nationwide, multi-source database of rooftop PV installations in mainland France, built by aggregating and reconciling complementary sources: a deep learning-based detection pipeline deployed on nationwide aerial imagery, OpenStreetMap and a probabilistic building-level detection dataset. The resulting database contains 1,135,850 installations, totaling approximately 15.01~GWp of installed capacity, each documented with its provenance, detection method, and, where available, a manual validation flag. Manual review of a stratified sample of 1,862 installations places the database's overall precision at approximately 74--75\%, with corroboration across independent sources bringing a substantial, quantified precision gain. By aggregating independent sources rather than relying on any single detection method, OpenPVMapper reaches a level of confidence beyond what any one source could provide on its own, while remaining extensible to further sources as they become available. It is released under an open (CC-BY) license alongside the full source code used to build it.
Chinese Translation
屋顶光伏（PV）系统占据了绝大多数的光伏电网连接，但目前尚不存在一个开放的、全面的、安装级别的数据集：公共登记处仅汇总超过容量阈值的数据，而基于遥感的检测工作虽然广泛，但通常局限于单一方法、有限的地理范围或单一时间点。我们介绍了OpenPVMapper，这是一个覆盖法国本土的全国性多源屋顶光伏安装数据库，旨在通过汇总和协调互补的数据源构建而成：一个基于深度学习的检测管道，部署在全国范围的航空影像上，OpenStreetMap以及一个概率建筑级别的检测数据集。最终数据库包含1,135,850个安装，总计约15.01 GWp的装机容量，每个安装都有其来源、检测方法的记录，并在可用时标注手动验证标志。对1,862个安装的分层样本进行的人工审查显示，该数据库的整体精度约为74-75%，而独立来源之间的相互验证带来了显著的、量化的精度提升。通过汇总独立来源而非依赖于任何单一检测方法，OpenPVMapper达到了超出任何单一来源所能提供的信心水平，同时在未来可扩展至更多来源。该数据库以开放（CC-BY）许可证发布，并附带构建所用的完整源代码。
cs.CV / 16 / 2607.25164
OrganLens: Organ-Specific Representation Learning for CT Foundation Models
OrganLens：用于CT基础模型的器官特定表示学习
Abstract
A CT examination captures multiple organs, but many biomedical questions concern abnormalities, prognosis, or longitudinal change in a specific organ. These questions require a separate representation for each organ within the same CT volume. Existing CT foundation models commonly produce a single volume-level representation, while recent anatomy-aware methods either encode pre-separated organ volumes or explicitly disentangle images into organ token groups. The former may remove clinically relevant surrounding context, while the latter does not condition a shared encoder on a selected organ before its features are formed. We introduce OrganLens for organ-specific representation learning through self-supervision. An organ identity conditions a shared CT encoder, while organ-specific distillation and anatomy-mask supervision shape features for anatomy-weighted pooling into organ-specific representations. At inference, the shared model produces 11 organ-specific representations without external segmentation masks. We evaluate OrganLens on CT-RATE, RAD-ChestCT, INSPECT, and NLST across diverse acquisitions and downstream evaluations. Relative to CT-pretrained DINOv2, heart representations raise CT-RATE cardiomegaly AUROC from 0.910 to 0.953, while lung representations improve the Harrell C-index for NLST lung-cancer mortality by 14.2\%. The global representation reaches INSPECT Recall@10 of 33.09\% and 32.04\% for text-to-image and image-to-text retrieval, respectively. Across organ-related tasks, anatomically matched representations provide stronger task-relevant signal, while the global representation retains broad utility. OrganLens offers a scalable approach to organ-specific CT representation learning with a shared encoder. More broadly, it provides the medical research community with a reusable framework for studying organ-specific disease across cohorts and clinical endpoints.
Chinese Translation
CT检查捕获多个器官，但许多生物医学问题涉及特定器官的异常、预后或纵向变化。这些问题需要在同一CT体积内为每个器官提供单独的表示。现有的CT基础模型通常生成单一的体积级表示，而最近的解剖学感知方法要么编码预先分离的器官体积，要么明确将图像解构为器官标记组。前者可能会去除临床相关的周围背景，而后者在形成特征之前并未对选定器官进行共享编码器的条件化。我们提出了OrganLens，通过自我监督实现器官特定的表示学习。器官身份条件化一个共享的CT编码器，而器官特定的蒸馏和解剖掩码监督则塑造特征，以进行解剖加权池化，形成器官特定的表示。在推理时，共享模型在没有外部分割掩码的情况下生成11个器官特定的表示。我们在CT-RATE、RAD-ChestCT、INSPECT和NLST上评估OrganLens，涵盖多种采集方式和下游评估。相较于CT预训练的DINOv2，心脏表示将CT-RATE心脏肥大AUROC从0.910提高到0.953，而肺部表示则使NLST肺癌死亡率的Harrell C-index提高了14.2%。全球表示在INSPECT Recall@10中达到了33.09%，在文本到图像和图像到文本检索中分别为32.04%。在与器官相关的任务中，解剖匹配的表示提供了更强的任务相关信号，而全球表示则保持了广泛的实用性。OrganLens提供了一种可扩展的方法，通过共享编码器实现器官特定的CT表示学习。更广泛地说，它为医学研究社区提供了一个可重用的框架，以研究跨队列和临床终点的器官特定疾病。
cs.CV / 17 / 2607.25197
LGFNet: A CTC-Guided Local-Global Fusion Framework for Single-Channel Sleep Staging
LGFNet：一种基于CTC指导的单通道睡眠分期局部-全局融合框架
Abstract
Sleep staging remains challenging due to long-range temporal dependencies, ambiguous stage transitions-particularly in N1-and substantial distribution shifts across subjects, sampling rates, and EEG montages. These difficulties are further amplified in single-channel, low-latency scenarios required by wearable and real-world applications. To address these issues, we propose LGFNet, a CTC-guided sequence-to-sequence framework for robust sleep staging. LGFNet introduces a Local-Global Fusion encoder that jointly models fine-grained temporal dynamics and long-range sleep structure, overcoming the limitations of conventional serial hybrid architectures. A CTC-Attention joint training paradigm is adopted to unify temporal alignment with context-dependent modeling, enabling more accurate recognition of stage boundaries and transitions. Furthermore, a three-stage decoding strategy is devised, leveraging CTC-guided decoding and Viterbi-based smoothing to reduce error accumulation and enforce physiological consistency. Extensive cross-dataset evaluations on five public benchmarks demonstrate that LGFNet consistently outperforms state-of-the-art single-channel methods. In particular, on Sleep-EDF-78, LGFNet surpasses DMIN by +1.27% accuracy, +1.74% macro-F1, and +1.93% kappa, with pronounced gains on N1 and transition segments, highlighting its robustness and strong generalization across diverse sampling rates, montages, and recording environments.
Chinese Translation
睡眠分期仍然面临挑战，主要由于长时间的时间依赖性、模糊的阶段过渡（特别是在N1阶段）以及在不同受试者、采样率和脑电图（EEG）布置之间的显著分布变化。这些困难在可穿戴设备和实际应用所需的单通道低延迟场景中进一步加剧。为了解决这些问题，我们提出了LGFNet，一种基于CTC指导的序列到序列框架，用于稳健的睡眠分期。LGFNet引入了一种局部-全局融合编码器，联合建模细粒度的时间动态和长范围的睡眠结构，克服了传统串行混合架构的局限性。采用CTC-注意力联合训练范式，将时间对齐与上下文依赖建模统一起来，从而实现对阶段边界和过渡的更准确识别。此外，设计了一种三阶段解码策略，利用CTC指导的解码和基于维特比（Viterbi）的平滑来减少错误累积并强化生理一致性。在五个公共基准上的广泛跨数据集评估表明，LGFNet始终优于最先进的单通道方法。特别是在Sleep-EDF-78数据集上，LGFNet的准确率比DMIN提高了1.27%，宏观F1提高了1.74%，Kappa值提高了1.93%，在N1和过渡段上表现出显著的提升，突显了其在不同采样率、布置和记录环境中的稳健性和强泛化能力。
cs.CV / 18 / 2607.25210
ObliCity: A Benchmark and Baseline for Roof-to-Ground Projection Displacement Correction
ObliCity：屋顶到地面投影位移校正的基准与基线
Abstract
Oblique-view urban remote sensing imagery inevitably exhibits geometric projection displacements between building roofs and footprints, leading to significant distortions in spatial structure. Existing approaches either ignore these deformations or handle them implicitly within segmentation-based frameworks, where progress is dominated by general segmentation advances rather than improvements in geometric correction. In this work, we explicitly define roof-to-footprint offset vector (RFOV) extraction as an independent learning task that decouples geometric alignment from semantic segmentation. To support this task, we introduce the Oblique City dataset (ObliCity), the first large-scale benchmark that integrates high-resolution UAV imagery and globally distributed satellite data, covering diverse city morphologies and camera perspectives. Methodologically, we reformulate DragOSM into DragRoof, an ODE-based framework inspired by human annotation behavior. By simulating the continuous process of dragging roofs toward their footprints, DragRoof learns deterministic, geometry-consistent offset fields and adaptively determines convergence through an end token. Extensive experiments on ObliCity demonstrate that DragRoof achieves state-of-the-art RFOV extraction performance, requiring fewer inference steps while delivering superior directional and length accuracy. Our dataset and model establish a principled foundation for studying projection displacement correction in oblique remote sensing imagery. The source code and dataset will be avaliable at https://github.com/likaiucas/DragRoof.
Chinese Translation
倾斜视角的城市遥感影像不可避免地表现出建筑屋顶与地面轮廓之间的几何投影位移，导致空间结构出现显著扭曲。现有的方法要么忽略这些变形，要么在基于分割的框架中隐式处理这些变形，其中进展主要依赖于一般分割的进步，而非几何校正的改进。在本研究中，我们明确将屋顶到地面轮廓偏移向量（RFOV）提取定义为一个独立的学习任务，从而将几何对齐与语义分割解耦。为支持这一任务，我们引入了Oblique City数据集（ObliCity），这是第一个整合高分辨率无人机影像和全球分布卫星数据的大规模基准，涵盖了多样的城市形态和相机视角。在方法论上，我们将DragOSM重新构建为DragRoof，这是一个基于常微分方程（ODE）的框架，灵感来源于人类标注行为。通过模拟将屋顶拖向其地面轮廓的连续过程，DragRoof学习确定性、几何一致的偏移场，并通过结束标记自适应地确定收敛。对ObliCity的广泛实验表明，DragRoof在RFOV提取性能上达到了最先进水平，所需推理步骤更少，同时提供了优越的方向和长度精度。我们的数据集和模型为研究倾斜遥感影像中的投影位移校正奠定了原则性基础。源代码和数据集将可在 https://github.com/likaiucas/DragRoof 获取。
cs.CV / 19 / 2607.25234
WHTMix: Efficient Stereo Depth Estimation via Walsh-Hadamard Token Mixing
WHTMix：通过Walsh-Hadamard令牌混合实现高效立体深度估计
Abstract
Stereo depth estimation for driving, robotics and augmented reality must run at high resolution under tight latency budgets, yet in transformer-based matchers the global self-attention that aggregates scene context grows quadratically with the number of pixels and comes to dominate runtime. We show that the joint self-attention stage of a stereo transformer, whose role is to spread context across both views, can be replaced by a data-independent Walsh-Hadamard token mixer that mixes tokens globally in the transform domain at log-linear cost, while the data-dependent cross-attention that performs left-right correspondence is retained. On synthetic driving data the mixer matches the attention baseline in end-point error while reducing model compute by a factor of 2.46 and single-image inference latency by a factor of 2.65. A complexity analysis shows the benefit is governed by the ratio of sequence length to channel width, which explains why high-resolution stereo matching is a particularly favorable setting and why classification transformers are not; we confirm this token-to-channel scaling on non-stereo long-sequence benchmarks. Furthermore, we introduce a hybrid log-disparity loss function designed to up-weight small-disparity pixels corresponding to long-range objects. This approach reduces the error on distant objects without incurring any additional computational overhead.
Chinese Translation
立体深度估计在驾驶、机器人技术和增强现实中必须在高分辨率下运行，并且具有严格的延迟预算。然而，在基于变换器的匹配器中，全局自注意力机制用于聚合场景上下文，其计算复杂度随着像素数量的增加呈二次增长，并主导了运行时间。我们展示了立体变换器的联合自注意力阶段，其作用是跨两个视图传播上下文，可以被一个数据无关的Walsh-Hadamard令牌混合器所替代，该混合器在变换域中以对数线性成本全局混合令牌，同时保留执行左右对应关系的数据相关交叉注意力。在合成驾驶数据上，该混合器在端点误差方面与注意力基线相匹配，同时将模型计算量减少了2.46倍，单图像推理延迟减少了2.65倍。复杂度分析表明，这一好处由序列长度与通道宽度的比率决定，这解释了为什么高分辨率立体匹配是一个特别有利的设置，而分类变换器则不是；我们在非立体长序列基准上确认了这种令牌与通道的缩放。此外，我们引入了一种混合对数视差损失函数，旨在对对应于远程物体的小视差像素进行加权。这种方法在不增加任何额外计算开销的情况下，减少了远处物体的误差。
cs.CV / 20 / 2607.25239
CD-RMOT-Bench: Benchmarking the Cross-Domain Referring Multi-Object Tracking
CD-RMOT-Bench：跨域引用多目标跟踪基准测试
Abstract
Referring multi-object tracking (RMOT) extends tracking from category-driven perception to language-guided understanding by grounding object trajectories in natural-language expressions. Despite recent progress, existing RMOT studies are largely conducted under in-domain settings, leaving the robustness of language-conditioned tracking under inevitable visual domain shifts unexplored. In this paper, we study Cross-Domain Referring Multi-Object Tracking (CD-RMOT), a new and challenging problem that evaluates whether an RMOT model trained on a labeled source domain can reliably follow natural-language expressions in an unlabeled target domain with different visual conditions. To support systematic study, we construct CD-RMOT-Bench, a unified benchmark that combines real clear-domain referring tracking data, aligned digital-twin variants, and real adverse-domain videos. CD-RMOT-Bench enables both controlled weather/viewpoint shift analysis and realistic synthetic-real transfer evaluation under a shared RMOT protocol. Further, we provide a Query-Centric Adaptation (QCA) framework, designed to stabilize the query space that bridges visual trajectories and referring expressions. Extensive experiments reveal that domain shifts severely degrade RMOT performance, where the failure is not merely caused by object detection errors but more critically by unstable expression-conditioned temporal association and target selection. QCA establishes a strong baseline, while CD-RMOT-Bench opens a new direction for robust language-guided tracking across visual domains.
Chinese Translation
引用多目标跟踪（RMOT）将跟踪从以类别为驱动的感知扩展到以语言为指导的理解，通过将物体轨迹与自然语言表达相结合。尽管近期取得了一定进展，现有的RMOT研究大多是在同域设置下进行的，尚未探讨在不可避免的视觉域变化下语言条件跟踪的鲁棒性。本文研究了跨域引用多目标跟踪（CD-RMOT），这是一个新的且具有挑战性的问题，评估在标注源域上训练的RMOT模型是否能够在具有不同视觉条件的未标注目标域中可靠地跟随自然语言表达。为了支持系统性研究，我们构建了CD-RMOT-Bench，一个统一的基准，结合了真实清晰域的引用跟踪数据、对齐的数字双胞胎变体和真实的不利域视频。CD-RMOT-Bench支持在共享的RMOT协议下进行受控天气/视角变化分析和现实的合成-真实迁移评估。此外，我们提供了一个查询中心适应框架（QCA），旨在稳定连接视觉轨迹和引用表达的查询空间。大量实验表明，域变化严重降低了RMOT性能，失败不仅仅是由物体检测错误引起的，更关键的是由不稳定的表达条件时间关联和目标选择引起的。QCA建立了一个强基线，而CD-RMOT-Bench为跨视觉域的鲁棒语言指导跟踪开辟了新的方向。
cs.CV / 21 / 2607.25242
Medical world models in healthcare: foundations, applications, and challenges for trustworthy clinical translation
医疗领域模型在医疗保健中的应用：可信临床转化的基础、应用与挑战
Abstract
Medical world models offer a framework for extending medical artificial intelligence beyond static prediction by representing evolving patient states and modelling how they change over time and in response to clinical interventions. This Review defines the conceptual boundaries, technical foundations, application domains, and evidence requirements of the field through a structured narrative synthesis with reproducible evidence mapping.We screened 1,455 unique records and assembled a corpus of 98 sources, including 14 studies that met a strict empirical definition of a medical world model. The field is organised around four capabilities: patient state representation, temporal dynamics modelling, intervention-conditioned simulation, and clinician-supervised planning. Evidence spans medical imaging, longitudinal electronic health records, treatment response modelling, physiological and multimodal state modelling, ultrasound and surgical interaction, and population and health-system simulation; clinical digital twins are treated as a cross-cutting integration framework.Current studies provide early evidence of technical feasibility for trajectory forecasting and comparison of candidate interventions, but most remain retrospective, task-specific, or preclinical. The evidence base is further limited by incomplete longitudinal intervention data, inconsistent action semantics, limited causal identifiability, long-horizon error accumulation, inadequate uncertainty estimation, and limited external validation. Clinical translation will therefore depend on precise intervention representations, robust causal and mechanistic grounding, calibrated trajectory-level uncertainty, safety-constrained planning, and prospective multicentre validation against clinically meaningful endpoints.
Chinese Translation
医疗领域模型提供了一种框架，通过表示不断变化的患者状态以及建模这些状态如何随时间和临床干预而变化，从而将医疗人工智能的应用扩展超越静态预测。本综述通过结构化叙述综合和可重复的证据映射，定义了该领域的概念边界、技术基础、应用领域和证据要求。我们筛选了1,455条独特记录，汇编了98个来源的文献，包括14项符合医疗领域模型严格实证定义的研究。该领域围绕四个能力组织：患者状态表示、时间动态建模、干预条件模拟和临床医生监督规划。证据涵盖医学影像、纵向电子健康记录、治疗反应建模、生理和多模态状态建模、超声与外科互动，以及人口和健康系统模拟；临床数字双胞胎被视为一种跨领域的集成框架。目前的研究提供了轨迹预测和候选干预比较的技术可行性早期证据，但大多数研究仍然是回顾性的、任务特定的或前临床的。证据基础还受到不完整的纵向干预数据、不一致的行动语义、有限的因果可识别性、长期误差积累、不充分的不确定性估计和有限的外部验证的进一步限制。因此，临床转化将依赖于精确的干预表示、稳健的因果和机制基础、校准的轨迹级不确定性、安全约束的规划，以及针对临床有意义终点的前瞻性多中心验证。
cs.CV / 22 / 2607.25266
FORGE: Frame Orthogonality in Relevance Geometry for Long-Form Video Understanding
FORGE：长视频理解中的相关几何框架正交性
Abstract
Multimodal large language models (MLLMs) have enabled long-form video understanding at a scale that was not previously possible. However, the density of relevant content decreases sharply as video sequence length increases, and exposing the model to more irrelevant content measurably reduces its accuracy. In this paper, we address the problem of maximizing query-relevant information in a frame subset selected at inference time, without training. FORGE (Frame Orthogonality in Relevance Geometry) is a model-agnostic method that induces a query-conditioned geometry on a pretrained multimodal embedding space, unifying relevance and diversity into a single objective. In this space, frames that cover independent query-relevant directions are far apart, and selecting the subset of maximum information captures diverse query-relevant content within the budget. Experiments on Video-MME and LongVideoBench at budgets of 16, 32, and 64 frames show that FORGE improves the unified keyframe selection score by 11.0-15.3 points over the strongest training-free baseline and up to doubles keyframe recall (0.415 vs. 0.204 at K=64 on Video-MME). The gains extend to question answering, where accuracy improves in every evaluated setting across eight open-source MLLMs spanning 4B to 32B parameters, by up to 8.7 points over uniform sampling and 5.2 points over the strongest baseline. Our findings suggest that aligning the embedding space with the query's high-dimensional structure is a promising direction for inference-time video understanding.
Chinese Translation
多模态大型语言模型（MLLMs）使得长视频理解达到了之前无法实现的规模。然而，随着视频序列长度的增加，相关内容的密度急剧下降，向模型暴露更多无关内容会显著降低其准确性。本文解决了在推理时选择的帧子集中最大化查询相关信息的问题，而无需训练。FORGE（相关几何中的帧正交性）是一种与模型无关的方法，它在预训练的多模态嵌入空间中引入了查询条件几何，将相关性和多样性统一为一个目标。在这个空间中，覆盖独立查询相关方向的帧相距较远，选择最大信息的子集能够在预算内捕获多样的查询相关内容。在Video-MME和LongVideoBench上进行的实验，预算为16、32和64帧，显示FORGE在最强的无训练基线之上提高了统一关键帧选择分数11.0-15.3分，并且关键帧召回率提高了近一倍（在Video-MME上K=64时为0.415对比0.204）。这些提升也扩展到问答任务，在八个开源MLLMs（参数从4B到32B）中，每个评估设置的准确性都有所提高，相较于均匀采样提高了最多8.7分，相较于最强基线提高了5.2分。我们的研究结果表明，将嵌入空间与查询的高维结构对齐是推理时视频理解的一个有前景的方向。
cs.CV / 23 / 2607.25275
ScaleResfusion: Residual Rectified Flow based on Residual Vector Field
ScaleResfusion：基于残差向量场的残差校正流
Abstract
Real-world Image Restoration (Real-IR) aims to recover high-quality (HQ) images from complex and unknown degradations. Although recent diffusion-based methods have substantially improved perceptual quality, their current designs leave two key challenges unresolved. Methods that start from Gaussian noise are slow and often less faithful to the degraded input. Residual-based methods usually train from scratch, which makes it hard to exploit modern pre-trained generative priors. In this paper, we present ScaleResfusion, a scalable diffusion framework for real-world image restoration built on pre-trained text-to-image rectified-flow models. The core of our method is Residual Rectified Flow, which introduces the residual term R into Standard Rectified Flow. Instead of starting from pure noise, it uses a residual transport path that starts from noisy low-quality (LQ) images and admits an exact acceleration point. By learning the residual vector field, Residual Rectified Flow keeps the output distribution and linear diffusion process consistent with the pre-trained rectified-flow models. This makes parameter-efficient fine-tuning possible at scale. We further introduce a knowledge-distillation pipeline to reduce sampling cost while maintaining restoration quality. Extensive experiments on multiple real-world restoration tasks show that ScaleResfusion achieves state-of-the-art performance with much higher efficiency. These results suggest a practical and scalable way to adapt large pre-trained diffusion models to real-world image restoration. Our code and models are available at https://github.com/YukinoshitaLove/ScaleResfusion.
Chinese Translation
现实世界图像恢复（Real-IR）旨在从复杂且未知的退化中恢复高质量（HQ）图像。尽管最近的基于扩散的方法在感知质量上有了显著改善，但它们当前的设计仍然存在两个关键挑战。从高斯噪声开始的方法速度较慢，并且通常对退化输入的忠实度较低。基于残差的方法通常从头开始训练，这使得利用现代预训练生成先验变得困难。在本文中，我们提出了ScaleResfusion，这是一个基于预训练文本到图像校正流模型的可扩展扩散框架。我们方法的核心是残差校正流（Residual Rectified Flow），它将残差项R引入标准校正流（Standard Rectified Flow）。它不是从纯噪声开始，而是使用从噪声低质量（LQ）图像开始的残差传输路径，并允许精确的加速点。通过学习残差向量场，残差校正流保持输出分布和线性扩散过程与预训练的校正流模型一致。这使得在大规模下进行参数高效的微调成为可能。我们进一步引入了知识蒸馏管道，以降低采样成本，同时保持恢复质量。在多个现实世界恢复任务上的广泛实验表明，ScaleResfusion以更高的效率实现了最先进的性能。这些结果表明了一种将大型预训练扩散模型适应于现实世界图像恢复的实用且可扩展的方法。我们的代码和模型可在 https://github.com/YukinoshitaLove/ScaleResfusion 获取。
cs.CV / 24 / 2607.25276
FunnelAL: Retrieve-then-Rank Active Learning for Single-Class Discovery
FunnelAL：用于单类发现的检索-再排序主动学习
Abstract
We present FunnelAL, a retrieve-then-rank active learning system for single-class discovery, which adapts the multi-stage funnel architecture of industrial recommender systems to data annotation. Large-scale supervised learning faces two challenges: efficiently finding relevant samples in a massive corpus, and distinguishing true positives from visually confusable negatives when embeddings do not cleanly separate classes. Conventional active learning offers a principled framework for reducing annotation cost, yet it treats sample selection as a single-stage process that addresses neither challenge efficiently. FunnelAL decomposes the problem into cascaded stages. Starting from a single positive and negative example, the system iterates through: (1) embedding-based retrieval scoring that narrows the corpus to a manageable candidate set; (2) a precision-triggered ranking stage that exploits a learned ranker (RankNet) while batch precision remains high, then automatically blends in committee-based exploration (QBC) once returns diminish; and (3) feedback from the annotator's labels that refines both stages in subsequent iterations. We evaluate on three diverse image classification benchmarks. With a perfect annotator, FunnelAL attains the best final F1 on all three benchmarks, the best annotation efficiency (first in AULC), and the fewest annotation rounds. The most recent single-class discovery methods (GAL, PF-MA) at best match its final quality, and only at consistently higher labeling cost. Under annotator labeling errors at realistic rates, FunnelAL remains first or statistically tied for first while classical uncertainty-based methods degrade two to three times faster. Our work provides a concrete bridge between multi-stage recommender systems and active learning.
Chinese Translation
我们提出了FunnelAL，这是一个用于单类发现的检索-再排序主动学习系统，它将工业推荐系统的多阶段漏斗架构适配于数据注释。大规模监督学习面临两个挑战：在庞大的语料库中高效找到相关样本，以及在嵌入无法清晰区分类别时从视觉上混淆的负样本中区分真正的正样本。传统的主动学习提供了一个原则性的框架来降低注释成本，但它将样本选择视为一个单阶段过程，未能有效解决这两个挑战。FunnelAL将问题分解为级联阶段。从一个正样本和一个负样本开始，该系统迭代经过：(1) 基于嵌入的检索评分，将语料库缩小到一个可管理的候选集；(2) 一个精度触发的排序阶段，在批量精度保持高的情况下利用学习到的排序器（RankNet），然后在回报减少时自动融合基于委员会的探索（QBC）；以及(3) 来自注释者标签的反馈，进一步细化后续迭代中的两个阶段。我们在三个不同的图像分类基准上进行了评估。在完美的注释者情况下，FunnelAL在所有三个基准上都达到了最佳的最终F1值，最佳的注释效率（在AULC中排名第一），以及最少的注释轮次。最新的单类发现方法（GAL，PF-MA）在最佳情况下与其最终质量相匹配，但仅在持续更高的标注成本下。在现实的标注错误率下，FunnelAL仍然保持第一或在统计上与第一平局，而经典的不确定性基础方法的性能下降速度是其两到三倍。我们的工作为多阶段推荐系统与主动学习之间提供了一个具体的桥梁。
cs.CV / 25 / 2607.25294
CLBench-V: Evaluating Multimodal Context Learning from Grounding to Knowledge Acquisition
CLBench-V：从基础到知识获取的多模态上下文学习评估
Abstract
Real-world tasks often require models to learn from task-specific context rather than relying only on pre-trained knowledge. While recent work has highlighted this capability as context learning, existing evaluations mainly focus on textual contexts. In many practical settings, however, the context to be learned from is multimodal: scientific findings are conveyed through figures and tables, financial indicators are scattered across converted reports, and spatial decisions depend on maps, scenes, or web pages. We introduce CLBench-V, a benchmark for multimodal context learning that addresses the difficulty of localizing where context use breaks down by organizing tasks around three dimensions: context grounding, new information application, and new knowledge learning. CLBench-V combines converted public benchmarks with newly constructed datasets spanning domains such as science, finance, long-document understanding, spatial reasoning, and web-based visual question answering. To reduce the cost of constructing domain-specific context-learning tasks, we further use automated construction and filtering procedures for our newly built datasets. Across 3,443 instances and six recent multimodal models, the best overall score is only 0.2847, indicating that multimodal context learning remains far from saturated. Moreover, InternVL3.5-30B-A3B performs best on context grounding and new knowledge learning, while Qwen3.5-Plus performs best on new information application. We further analyze judge reliability, context length, image count, and representative failure cases. Code is available at https://github.com/IamLihua/CLBench-V.
Chinese Translation
现实世界的任务通常要求模型从特定任务的上下文中学习，而不仅仅依赖于预训练知识。尽管近期的研究强调了这种能力为上下文学习，但现有的评估主要集中在文本上下文上。然而，在许多实际场景中，待学习的上下文是多模态的：科学发现通过图表和表格传达，金融指标散布在转换后的报告中，空间决策依赖于地图、场景或网页。我们引入了CLBench-V，这是一个多模态上下文学习的基准，旨在解决上下文使用中断裂的定位难题，通过围绕三个维度组织任务：上下文基础、新信息应用和新知识学习。CLBench-V结合了转换的公共基准和新构建的数据集，涵盖科学、金融、长文档理解、空间推理和基于网络的视觉问答等领域。为了降低构建特定领域上下文学习任务的成本，我们进一步使用自动化构建和过滤程序来处理新构建的数据集。在3,443个实例和六个最新的多模态模型中，最佳总体得分仅为0.2847，表明多模态上下文学习仍远未饱和。此外，InternVL3.5-30B-A3B在上下文基础和新知识学习方面表现最佳，而Qwen3.5-Plus在新信息应用方面表现最佳。我们进一步分析了评审可靠性、上下文长度、图像数量和典型失败案例。代码可在https://github.com/IamLihua/CLBench-V获取。
cs.CV / 26 / 2607.25300
MEDit-Bench: A Dataset for Evaluating Message-Driven Narrative Video Editing
MEDit-Bench：用于评估基于信息驱动的叙事视频编辑的数据集
Abstract
Video editing is fundamentally message-driven: even from the same source footage, the selected shots change depending on the narrative the editor wishes to convey. Benchmarks for a closely related task, video summarization, reduce editorial intent to a single, message-agnostic notion of saliency and thus do not account for this diversity. For evaluating message-driven video editing, we present \textbf{MEDit-Bench}, a dataset and benchmark, which pairs long-form videos with multiple editing messages and multiple professionally produced edits per message, demonstrating that different messages yield substantially different edits from the same source. We define an automatic evaluation protocol based on temporal alignment metrics, and find that an LLM-as-a-judge preference, a natural proxy for narrative quality, is unreliable for this task due to severe position bias. We additionally annotate each message with ambiguity and contextfulness scores, and show that both dimensions negatively correlate with model performance, establishing message difficulty as a meaningful stratification factor. Experiments with state-of-the-art MLLMs and reinforcement fine-tuned baselines show that while strong models approach human temporal alignment at lenient thresholds, all models fall behind humans at stricter criteria. A human perceptual study further confirms a large quality gap, with professional human edits remaining consistently preferred over model outputs.
Chinese Translation
视频编辑本质上是信息驱动的：即使来自相同的素材，所选镜头也会根据编辑者希望传达的叙事而变化。与视频摘要这一密切相关的任务的基准将编辑意图简化为单一的、与信息无关的显著性概念，因此未能考虑这种多样性。为了评估基于信息驱动的视频编辑，我们提出了 extbf{MEDit-Bench}，这是一个数据集和基准，它将长视频与多个编辑信息和每个信息的多个专业制作的编辑配对，展示了不同的信息会从相同的源产生显著不同的编辑。我们定义了一种基于时间对齐指标的自动评估协议，并发现 LLM-as-a-judge 偏好作为叙事质量的自然代理，在此任务中由于严重的位置信息偏差而不可靠。我们还为每个信息注释了模糊性和上下文相关性评分，并显示这两个维度与模型性能呈负相关，确立了信息难度作为一个有意义的分层因素。与最先进的 MLLMs 和强化微调基线的实验表明，尽管强模型在宽松阈值下接近人类的时间对齐，但在更严格的标准下，所有模型都落后于人类。一项人类感知研究进一步确认了巨大的质量差距，专业人类编辑始终被优先选择，而不是模型输出。
cs.CV / 27 / 2607.25310
Human-in-the-Loop Signature Bootstrapping for UAV Hyperspectral PFM-1 Mine Detection
基于人机协作的无人机高光谱PFM-1地雷检测签名自助引导
Abstract
Hyperspectral imaging (HSI) is useful for material discrimination, but operational mine screening also depends on how many false alarms must be inspected before targets are found. This paper studies PFM-1 landmine detection in unmanned aerial vehicle (UAV) visible and near-infrared (VNIR) HSI using spectral angle mapper (SAM), matched filter (MF), adaptive coherence estimator (ACE), and constrained energy minimization (CEM). We compare a ground-measured SVC signature, a fully informed in-scene core-pixel signature, and a simulated human-in-the-loop signature bootstrap. Besides receiver operating characteristic area under the curve and average precision, we report target-discovery curves and spatial candidate-review counts. Full-review bootstrapping reaches the fully informed in-scene signature case after all seven target regions are verified, but the required inspection effort varies strongly: ACE confirms all regions in two rounds and nine candidate inspections, whereas the SAM variants need thousands of candidate reviews for their final target locations. Code is available at https://github.com/SagarLekhak/IEEE_WHISPERS_2026_UAV_HSI_PFM1.
Chinese Translation
高光谱成像（HSI）在材料识别中具有重要应用，但操作性地雷筛查还依赖于在发现目标之前需要检查多少虚假警报。本文研究了在无人机（UAV）可见光和近红外（VNIR）高光谱成像中使用光谱角映射器（SAM）、匹配滤波器（MF）、自适应相干估计器（ACE）和约束能量最小化（CEM）进行PFM-1地雷检测。我们比较了地面测量的SVC签名、完全知情的场景内核心像素签名以及模拟的人机协作签名自助引导。除了接收者操作特征曲线下面积和平均精度外，我们还报告了目标发现曲线和空间候选审查计数。完全审查自助引导在验证所有七个目标区域后达到了完全知情的场景内签名情况，但所需的检查工作量差异很大：ACE在两轮和九次候选检查中确认了所有区域，而SAM变体则需要数千次候选审查才能确定最终目标位置。代码可在 https://github.com/SagarLekhak/IEEE_WHISPERS_2026_UAV_HSI_PFM1 获取。
cs.CV / 28 / 2607.25314
Sense it with your eyes: Sensation Generation and Understanding for Advertisements
用眼睛感知：广告中的感知生成与理解
Abstract
Sensory advertising evokes human senses through visual cues, enabling audiences to mentally simulate experiences and increasing persuasive impact. Despite the recent increase in using AI in generating and understanding creative and persuasive content, how advertisements visually evoke sensations remains largely unexplored. In this work, we introduce the first study of understanding, evaluating, and generating sensory ads. We introduce the Sensory Ad dataset, and define sensation classification tasks (SenseClass) to benchmark LLMs and MLLMs. We further propose SenseScore, an automated evaluation metric for sensation evocation achieving strong agreement with human judgments. Finally, we introduce the Sensory Ad Generation (SenseGen) task and propose SAGA, a multi-agent framework that improves message image alignment, sensory evocation, and persuasion. Our work establishes a foundation for sensory-aware visual persuasion.
Chinese Translation
感官广告通过视觉线索唤起人类感官，使观众能够在心理上模拟体验，从而增强说服力。尽管最近在生成和理解创意及说服内容方面使用人工智能的趋势有所增加，但广告如何在视觉上唤起感官仍然很大程度上未被探索。在本研究中，我们首次介绍了对感官广告的理解、评估和生成的研究。我们引入了感官广告数据集，并定义了感知分类任务（SenseClass），以基准化大型语言模型（LLMs）和多模态大型模型（MLLMs）。我们进一步提出了SenseScore，这是一种自动化评估指标，用于感官唤起，且与人类判断具有较强的一致性。最后，我们介绍了感官广告生成（SenseGen）任务，并提出了SAGA，这是一个多智能体框架，旨在改善信息与图像的对齐、感官唤起和说服力。我们的工作为感官意识的视觉说服奠定了基础。
cs.CV / 29 / 2607.25318
Beyond Background Bias: Saliency-Driven Prototype Alignment for Dataset Distillation
超越背景偏差：基于显著性的原型对齐用于数据集蒸馏
Abstract
Dataset distillation aims to synthesize compact datasets that can approximate the performance of full-data training while significantly reducing computational and storage costs. However, diffusion-based distillation methods often struggle to preserve structural coherence and generalization, especially in visually complex domains. This issue often stems from latent prototypes that are weakly aligned with class-discriminative regions and contaminated by irrelevant background, thereby degrading generation quality and generalization. To address this limitation, we propose a saliency-driven distillation framework that constructs class-discriminative latent prototypes to enhance representativeness and generalization. The framework proceeds in two stages: (1) ensemble Grad-CAM saliency is used to construct prototypes emphasizing high-confidence regions, and (2) hard prototype refinement is then applied to construct challenging yet class-consistent prototypes, thereby enhancing discriminability and diversity. Importantly, the diffusion backbones (e.g., LDM and DiT) remain frozen; only lightweight classifiers used for saliency extraction are trained. Extensive experiments across multiple benchmarks demonstrate consistent performance improvements over strong baselines. Code will be released.
Chinese Translation
数据集蒸馏旨在合成紧凑的数据集，以近似全数据训练的性能，同时显著降低计算和存储成本。然而，基于扩散的方法在保持结构一致性和泛化能力方面常常面临挑战，尤其是在视觉复杂的领域。这一问题通常源于与类区分区域弱对齐且受到无关背景污染的潜在原型，从而降低了生成质量和泛化能力。为了解决这一局限性，我们提出了一种基于显著性的蒸馏框架，该框架构建类区分的潜在原型，以增强代表性和泛化能力。该框架分为两个阶段：（1）使用集成的 Grad-CAM 显著性构建强调高置信度区域的原型；（2）随后应用困难原型细化，以构建具有挑战性且类一致的原型，从而增强可区分性和多样性。重要的是，扩散主干（例如 LDM 和 DiT）保持冻结；仅对用于显著性提取的轻量级分类器进行训练。在多个基准测试中的广泛实验表明，相较于强基线有一致的性能提升。代码将会发布。
cs.CV / 30 / 2607.25324
Balanced Soft mixture-of-expert model for Glaucoma Detection
用于青光眼检测的平衡软混合专家模型
Abstract
Glaucoma is a group of eye diseases that damage the optic nerve, often caused by elevated intraocular pressure. It is a leading cause of irreversible vision loss and is typically developed slowly and painlessly, making it difficult to notice until significant damage has occurred. Therefore, early detection is crucial to prevent or slow the progression of vision loss. In recent years, deep learning based uni-modal models have improved the accuracy and efficiency of glaucoma detection, empowering doctors with tools for earlier diagnosis, better monitoring, and timely treatment. Building on this, multi-modal models have emerged, leveraging the strengths of different imaging modalities to learn richer and more robust representations, further enhancing glaucoma detection accuracy. However, multi-modal learning faces challenges such as imbalanced and under-optimized uni-modal representations due to joint learning objectives. To address this, we propose a balanced soft mixture-experts model with three experts and load balancing loss. The performance is measured by AUC, our proposed method surpasses the performance of all uni-modal baselines, conventional multi-modal models, and current stateof- the-art balanced multi-modal models. The proposed model can be generalized to other disease detections such as diabetic retinopathy.
Chinese Translation
青光眼是一组损害视神经的眼病，通常由眼内压升高引起。它是不可逆视力丧失的主要原因，通常发展缓慢且无痛，使得在发生显著损害之前难以察觉。因此，早期检测对于预防或减缓视力丧失的进展至关重要。近年来，基于深度学习的单模态模型提高了青光眼检测的准确性和效率，为医生提供了更早诊断、更好监测和及时治疗的工具。在此基础上，多模态模型应运而生，利用不同成像模态的优势，学习更丰富和更强健的表征，进一步提高青光眼检测的准确性。然而，多模态学习面临着由于联合学习目标导致的单模态表征不平衡和未优化的问题。为了解决这一问题，我们提出了一种平衡软混合专家模型，该模型包含三个专家和负载平衡损失。通过AUC指标进行性能评估，我们提出的方法超越了所有单模态基线、传统多模态模型和当前最先进的平衡多模态模型。该模型可以推广到其他疾病的检测，如糖尿病视网膜病变。
cs.CV / 31 / 2607.25362
PanoLess: Environment Reconstruction from Partial Reflective Views
PanoLess：基于部分反射视图的环境重建
Abstract
Reflections from shiny objects and glass facades naturally extend the field of view of a camera, capturing the surrounding environment without the need to pan the camera or acquire a full panorama. We propose PanoLess, a Gaussian-splat-based framework that reconstructs the surrounding environment as a distant illumination map from images captured on only one side of a reflective surface. PanoLess leverages surface-aligned 2D Gaussian splats with deferred shading to recover accurate per-pixel normals and reflection cues, which are fused into a neural cubemap representation of the environment. In addition, PanoLess produces a visibility map that explicitly denotes which regions of the environment are supported by the partial reflective observations. Unlike existing inverse-rendering and reflection-aware Gaussian-splatting approaches, which typically require full 360-degree coverage and struggle under incomplete views, PanoLess enables consistent, physically grounded illumination estimation from partial-view input. We show that PanoLess achieves high-fidelity and geometrically consistent environment reconstruction, outperforming reflection-aware baselines on a new custom synthetic benchmark and publicly available datasets, and demonstrating generalization to real-world reflective captures.
Chinese Translation
来自光滑物体和玻璃外立面的反射自然扩展了相机的视野，捕捉周围环境而无需移动相机或获取完整的全景图。我们提出了PanoLess，一个基于高斯点云的框架，它从仅在反射表面一侧捕获的图像中重建周围环境的远程照明图。PanoLess利用与表面对齐的2D高斯点云和延迟着色技术来恢复准确的每像素法线和反射线索，这些信息被融合成环境的神经立方体映射。此外，PanoLess生成一个可见性图，明确标示出哪些区域的环境是由部分反射观测支持的。与现有的逆渲染和反射感知高斯点云方法不同，这些方法通常需要完整的360度覆盖并在不完整视图下表现不佳，PanoLess能够从部分视图输入中实现一致的、物理基础的照明估计。我们展示了PanoLess在新的自定义合成基准和公开可用的数据集上实现了高保真度和几何一致的环境重建，超越了反射感知基线，并展示了对真实世界反射捕捉的泛化能力。
cs.CV / 32 / 2607.25367
Leak-Free Cross-Validated Stacking with Per-Architecture Calibration for Sand-Boil Segmentation in Earthen Levees
无泄漏的交叉验证堆叠与每种架构校准用于土堤沙喷涌分割
Abstract
Sand boils, points where water seeping beneath an earthen levee re-emerges at the surface, are early warnings of internal erosion, and deep segmentation networks are increasingly used to find them in inspection photographs. Annotated examples are scarce, and two common ways of working around that scarcity quietly inflate reported accuracy: tuning ensemble weights on the same images later used to score them, and training on synthetic images derived from the very photographs held out for testing. We present a sand-boil segmentation framework that closes both loopholes. Every synthetic image carries a pointer to its real parent, and a per-fold filter excludes any image whose parent is held out; five encoder-decoder backbones are trained under five-fold cross-validation, calibrated by one temperature scalar each, and combined by a per-pixel meta-learner fitted only on out-of-fold predictions. On the held-out test set the proposed Updated SandBoilNet reaches an intersection-over-union of 0.707 over three seeds, against 0.608 for the published original re-evaluated on the same split. Under the stacking protocol the calibrated stack reaches 0.681 against 0.694 for the strongest fold-averaged member, so it does not improve on the best single model; eight meta-learner families reproduce that outcome, which we trace to a mean pairwise error correlation of 0.894 among members. A synthetic pool filtered for label fidelity lifts the champion to 0.718 over three seeds against a 0.707 control. We also introduce a mask-conditioned synthesis route that makes the conditioning mask the label by construction, giving labelled training images at zero annotation cost.
Chinese Translation
沙喷涌是指水在土堤下渗透后重新出现在表面的点，是内部侵蚀的早期警告，深度分割网络越来越多地用于在检查照片中寻找它们。标注的示例稀缺，而两种常见的应对这种稀缺的方法在不知不觉中夸大了报告的准确性：在同一图像上调整集成权重，随后用于评分，以及在从保留用于测试的照片中派生的合成图像上进行训练。我们提出了一种沙喷涌分割框架，解决了这两个漏洞。每个合成图像都指向其真实母图像，而每折过滤器排除了任何母图像被保留的图像；五个编码器-解码器骨干网络在五折交叉验证下训练，每个网络由一个温度标量进行校准，并通过一个仅在折外预测上拟合的每像素元学习器进行组合。在保留的测试集上，所提出的更新版 SandBoilNet 在三个种子下达到了 0.707 的交并比，而在同一拆分上重新评估的已发布原始模型为 0.608。在堆叠协议下，经过校准的堆叠达到了 0.681，而最强的折平均成员为 0.694，因此并未改善最佳单一模型；八个元学习器家族重现了这一结果，我们追溯到成员之间的平均成对误差相关性为 0.894。经过标签保真度过滤的合成池将冠军提升至 0.718，三个种子对比控制组的 0.707。我们还引入了一种掩模条件合成路径，通过构造使条件掩模成为标签，从而以零标注成本获得标注的训练图像。
cs.CV / 33 / 2607.25371
Hyperspectral Intrinsic Decomposition: Joint Recovery of Reflectance and Photometric Components for Non-Lambertian Scenes
高光谱内在分解：非朗伯场景中反射率与光度成分的联合恢复
Abstract
Hyperspectral intrinsic decomposition (HID) aims to disentangle material-related spectral properties and photometric effects in hyperspectral images (HSIs), which is essential for understanding real-world imaging processes and benefits a variety of downstream applications. Most existing HID studies have been developed under Lambertian or near-Lambertian assumptions. The few prior non-Lambertian efforts rely on simplified specular assumptions insufficient to handle diverse real-world specularity, and typically require auxiliary inputs or recover only a subset of the coupled reflectance and photometric components, hindering complete and blind decomposition. In this paper, we revisit the dichromatic reflection model (DRM) and develop a unified inversion paradigm that reformulates the recovery of four coupled reflectance and photometric components as the estimation of two spectral--spatial target variables. Building on this reformulation, we propose a dual-scale decomposition scheme to handle non-Lambertian effects with distinct spatial characteristics. At the global scale, photometrically invariant descriptors serve as edge priors for high-fidelity intrinsic boundary preservation; at the local scale, specularity-guided attention directs refinement with emphasis on specularity-dominated regions, including those affected by clipping distortion. To facilitate future research, we establish CITE, the first public real-world HID dataset for non-Lambertian objects, and develop a Physically-faithful Intrinsic Set Generator (PISG) for controllable data synthesis. Extensive ablation studies and experiments on the CITE and additional HSIs demonstrate the effectiveness of our method and its robustness across diverse scenes.
Chinese Translation
高光谱内在分解（HID）旨在解开高光谱图像（HSIs）中与材料相关的光谱特性和光度效应，这对于理解现实世界的成像过程至关重要，并且有利于多种下游应用。现有的大多数HID研究是在朗伯或近朗伯假设下发展起来的。少数先前的非朗伯研究依赖于简化的镜面假设，这不足以处理多样的现实世界镜面反射，并且通常需要辅助输入或仅恢复耦合的反射率和光度成分的一个子集，从而阻碍了完整和盲目的分解。在本文中，我们重新审视了二色反射模型（DRM），并开发了一种统一的反演范式，将四个耦合的反射率和光度成分的恢复重新表述为两个光谱-空间目标变量的估计。基于这种重新表述，我们提出了一种双尺度分解方案，以处理具有不同空间特征的非朗伯效应。在全局尺度上，光度不变描述符作为高保真内在边界保持的边缘先验；在局部尺度上，镜面导向的注意力引导细化，强调镜面主导区域，包括那些受到剪切失真的区域。为了促进未来的研究，我们建立了CITE，这是第一个针对非朗伯物体的公共真实世界HID数据集，并开发了一个物理真实内在集生成器（PISG）以实现可控的数据合成。在CITE和其他高光谱图像上的广泛消融研究和实验表明了我们方法的有效性及其在多样场景中的鲁棒性。
cs.CV / 34 / 2607.25377
Gaussian Volumetric Representation for Efficient Shear-Warp Visualization
高斯体积表示法用于高效剪切-扭曲可视化
Abstract
Medical image visualization requires volumetric rendering algorithms that preserve anatomical fidelity while maintaining high rendering speeds. To address the high computational cost of large volumetric datasets, we propose a Gaussian-based volumetric representation for efficient visualization of dense medical volumes without compromising structural and radiometric details. We optimize the proposed representation using Monte Carlo volumetric estimation, which enables training on a highly sparse subset of voxels while maintaining consistency with the dense volumetric objective. In addition, we introduce a curriculum learning strategy that progressively incorporates structured slice-based sampling during training. Sparse voxel samples provide an early global coverage of the volume, while slice samples capture spatially correlated regions that aid geometric structure and texture continuity. This combination enables the Gaussian representation to learn anatomical details of various structures and corresponding textures from sparse supervision while significantly reducing the computational cost associated with dense voxel processing. The learned representation supports slice-based rendering methods such as shear-warp volume rendering, enabling efficient visualization of multimodal medical datasets including MRI and Cryosection volumes while preserving anatomical structures. Using sparse supervision, our method achieves up to 43.86 FPS rendering with a compression ratio of 11.31:1.
Chinese Translation
医学图像可视化需要在保持解剖学忠实度的同时实现高渲染速度的体积渲染算法。为了解决大体积数据集的高计算成本，我们提出了一种基于高斯的体积表示法，以高效可视化密集的医学体积，同时不妥协结构和辐射度细节。我们通过蒙特卡洛体积估计优化所提的表示法，使其能够在高度稀疏的体素子集上进行训练，同时保持与密集体积目标的一致性。此外，我们引入了一种课程学习策略，在训练过程中逐步结合结构化的切片基础采样。稀疏体素样本为体积提供了早期的全局覆盖，而切片样本则捕捉空间相关区域，帮助几何结构和纹理的连续性。这种组合使得高斯表示法能够从稀疏监督中学习各种结构的解剖细节及其相应的纹理，同时显著降低与密集体素处理相关的计算成本。所学习的表示法支持基于切片的渲染方法，如剪切-扭曲体积渲染，从而实现对包括MRI和冷冻切片体积在内的多模态医学数据集的高效可视化，同时保持解剖结构的完整性。通过稀疏监督，我们的方法在压缩比为11.31:1的情况下实现了高达43.86帧每秒的渲染速度。
cs.CV / 35 / 2607.25389
HOME: Robust Hough-space Matching Method for Structured and Textureless Videos
HOME：用于结构化和无纹理视频的鲁棒霍夫空间匹配方法
Abstract
Visual front-ends for robotic localization typically rely on point-based features such as Oriented FAST and Rotated BRIEF (ORB), which frequently fail in structured environments dominated by strong linear structures or textureless surfaces. While line-based Simultaneous Localization and Mapping (SLAM) systems mitigate this by utilizing line segments, conventional line extraction and description algorithms are computationally prohibitive for real-time edge robotics. To address this fundamental bottleneck, we propose HOME (Hough-space One-dimensional Matching of Extrema), an ultra-lightweight, training-free feature matching framework. HOME transforms images into Hough space, mapping global linear structures to stable local extrema, which serve as keypoints, thereby reformulating complex line matching into highly efficient one-dimensional point matching. The proposed 1D radial descriptor mathematically guarantees rotational and translational invariance without the overhead of explicit orientation estimation. As a proof of concept to validate the matching accuracy and efficiency of HOME, this paper focuses on homography estimation. Extensive evaluations demonstrate that HOME achieves robust registration in challenging scenarios where point-based methods fail, operating at a much faster speed than existing line-based methods. Extending this robust matching engine to full 3D pose estimation remains a highly promising future direction.
Chinese Translation
机器人定位的视觉前端通常依赖于基于点的特征，如定向FAST和旋转BRIEF（ORB），但在以强线性结构或无纹理表面为主的结构化环境中，这些方法常常失效。虽然基于线段的同时定位与地图构建（SLAM）系统通过利用线段来缓解这一问题，但传统的线提取和描述算法在实时边缘机器人中计算开销过大。为了解决这一根本瓶颈，我们提出了HOME（霍夫空间极值的一维匹配），这是一种超轻量级、无需训练的特征匹配框架。HOME将图像转换为霍夫空间，将全局线性结构映射到稳定的局部极值，这些极值作为关键点，从而将复杂的线匹配重新构造为高效的一维点匹配。所提出的一维径向描述符在数学上保证了旋转和位移的不变性，而无需显式的方向估计开销。作为验证HOME匹配精度和效率的概念证明，本文重点关注单应性估计。广泛的评估表明，HOME在点基方法失效的挑战场景中实现了鲁棒的配准，且运行速度远快于现有的基于线的方法。将这一鲁棒匹配引擎扩展到完整的三维姿态估计仍然是一个极具前景的未来方向。
cs.CV / 36 / 2607.25390
Noise-Free One-Step LoRA for Task-Driven Image Restoration with Diffusion Priors
无噪声的一步式 LoRA 任务驱动图像恢复与扩散先验
Abstract
Degraded images not only reduce visual quality but also impair downstream high-level vision tasks. Task-driven image restoration (TDIR) addresses this issue by jointly optimizing restoration quality and task performance. Recent works show that pretrained diffusion priors benefit TDIR, yet diffusion-based restoration is inherently stochastic, as the sampling process depends on a random noise term, which can undermine task consistency. In this paper, we show that a deterministic, noise-free one-step forward pass with pretrained diffusion priors can substantially improve TDIR, but the benefit critically depends on the adaptation module: LoRA yields consistent gains, whereas ControlNet-style conditioning does not. This enables one-step forwarding that surpasses conventional multi-step diffusion TDIR baselines. Furthermore, we introduce a task-preserving GAN training strategy that improves perceptual quality without sacrificing task performance. Extensive experiments on classification, segmentation, and detection demonstrate consistent gains over prior TDIR methods, and we further validate generalization on real-world degraded images and OCR.
Chinese Translation
退化图像不仅降低了视觉质量，还损害了下游高层次视觉任务。任务驱动图像恢复（TDIR）通过联合优化恢复质量和任务性能来解决这一问题。近期研究表明，预训练的扩散先验对 TDIR 有益，然而基于扩散的恢复本质上是随机的，因为采样过程依赖于随机噪声项，这可能会削弱任务一致性。本文展示了使用预训练扩散先验的确定性、无噪声的一步前向传递可以显著改善 TDIR，但这种好处在很大程度上依赖于适应模块：LoRA 产生了一致的增益，而 ControlNet 风格的条件化则没有。这使得一步前向传递超越了传统的多步扩散 TDIR 基线。此外，我们引入了一种保持任务的 GAN 训练策略，改善了感知质量而不牺牲任务性能。在分类、分割和检测方面的广泛实验表明，相较于先前的 TDIR 方法具有一致的增益，我们进一步验证了在现实世界退化图像和光学字符识别（OCR）上的泛化能力。
cs.CV / 37 / 2607.25392
RDVSv2: A Large-scale Benchmark for RGB-D Video Salient Object Detection
RDVSv2：一个大规模的RGB-D视频显著目标检测基准
Abstract
We introduce RDVSv2, a large-scale benchmark for RGB-D video salient object detection (RGB-D VSOD) with dense frame-level annotations. Existing datasets in this emerging field are often limited in scale and annotation quality, while also relying on less geometry-consistent depth cues. To address these limitations, RDVSv2 is built from publicly accessible stereoscopic online videos and contains 249 video sequences with 29,077 annotated frames. It includes depth maps derived from stereoscopic videos, together with frame-wise salient object masks annotated with eye-tracking guidance. Compared with existing datasets, RDVSv2 is much larger in scale and covers more diverse and challenging scenarios. In addition, we establish a strong baseline for RGB-D VSOD based on Segment Anything Model 2 (SAM2). Specifically, we employ a parameter-efficient fine-tuning (PEFT) strategy to adapt the SAM2 encoder to jointly encode RGB, depth, and optical flow cues. Extensive experiments show that RDVSv2 is substantially more challenging for existing RGB-D VSOD methods. Meanwhile, the proposed baseline achieves state-of-the-art results on RDVSv2 and existing RGB-D VSOD benchmarks. We hope that RDVSv2 and the provided baseline will serve as useful resources for future research on RGB-D VSOD and related multi-modal video understanding tasks. Our dataset and code will be available at https://github.com/ltynick/RDVSv2.
Chinese Translation
我们介绍了RDVSv2，这是一个用于RGB-D视频显著目标检测（RGB-D VSOD）的大规模基准，具有密集的帧级注释。现有的数据集在这一新兴领域通常在规模和注释质量上受到限制，同时也依赖于较少的几何一致性深度线索。为了解决这些局限性，RDVSv2基于公开可访问的立体在线视频构建，包含249个视频序列和29,077个注释帧。它包括从立体视频中提取的深度图，以及基于眼动追踪指导注释的逐帧显著目标掩码。与现有数据集相比，RDVSv2在规模上大得多，并涵盖了更多样化和具有挑战性的场景。此外，我们基于Segment Anything Model 2（SAM2）建立了RGB-D VSOD的强基线。具体而言，我们采用了一种参数高效微调（PEFT）策略，以适应SAM2编码器共同编码RGB、深度和光流线索。大量实验表明，RDVSv2对现有的RGB-D VSOD方法具有显著的挑战性。同时，所提出的基线在RDVSv2和现有的RGB-D VSOD基准上达到了最先进的结果。我们希望RDVSv2及提供的基线能够为未来的RGB-D VSOD及相关多模态视频理解任务的研究提供有用的资源。我们的数据集和代码将发布在https://github.com/ltynick/RDVSv2。
cs.CV / 38 / 2607.25393
Towards Reliable Stain Transfer: An Iterative Data-Model Co-Optimization Framework Based on Multimodal Expert-Guided Assessment
迈向可靠的染色转移：基于多模态专家指导评估的迭代数据-模型协同优化框架
Abstract
Histopathological examination primarily relies on hematoxylin and eosin (H&E) and immunohistochemistry (IHC) staining. Although IHC provides critical molecular information, it is costly and requires specialized expertise. Stain transfer provides an efficient alternative by computationally generating IHC from H&E images, but remains challenged by unified and interpretable modeling for heterogeneous biomarkers under pixel-unaligned supervision. We propose DMCoStain, a novel Data-Model Co-optimization framework for Stain transfer. It iteratively co-refines training data and model capability, improving staining accuracy and interpretability in both pathological and structural consistency. To refine training data in a clinically meaningful manner, it incorporates the Multimodal Expert-Guided Finer Selection (MEGFS) strategy, built upon a pioneering IHC-positive-expression (IPE) vision-language model (VLM) that emulates pathologist reasoning. To support MEGFS, we construct ImmunoInstruction, the first large-scale IPE instruction-following dataset with 150K VQA samples. Extensive experiments on multiple tissues and biomarkers demonstrate that DMCoStain achieves state-of-the-art (SOTA) accuracy. This paradigm offers strong practical value, and MEGFS also functions as a specialized evaluation tool for future model development. Dataset, code, and more details are in https://github.com/SikangSHU/DMCoStain.
Chinese Translation
组织病理学检查主要依赖于苏木精-伊红（H&E）和免疫组化（IHC）染色。尽管IHC提供了重要的分子信息，但其成本高昂且需要专业的技术支持。染色转移通过计算生成H&E图像的IHC，提供了一种高效的替代方案，但在像素不对齐的监督下，针对异质生物标志物的统一和可解释建模仍然面临挑战。我们提出了DMCoStain，一种用于染色转移的新型数据-模型协同优化框架。该框架迭代地共同优化训练数据和模型能力，提高了染色的准确性和在病理及结构一致性方面的可解释性。为了以临床上有意义的方式优化训练数据，它结合了多模态专家指导的精细选择（MEGFS）策略，该策略基于一种开创性的IHC阳性表达（IPE）视觉-语言模型（VLM），模拟病理学家的推理。为了支持MEGFS，我们构建了ImmunoInstruction，这是第一个具有15万VQA样本的大规模IPE指令跟随数据集。对多种组织和生物标志物的广泛实验表明，DMCoStain达到了最先进的（SOTA）准确性。该范式具有强大的实际价值，MEGFS也作为未来模型开发的专门评估工具。数据集、代码及更多细节见 https://github.com/SikangSHU/DMCoStain。
cs.CV / 39 / 2607.25407
ANFI: Rethinking Neighbor Feature Interaction in Person Re-ID
ANFI：重新思考行人重识别中的邻居特征交互
Abstract
In person re-identification, neighbor-based methods have achieved significant success by interacting with neighbor samples to obtain more robust representations. However, existing methods rely only on affinity relations, causing their success to depend heavily on the reliability of selected neighbors. We find that affinity-only interaction often fails in challenging scenarios due to the inevitable presence of noisy neighbors. To enable effective interactions under noisy neighborhoods, we revisit neighbor-based methods under distinct reliability conditions and propose a novel Adaptive Neighbor Feature Interaction (ANFI) method. The core idea of ANFI is to account for negative effects from noisy neighbors, allowing samples to remain distinguishable from false positive neighbors. Unlike existing methods, ANFI models not only affinity relations but also discrepancy relations, and employs sample-wise adaptive weighting for these two types of relations. Given that capturing negative effects from noisy neighbors differs significantly from traditional relation learning, we derive discrepancy relations from a new neighborhood similarity, which provides more information than pairwise similarity. In addition, we propose Noisy Relation Supervision (NRS) to train ANFI, gradually injecting robustness to noisy relations into the model. Extensive experiments conducted under standard, cross-modal, and cross-domain settings, including comparisons with neighbor-based methods and re-ranking methods, demonstrate the superiority of our method across various neighbor distributions.
Chinese Translation
在行人重识别中，基于邻居的方法通过与邻居样本的交互取得了显著成功，从而获得更为稳健的表示。然而，现有方法仅依赖于亲和关系，导致其成功在很大程度上依赖于所选邻居的可靠性。我们发现，仅依靠亲和交互在具有挑战性的场景中往往失败，因为不可避免地存在噪声邻居。为了在噪声邻域中实现有效的交互，我们在不同的可靠性条件下重新审视基于邻居的方法，并提出了一种新颖的自适应邻居特征交互（Adaptive Neighbor Feature Interaction，ANFI）方法。ANFI的核心思想是考虑噪声邻居带来的负面影响，使样本能够与假阳性邻居保持可区分性。与现有方法不同，ANFI不仅建模亲和关系，还建模差异关系，并对这两种关系采用样本自适应加权。鉴于捕捉噪声邻居的负面影响与传统关系学习有显著不同，我们从一种新的邻域相似性中推导出差异关系，这提供了比成对相似性更多的信息。此外，我们提出了噪声关系监督（Noisy Relation Supervision，NRS）来训练ANFI，逐步将对噪声关系的鲁棒性注入模型。在标准、跨模态和跨领域设置下进行的大量实验，包括与基于邻居的方法和重排序方法的比较，证明了我们的方法在各种邻居分布下的优越性。
cs.CV / 40 / 2607.25432
Bi-Level Collaborative Learning for Few-Shot Scribble-Supervised Medical Image Segmentation
双层协作学习用于少样本涂鸦监督的医学图像分割
Abstract
Scribble annotations offer an efficient alternative to costly pixel-wise labeling for medical image segmentation, yet in real clinical scenarios, scribble-annotated samples are often still limited, imposing the dual challenges of sparse supervision and annotated sample scarcity. These compounded constraints severely deprive models of the structural evidence needed for complete region recovery and precise boundary delineation. To break this bottleneck, we propose a bi-level collaborative learning framework for few-shot scribble-supervised medical image segmentation. Specifically, an upper-level learnable superpixel model is introduced to provide region-structural priors for lower-level segmentation, while superpixel-based region-wise pseudo-label propagation and a spatial-prior-guided filtering strategy are performed to generate reliable dense pseudo-labels for segmentation learning. Meanwhile, the anatomical semantics learned by the lower-level segmentation model under the guidance of the current superpixels are fed back to the upper level, further driving it to learn region-structural representations better aligned with the segmentation task. Through bidirectional interaction and collaborative learning between the upper and lower levels, the proposed framework significantly outperforms existing state-of-the-art scribble-supervised methods on the ACDC and Prostate datasets under the few-shot scribble-supervised setting.
Chinese Translation
涂鸦标注为医学图像分割提供了一种高效的替代方案，取代了成本高昂的逐像素标注。然而，在实际临床场景中，涂鸦标注样本往往仍然有限，这带来了稀疏监督和标注样本稀缺的双重挑战。这些复合约束严重剥夺了模型所需的结构证据，从而影响了完整区域恢复和精确边界划定。为了解决这一瓶颈，我们提出了一种双层协作学习框架，用于少样本涂鸦监督的医学图像分割。具体而言，引入了一个上层可学习的超像素模型，为下层分割提供区域结构先验，同时通过基于超像素的区域伪标签传播和空间先验引导的过滤策略，生成可靠的密集伪标签以进行分割学习。同时，在当前超像素的指导下，下层分割模型学习到的解剖语义会反馈到上层，进一步推动其学习与分割任务更好对齐的区域结构表示。通过上下层之间的双向互动和协作学习，所提出的框架在少样本涂鸦监督设置下，在ACDC和前列腺数据集上显著超越了现有的最先进的涂鸦监督方法。
cs.CV / 41 / 2607.25455
Reading Legends on Ancient Coins: An Object Detection Approach for Character Recognition on a Novel Roman Republican Dataset
古币上的铭文解读：基于物体检测的古罗马共和时期数据集字符识别方法
Abstract
When it comes to the proper classification of ancient coins with respect to their time and issuer, the textual inscriptions on these coins, also known as legends, are of paramount importance. These legends consist of alphabets or characters still used in English. This paper addresses image based character recognition on ancient Roman Republican coins via a deep learning based object detection strategy. However, legends on these coins pose high variation due to non-uniform placement, primitive inscription techniques, and wear and tear. Additional challenges include inconsistent imaging conditions such as illumination, orientation, and scale. To accommodate these, we gathered a novel large-scale dataset of 5,654 Roman Republican coin images, manually annotated with 21 character labels, totaling 38,808 annotations. For recognition, we use You Only Look Once (YOLO) variants: YOLOv3, v4, v5, v7, and v8. YOLOv7-Large achieves the best mAP50 of 90.4%, followed by YOLOv7-Extended and YOLOv7-xl with 90.2% and 90.1%, respectively.
Chinese Translation
在对古币进行时间和发行者的正确分类时，这些硬币上的文字铭文（即铭文）至关重要。这些铭文由仍在英语中使用的字母或字符组成。本文通过基于深度学习的物体检测策略，探讨了古罗马共和时期硬币上的图像字符识别。然而，这些硬币上的铭文由于不均匀的放置、原始的铭刻技术以及磨损等原因，呈现出高度的变化性。其他挑战还包括不一致的成像条件，如光照、方向和比例。为了解决这些问题，我们收集了一个新的大规模数据集，包含5,654张古罗马共和时期硬币图像，手动标注了21个字符标签，总计38,808个标注。为了进行识别，我们使用了You Only Look Once (YOLO)的多个变体：YOLOv3、v4、v5、v7和v8。其中，YOLOv7-Large实现了最佳的mAP50为90.4%，其次是YOLOv7-Extended和YOLOv7-xl，分别为90.2%和90.1%。
cs.CV / 42 / 2607.25465
DensFiLM: Density-Conditioned Video Saliency for Crowd Scenes
DensFiLM：基于密度条件的视频显著性模型用于人群场景
Abstract
Video saliency models typically apply a single fixation strategy across crowd scenes, despite systematic changes in attention with crowd density. Sparse scenes encourage tracking individuals, whereas dense scenes shift attention toward collective motion and scene-level landmarks. We introduce DensFiLM, a density-conditioned video saliency model that inserts a lightweight Feature-wise Linear Modulation layer at the bottleneck of a Video Swin Transformer. A learned density embedding produces channel-wise scale and shift parameters, allowing the decoder to reconstruct saliency from features selected for each density regime. The module adds only ~100K parameters and can use either CrowdFix density labels or the model's own density prediction. On CrowdFix, DensFiLM achieves mean NSS 1.434 and CC 0.517 over four seeds, improving over ACLNet by 14.7% and 14.9%, respectively, while predicted-density conditioning matches oracle-label performance. Ablations show that explicit RAFT optical flow and larger temporal and social-force extensions provide no further improvement in this setting. In a centre-prior-subtraction diagnostic, density conditioning yields an NSS gain of 0.462 over the unconditioned backbone, compared with 0.124 under standard evaluation. These results show that lightweight bottleneck conditioning provides a more effective inductive bias than increasing model capacity for crowd-video saliency. Our code is available at https://github.com/aniskhan25/crowdfix-saliency.
Chinese Translation
视频显著性模型通常在拥挤场景中应用单一的注视策略，尽管人群密度会系统性地改变注意力。稀疏场景鼓励跟踪个体，而密集场景则将注意力转向集体运动和场景级地标。我们提出了DensFiLM，一种基于密度条件的视频显著性模型，该模型在视频Swin Transformer的瓶颈处插入了轻量级的特征线性调制（Feature-wise Linear Modulation）层。学习到的密度嵌入生成通道级的缩放和偏移参数，使解码器能够从为每个密度状态选择的特征中重建显著性。该模块仅增加约10万参数，并可以使用CrowdFix密度标签或模型自身的密度预测。在CrowdFix上，DensFiLM在四个种子上实现了平均NSS为1.434和CC为0.517，分别比ACLNet提高了14.7%和14.9%，同时预测的密度条件与oracle标签性能相匹配。消融实验表明，在这种设置下，显式的RAFT光流和更大的时间及社会力扩展并未提供进一步的改善。在中心优先减法诊断中，密度条件相比于未条件化的主干网络，获得了0.462的NSS增益，而在标准评估下仅为0.124。这些结果表明，轻量级的瓶颈条件提供了比增加模型容量更有效的归纳偏差，用于人群视频显著性。我们的代码可在https://github.com/aniskhan25/crowdfix-saliency获取。
cs.CV / 43 / 2607.25467
Seen, Said, or Forgotten? A Causal Audit of Visual KV Memory Across Dialog Turns
看见、说出还是遗忘？对对话轮次中视觉KV记忆的因果审计
Abstract
Stateful multimodal assistants encode an image once but may answer questions about it many turns later. Attention-guided visual-KV eviction assumes that evidence irrelevant now will remain dispensable, although future questions are unknown. We ask when a visual fact is actually safe to forget and introduce the Causal Visual Memory Audit (CVMA), a paired single-prefill framework that tests what later answers lose when a visual region, the whole image, or prior assistant text becomes unavailable. On VisDial and ConvBench, current attention can rank future-useful regions worse than random even though a diagnostic marginal-utility control shows substantial selection headroom. Aggregate scores hide this failure when later turns do not need vision; controlled and stock-generated histories reveal a second escape route, in which assistant-text KV replaces image KV for facts already stated but not reliably for unstated facts. In the tested stacks, safe forgetting is supported by low future visual dependence or fact-specific verbalization---not by low current attention.
Chinese Translation
状态感知的多模态助手只需对图像进行一次编码，但可能在多轮对话中回答与之相关的问题。基于注意力引导的视觉-KV驱逐假设当前无关的证据将保持可丢弃，尽管未来的问题尚不确定。我们探讨何时视觉事实实际上可以安全遗忘，并引入因果视觉记忆审计（Causal Visual Memory Audit, CVMA），这是一种配对的单一预填充框架，用于测试当视觉区域、整个图像或先前助手文本不可用时，后续答案会损失什么。在VisDial和ConvBench上，目前的注意力可能会将未来有用的区域的排名评估得比随机选择还要差，尽管诊断性的边际效用控制显示出相当大的选择余地。当后续轮次不需要视觉信息时，汇总得分掩盖了这一失败；受控和生成的历史揭示了第二条逃逸路径，其中助手文本KV替代了图像KV，用于已经陈述的事实，但对未陈述的事实则不可靠。在测试的堆栈中，安全遗忘得到了低未来视觉依赖性或事实特定语言表达的支持，而不是低当前注意力的支持。
cs.CV / 44 / 2607.25468
Safety-Aware Cascaded Inference for Crop Damage Assessment with Controlled Error Trade-offs
安全意识的级联推理用于作物损害评估，控制误差权衡
Abstract
In picture-based agricultural insurance for smallholder farmers, missed damage detections carry substantially higher cost than false alarms: a farmer who sustained real losses receives no payout, while unnecessary expert review is operationally costly but reversible. Standard multi-class classifiers optimize global accuracy but provide no mechanism to operationalize or control this asymmetric cost structure at inference time. We propose CascadeCropNet, a two-stage cascade architecture calibrated to satisfy a target recall constraint (Rec-Damaged >= 0.95) through threshold selection. A lightweight Sentinel model performs binary health triage; samples exceeding a calibrated damage probability threshold tau are escalated to a specialist Expert model for fine-grained diagnosis. This design provides explicit, deployment-time control over the safety-efficiency trade-off without retraining. Evaluated on the Eyes on the Ground dataset (23,804 images from Kenyan smallholder maize farms), the cascade achieves Rec-Damaged = 0.974 at tau = 0.5, reducing missed damage cases by up to 54% relative to a flat baseline. Under evaluation alignment, the representational gap reduces to +0.008 F1-macro, confirming the contribution is architectural rather than representational. Under input degradation, the system prioritizes escalation over confident misclassification, reflecting error containment through architectural isolation rather than intrinsic model robustness. These results demonstrate that cascade architectures can operationalize safety-oriented decision constraints through calibrated routing in settings where reliability matters more than aggregate accuracy. These properties depend on threshold calibration and deployment conditions and do not constitute guarantees under arbitrary distribution shift.
Chinese Translation
在针对小农户的基于图像的农业保险中，漏检损害的成本远高于误报：遭受实际损失的农民将不会获得赔偿，而不必要的专家审查在操作上成本高昂但可逆。标准的多类分类器优化全局准确率，但在推理时并未提供机制来操作或控制这种不对称的成本结构。我们提出了CascadeCropNet，一种两阶段的级联架构，通过阈值选择来校准以满足目标召回约束（Rec-Damaged >= 0.95）。一个轻量级的Sentinel模型执行二元健康筛查；超过校准损害概率阈值tau的样本将被升级到专家模型进行细粒度诊断。该设计在不重新训练的情况下，提供了对安全与效率权衡的明确控制。根据“Eyes on the Ground”数据集（来自肯尼亚小农户玉米农场的23,804张图像）进行评估，该级联模型在tau = 0.5时实现了Rec-Damaged = 0.974，相较于平坦基线减少了多达54%的漏检损害案例。在评估对齐下，表示差距缩小至+0.008 F1-macro，确认该贡献是架构性的而非表示性的。在输入降级的情况下，该系统优先考虑升级而非自信的误分类，反映了通过架构隔离实现的错误控制，而非内在模型的鲁棒性。这些结果表明，级联架构可以通过校准路由在可靠性比整体准确性更重要的环境中实现安全导向的决策约束。这些特性依赖于阈值校准和部署条件，并不构成在任意分布变化下的保证。
cs.CV / 45 / 2607.25489
Agentic AI in medicine: architectures, applications, evaluation, and challenges for clinical translation
医学中的代理人工智能：架构、应用、评估及临床转化的挑战
Abstract
Large language models and multimodal foundation models are enabling medical artificial intelligence (AI) systems to move beyond isolated prediction and undertake multistep clinical tasks that require planning, tool use, memory, iterative correction, and coordination among specialized agents. However, the scope of agentic AI in medicine remains unsettled, and current evaluation practices are not yet aligned with the requirements of clinical use. We conducted a scoping review with systematic evidence mapping across five electronic sources, screened 1,649 exportable records, and provisionally included 557 unique studies that met predefined criteria for goal-directed task execution, tool use, interaction with external resources, feedback-based refinement, or multi-agent collaboration. The included studies describe single agents that use external tools, workflows supported by retrieval and external knowledge, multimodal agents, and multi-agent systems applied to medical question answering, image interpretation, electronic health record analysis, drug safety, and clinical trial prediction. The evidence base remains dominated by public benchmarks, simulated settings, retrospective datasets, and small-scale expert evaluation. Process reliability, evidence traceability, uncertainty, safety, workflow impact, and external validity are evaluated less consistently. Clinical translation will depend on clearer definitions, reproducible evaluation, auditable oversight, interoperable system design, and prospective validation in real-world clinical workflows.
Chinese Translation
大型语言模型和多模态基础模型使医学人工智能（AI）系统能够超越孤立的预测，承担需要规划、工具使用、记忆、迭代修正和专业代理之间协调的多步骤临床任务。然而，医学中代理人工智能的范围尚未确定，当前的评估实践尚未与临床使用的要求对齐。我们进行了范围审查，并在五个电子来源中进行了系统的证据映射，筛选了1,649条可导出记录，初步纳入了557项符合预定义目标导向任务执行、工具使用、与外部资源互动、基于反馈的改进或多代理协作标准的独特研究。这些纳入的研究描述了使用外部工具的单一代理、由检索和外部知识支持的工作流程、多模态代理以及应用于医学问答、图像解读、电子健康记录分析、药物安全和临床试验预测的多代理系统。证据基础仍然以公共基准、模拟环境、回顾性数据集和小规模专家评估为主。过程可靠性、证据可追溯性、不确定性、安全性、工作流程影响和外部有效性等方面的评估不够一致。临床转化将依赖于更清晰的定义、可重复的评估、可审计的监督、可互操作的系统设计以及在真实临床工作流程中的前瞻性验证。
cs.CV / 46 / 2607.25497
Beyond Counts: A Distributional Robustness Margin For Pathology Foundation Models
超越计数：病理基础模型的分布鲁棒性边际
Abstract
Pathology foundation models are approaching clinical deployment, yet remain vulnerable to systematic non-biological variation across centres. Differences in tissue preparation, staining and scanning are strongly encoded in their representations, enabling shortcut learning and weakening generalisation across cohorts and institutions. The Robustness Index (RI) quantifies whether local representation geometry is dominated by biology or by non-biological variation, but its count-based formulation discards distance information. We show that adding distance weights changes little because the deeper limitation lies in RI's pooled, fixed-neighbourhood design, which obscures sample-level heterogeneity and effectively evaluates only a model-dependent subset of samples. We introduce the Cross-confounder Robustness Margin (CRoMa), a sample-resolved measure that directly compares distances to cross-confounder biological matches and same-confounder biological distractors. CRoMa recasts robustness as a cohort-wide margin distribution rather than a single pooled score. We evaluated frozen representations from 20 tile-level encoders across three benchmarks and 4 slide-level encoders on a fourth. Rankings by median CRoMa were broadly consistent across datasets, while the underlying distributions revealed substantial within-model heterogeneity. Every tile encoder retained a confounder-dominated lower tail, whose prevalence and severity varied markedly across models. These distinct robustness profiles frame model selection as a Pareto trade-off between typical and lower-tail robustness. Higher CRoMa was also associated with smaller shortcut-induced performance drops after supervised adaptation. By turning representation geometry into a distributional robustness readout that anticipates downstream shortcut susceptibility, CRoMa provides a principled basis for robustness assessment and model selection.
Chinese Translation
病理基础模型正接近临床部署，但仍然容易受到中心间系统性非生物变异的影响。组织准备、染色和扫描的差异在其表示中被强烈编码，导致捷径学习并削弱了在不同队列和机构间的泛化能力。鲁棒性指数（Robustness Index, RI）量化了局部表示几何是否受到生物因素或非生物变异的主导，但其基于计数的公式丢弃了距离信息。我们展示了添加距离权重的改变不大，因为更深层次的限制在于RI的汇总固定邻域设计，这掩盖了样本级异质性，并有效地仅评估模型依赖的样本子集。我们引入了交叉混淆鲁棒性边际（Cross-confounder Robustness Margin, CRoMa），这是一种样本解析度的度量，直接比较与交叉混淆生物匹配和同混淆生物干扰者的距离。CRoMa将鲁棒性重塑为一个队列范围内的边际分布，而不是单一的汇总分数。我们评估了来自20个瓷砖级编码器的冻结表示，涵盖三个基准和一个第四个滑动级编码器。按中位数CRoMa排名在各数据集间大体一致，而潜在分布揭示了模型内显著的异质性。每个瓷砖编码器都保留了一个以混淆因素为主导的下尾，其普遍性和严重性在模型间显著不同。这些不同的鲁棒性特征将模型选择框架化为典型鲁棒性与下尾鲁棒性之间的帕累托权衡。更高的CRoMa也与经过监督适应后较小的捷径引起的性能下降相关。通过将表示几何转化为一种分布鲁棒性读数，CRoMa预见了下游捷径的易感性，为鲁棒性评估和模型选择提供了原则性基础。
cs.CV / 47 / 2607.25503
Group Equivariant Diffusion for Anomaly Detection in Computational Cytology
用于计算细胞学异常检测的群等变扩散
Abstract
Computational cytology on whole-slide images is challenging because malignant cells are rare, heterogeneous, and annotated slides are scarce. Anomaly detection frameworks can be trained on normal slide-negative patches and then applied at test time to flag abnormal patches in held-out slides. Most unsupervised anomaly detection approaches including generative ones (GAN-based and diffusion-based), are tuned to organ-level imaging and require large curated datasets. In cytology the signal is cell-centric: rotating or flipping a single-cell patch does not change its diagnostic class, yet standard diffusion models treat transformed views as distinct inputs, leading to transformation-dependent reconstructions and unstable anomaly scores. We propose a D4-equivariant diffusion framework that enforces rotation and reflection symmetry both architecturally, via a D4-equivariant U-Net, and at inference, via equivariant noise coupling and (optionally) frame averaging. This alignment with biological invariance yields transformation-consistent pseudo-healthy reconstructions and more stable anomaly ranking under symmetry. On two publicly available cytology datasets of bone marrow and peripheral blood smears, our D4-equivariant diffusion models achieve higher AUC and retrieve more abnormal cells in the top K predictions than non-equivariant generative baselines, a deep one-class, and a multiple instance learning based method, while substantially reducing score variance across rotations and flips. Code is available at https://swchmida.github.io/D4diffCyto/.
Chinese Translation
在全切片图像上的计算细胞学面临挑战，因为恶性细胞稀少且异质，且标注的切片也很稀缺。异常检测框架可以在正常切片负片段上进行训练，然后在测试时应用于标记保留切片中的异常片段。大多数无监督异常检测方法，包括生成式方法（基于GAN和基于扩散），都针对器官级成像进行调优，并且需要大量经过整理的数据集。在细胞学中，信号是以细胞为中心的：旋转或翻转单细胞片段并不会改变其诊断类别，但标准扩散模型将变换后的视图视为不同的输入，导致依赖于变换的重建和不稳定的异常评分。我们提出了一种D4等变扩散框架，通过D4等变U-Net在架构上以及通过等变噪声耦合和（可选的）帧平均在推理时强制执行旋转和反射对称性。这种与生物不变性的对齐产生了一致的伪健康重建，并在对称性下提供了更稳定的异常排名。在两个公开可用的骨髓和外周血涂片细胞学数据集上，我们的D4等变扩散模型在AUC上取得了更高的成绩，并在前K个预测中检索到更多异常细胞，相较于非等变生成基线、深度单类方法和基于多实例学习的方法，同时显著减少了旋转和翻转下的评分方差。代码可在 https://swchmida.github.io/D4diffCyto/ 获取。
cs.CV / 48 / 2607.25522
I2VShield: An Efficient Proactive Defense Framework against DiT-based Image-to-Video Models
I2VShield：针对基于DiT的图像到视频模型的高效主动防御框架
Abstract
The rapid advancement of video generation models has led to the increasing misuse of image-to-video (I2V) models. Although substantial progress has been made in detecting AI-generated videos, proactive defenses against I2V models remain underexplored. In particular, current proactive defenses against I2V models predominantly rely on gradient-based adversarial attacks, which require defenders to possess GPUs with substantial memory resources (VRAM) to generate adversarial examples. To address this issue, we propose I2VShield, a privacy protection method based on generative adversarial attacks tailored to Diffusion Transformer (DiT)-based I2V models. The proposed method primarily consists of two components: (1) a text-adaptive perturbation generation framework integrating adversarial learning to mitigate computational overhead while maintaining visual imperceptibility; and (2) an untargeted Multimodal Attention Disruption (MAD) attack that exploits the inherent vulnerabilities of DiT-based I2V models, maximizing the deviation of the internal attention features from their clean states. Extensive experiments demonstrate that our approach achieves highly competitive protection performance across various datasets and mainstream DiT-based I2V models, particularly in disrupting spatiotemporal coherence, while substantially reducing computational costs.
Chinese Translation
视频生成模型的快速发展导致了图像到视频（I2V）模型的日益滥用。尽管在检测AI生成视频方面取得了显著进展，但针对I2V模型的主动防御仍然未得到充分探索。特别是，目前针对I2V模型的主动防御主要依赖于基于梯度的对抗攻击，这要求防御者拥有具备大量内存资源（VRAM）的GPU来生成对抗样本。为了解决这一问题，我们提出了I2VShield，一种基于生成对抗攻击的隐私保护方法，专门针对基于扩散变换器（Diffusion Transformer, DiT）的I2V模型。该方法主要由两个组件组成：（1）一个文本自适应扰动生成框架，结合对抗学习，以减轻计算开销，同时保持视觉不可感知性；（2）一种无目标的多模态注意力干扰（Multimodal Attention Disruption, MAD）攻击，利用基于DiT的I2V模型固有的脆弱性，最大化内部注意力特征与其干净状态之间的偏差。大量实验表明，我们的方法在各种数据集和主流基于DiT的I2V模型中实现了高度竞争的保护性能，特别是在干扰时空一致性方面，同时显著降低了计算成本。
cs.CV / 49 / 2607.25524
ReLATE: Reliability-Guided Evidence Fusion for Robust UAV--Satellite cross-view Geo-Localization
ReLATE：基于可靠性引导的证据融合用于稳健的无人机-卫星交叉视角地理定位
Abstract
Unmanned aerial vehicle (UAV)-satellite cross-view geo-localization matches UAV images against satellite imagery and has achieved impressive accuracy on clean (non-degraded) image benchmarks. In real-world flights, however, UAV observations are frequently affected by adverse weather, illumination changes, platform motion, sensor noise, and compression, while the robustness of existing methods under such degradations remains largely unexamined. In this paper, we present UAVSat-Deg, a large-scale robustness benchmark for degraded UAV-satellite geo-localization, comprising University-1652-Deg and SUES-200-Deg. UAVSat-Deg covers 27 corruption types, including 19 core and 8 compound corruptions, at three severity levels, supports bidirectional drone-to-satellite and satellite-to-drone retrieval as well as multi-height UAV acquisition, and contains more than 11.7 million pre-generated corrupted test images. Benchmarking representative methods under this protocol reveals substantial robustness gaps, particularly under severe and compound corruptions. To address this problem, we propose ReLATE, a Reliable Evidence Learning framework with Adaptive Token Evidence Regulation, which realizes reliability-adaptive feature fusion during descriptor construction. ReLATE estimates a structure-smoothed reliability field over visual tokens, aggregates trustworthy local evidence, and adaptively integrates it into query-derived representations; the regulated query representations are then combined with the CLS-token and GeM-pooled branches to form the final cross-view descriptor. Across both test sets and retrieval directions, ReLATE achieves the best average corrupted-test performance among the compared methods while maintaining competitive accuracy on clean images. The code and dataset will be available at https://github.com/JHC626/ReLATE.
Chinese Translation
无人机（UAV）-卫星交叉视角地理定位通过将无人机图像与卫星影像进行匹配，在干净（未退化）图像基准测试中取得了令人印象深刻的准确性。然而，在实际飞行中，无人机观测常常受到恶劣天气、光照变化、平台运动、传感器噪声和压缩的影响，而现有方法在这些退化条件下的稳健性仍然未得到充分检验。本文提出了UAVSat-Deg，这是一个针对退化无人机-卫星地理定位的大规模稳健性基准，包含University-1652-Deg和SUES-200-Deg。UAVSat-Deg涵盖27种退化类型，包括19种核心退化和8种复合退化，分为三个严重程度级别，支持双向无人机到卫星和卫星到无人机的检索，以及多高度无人机采集，包含超过1170万张预生成的退化测试图像。在该协议下对代表性方法的基准测试揭示了显著的稳健性差距，尤其是在严重和复合退化情况下。为了解决这一问题，我们提出了ReLATE，一个具有自适应令牌证据调节的可靠证据学习框架，该框架在描述符构建过程中实现了可靠性自适应特征融合。ReLATE在视觉令牌上估计结构平滑的可靠性场，聚合可信的局部证据，并自适应地将其整合到查询派生的表示中；调节后的查询表示随后与CLS-token和GeM池化分支结合，形成最终的交叉视角描述符。在两个测试集和检索方向上，ReLATE在比较方法中实现了最佳的平均退化测试性能，同时在干净图像上保持了竞争力的准确性。代码和数据集将发布在https://github.com/JHC626/ReLATE。
cs.CV / 50 / 2607.25527
Argus-Unified: Towards A Compact and Economical Unified Model for Image Understanding and Generation
Argus-Unified：朝着一个紧凑且经济的统一图像理解与生成模型迈进
Abstract
Unifying visual understanding and generation in one model holds immense promise, but remains challenging and expensive due to heavy compute and data demands and conflicts between the visual features needed for these two capabilities. To address these challenges, we present Argus-Unified, a compact, effective and unified multimodal model built with low demand on computation and data. Instead of aligning modalities from scratch, Argus-Unified effectively leverages pretrained vision-language models (VLMs) that provide strong multimodal priors. Specifically, we introduce hybrid visual tokens that preserve continuous tokens for understanding while learning discrete tokens for generation from a frozen unified vision encoder. Our training pipeline includes two stages: the first stage learns a quantizer and image decoder on top of the frozen vision encoder, the second stage trains the LLM initialized from a pretrained VLM for the unified multimodal modeling. Using by far the least amount of data (15.6M) and the lowest cost (~$2,000), we demonstrate that unified multimodal models can be trained economically while achieving strong performance in both understanding and generation. Notably, our model attains state-of-the-art multimodal understanding on GQA, POPE, and VQAv2, and competitive generation quality compared to models with dedicated vision encoders (e.g., Janus, Janus-Pro), all at ~10x lower cost and with ~5x less data. We envision Argus-Unified as a useful baseline that lowers the development barrier for unified models.
Chinese Translation
将视觉理解与生成统一为一个模型具有巨大的潜力，但由于对计算和数据的高需求以及这两种能力所需视觉特征之间的冲突，这一目标仍然具有挑战性且成本高昂。为了解决这些挑战，我们提出了Argus-Unified，这是一种紧凑、高效且统一的多模态模型，具有较低的计算和数据需求。Argus-Unified并不是从头开始对齐模态，而是有效利用预训练的视觉-语言模型（VLMs），这些模型提供了强大的多模态先验。具体而言，我们引入了混合视觉标记，这些标记在理解时保留连续标记，而在生成时从冻结的统一视觉编码器学习离散标记。我们的训练流程包括两个阶段：第一阶段在冻结的视觉编码器上学习量化器和图像解码器，第二阶段则训练从预训练VLM初始化的语言模型（LLM）以实现统一的多模态建模。我们使用迄今为止最少的数据（15.6M）和最低的成本（约$2,000），展示了统一多模态模型可以以经济的方式进行训练，同时在理解和生成方面都能取得强大的性能。值得注意的是，我们的模型在GQA、POPE和VQAv2上达到了最先进的多模态理解水平，并且在生成质量上与具有专用视觉编码器的模型（例如Janus、Janus-Pro）相比具有竞争力，所有这些都以约10倍更低的成本和约5倍更少的数据实现。我们设想Argus-Unified作为一个有用的基准，降低了统一模型的开发门槛。
cs.CV / 51 / 2607.25537
Visual prompt engineering for video models
视频模型的视觉提示工程
Abstract
In the age of foundation models, a model is only as good as its prompt. For this reason, prompt engineering has become an essential technique for improving language model performance. Since video models are currently becoming foundation models for visual tasks (e.g., visual reasoning), we here ask whether they similarly benefit from visual prompt engineering: automatically modifying the task image to improve model performance. For example, for a visual physics reasoning task ("Where does the ball land, after passing a set of obstacles?"), an abstract sketch-like scene can be turned into a photorealistic version with a simple call to an image editing model. We find that visual prompt engineering, or VIPE for short, improves video reasoning performance across tasks. In fact, for video models, visual prompt engineering can be even more effective than classic text-based prompt engineering or test-time scaling. Ultimately, just as text-based prompt engineering systematically improves language model performance, visual prompt engineering can serve as a simple, compute-efficient approach to elicit better visual reasoning performance from video models. Example videos on our project page at https://visual-prompt-engineering.github.io/.
Chinese Translation
在基础模型的时代，模型的表现取决于其提示。因此，提示工程已成为提高语言模型性能的重要技术。由于视频模型目前正在成为视觉任务（例如视觉推理）的基础模型，我们在此探讨它们是否同样受益于视觉提示工程：自动修改任务图像以提升模型性能。例如，对于一个视觉物理推理任务（“在经过一系列障碍后，球会落在哪里？”），可以通过简单调用图像编辑模型将抽象的草图场景转变为逼真的版本。我们发现，视觉提示工程（简称 VIPE）在各项任务中提高了视频推理性能。实际上，对于视频模型而言，视觉提示工程的效果甚至可能优于经典的基于文本的提示工程或测试时缩放。最终，就像基于文本的提示工程系统性地提高语言模型性能一样，视觉提示工程可以作为一种简单且计算高效的方法，从视频模型中引出更好的视觉推理性能。有关示例视频，请访问我们的项目页面 https://visual-prompt-engineering.github.io/。
cs.CV / 52 / 2607.25543
Less is More: Modality-Decoupling for General AIGC Audio-Video Detection
少即是多：用于通用AIGC音视频检测的模态解耦
Abstract
Generative AI has rapidly expanded audio-visual forgery beyond human-centric deepfakes into general scenes. Existing AIGC detection methods assume audio-visual content correspondence, identifying forgeries by spotting cross-modal inconsistencies. However, we empirically find that this assumption does not consistently hold in general scenarios. We argue that, for general audio-visual AIGC detection, decision-level fusion is a more robust alternative to feature-level fusion. Therefore, we propose DAV-Det, a decoupled audio-visual AIGC detection system that independently models forensic evidence from each modality. The visual detector leverages multi-granularity representations at global, patch, and segment levels to capture spatial forgery cues, while the audio detector exploits both temporal and spectral irregularities via a gated temporal-spectral dual-branch architecture to model acoustic artifacts. Our method ranks 1st in the General AIGC Audio-Video Detection Challenge of the IJCAI-ECAI 2026 DDL 2.0 Workshop, with a final score of 0.8460. Code is available at https://github.com/tuffy-studio/DAV-Det.
Chinese Translation
生成性人工智能迅速将音视频伪造扩展到超越以人为中心的深度伪造，涵盖一般场景。现有的AIGC检测方法假设音视频内容之间存在对应关系，通过识别跨模态不一致性来识别伪造。然而，我们实证发现这一假设在一般场景中并不总是成立。我们认为，对于通用音视频AIGC检测，决策层融合是一种比特征层融合更为稳健的替代方案。因此，我们提出了DAV-Det，一个解耦的音视频AIGC检测系统，独立建模每种模态的取证证据。视觉检测器利用全局、补丁和片段级别的多粒度表示来捕捉空间伪造线索，而音频检测器则通过门控时频双分支架构利用时间和频谱不规则性来建模声学伪造。我们的方法在IJCAI-ECAI 2026 DDL 2.0研讨会的通用AIGC音视频检测挑战中排名第一，最终得分为0.8460。代码可在 https://github.com/tuffy-studio/DAV-Det 获取。
cs.CV / 53 / 2607.25563
Few-Shot Open-Vocabulary Remote Sensing Segmentation via Textual Inversion
通过文本反演实现少样本开放词汇遥感分割
Abstract
Open-vocabulary segmentation labels arbitrary categories from a text query without per-class training, yet on remote sensing imagery it underperforms on categories it handles reliably elsewhere. We find that much of this gap traces to the text query rather than to the segmentation model. Because these models are not specialized for overhead imagery, the class name that serves as the query is often a weak address into the vision-language embedding space. We show that a better name repairs part of the gap, while the remaining failures call for an address that the tested natural-language rephrasings do not provide. We recover that address from a few examples through textual inversion on a frozen model, keeping inference text only. On a representative benchmark this raises the mean intersection over union on the affected categories from 3.9 to 39.4, and across eight remote sensing datasets it improves over few-shot methods that instead inject visual prompts at inference.
Chinese Translation
开放词汇分割能够根据文本查询为任意类别标注，而无需进行每类的训练，但在遥感图像上，它在处理某些类别时的表现不如在其他地方可靠。我们发现，这一差距主要源于文本查询，而非分割模型。由于这些模型并未针对高空图像进行专门化，因此作为查询的类名往往在视觉-语言嵌入空间中是一个较弱的指向。我们展示了一个更好的名称可以修复部分差距，而剩余的失败则需要一个测试的自然语言重述所无法提供的指向。我们通过在一个冻结模型上进行文本反演，从少量示例中恢复该指向，仅保持推理文本。在一个具有代表性的基准测试中，这使受影响类别的平均交并比从3.9提高到39.4，并且在八个遥感数据集上，它的表现优于通过在推理时注入视觉提示的少样本方法。
cs.CV / 54 / 2607.25565
ReDesign: Recovering Editable Design Structures from Images via Agentic Decomposition
ReDesign：通过代理分解从图像中恢复可编辑设计结构
Abstract
Recovering an editable design file from a raster image is a common and costly bottleneck in modern design workflows, yet remains challenging since editability depends on recovering multi-modal attributes, such as typography, vector geometry, colors, grouping, and layer ordering. We present ReDesign, an agentic framework that grows an editable layer hierarchy by selecting and composing specialized tools across modalities. To keep this long decision process reliable despite imperfect tool outputs, we introduce graceful verification at each expansion, which provides local accept, prune, or retry feedback that prevents error accumulation and avoids large scale reruns. To evaluate editability at scale, we introduce the Figma Edit Replay Benchmark, consisting of 909 raw Figma files and 14,796 controlled edit instructions that replay edits on reconstructed outputs. Across this benchmark and standard reconstruction metrics, ReDesign achieves strong visual fidelity while delivering the highest editability across layout, color, and text edits, outperforming layered decomposition baselines and serial tool use pipelines.
Chinese Translation
从光栅图像中恢复可编辑设计文件是现代设计工作流程中的一个常见且昂贵的瓶颈，但由于可编辑性依赖于恢复多模态属性（如排版、矢量几何、颜色、分组和图层顺序），因此仍然具有挑战性。我们提出了ReDesign，一个代理框架，通过在不同模态中选择和组合专业工具，构建可编辑的图层层次结构。为了在这一漫长的决策过程中保持可靠性，尽管工具输出不完美，我们在每次扩展时引入了优雅的验证机制，提供局部接受、修剪或重试反馈，以防止错误累积并避免大规模重跑。为了在大规模上评估可编辑性，我们引入了Figma Edit Replay Benchmark，该基准包含909个原始Figma文件和14,796条控制编辑指令，用于在重建输出上重放编辑。在该基准和标准重建指标中，ReDesign在视觉保真度方面表现出色，同时在布局、颜色和文本编辑方面提供了最高的可编辑性，超越了分层分解基线和串行工具使用管道。
cs.CV / 55 / 2607.25570
The LAIA Dataset: Labelled Attention for Intelligent Automobiles
LAIA 数据集：智能汽车的标注注意力
Abstract
The development of autonomous vehicles (AVs) usually relies heavily on data-driven artificial intelligence (AI) models that require large volumes of sensor data with ground-truth annotations. While modular architectures are widely used, end-to-end driving paradigms offer a promising alternative by directly mapping sensor inputs to control actions. However, their adoption is limited by challenges in interpretability and explainability. To address this, we present LAIA (Labelled Attention for Intelligent Automobiles), a novel synthetic dataset designed to enrich end-to-end driving research with human attention data. Collected using the CARLA simulator in closed-loop environments, LAIA comprises over 15 hours of driving from 44 participants across carefully crafted scenarios designed to evoke natural responses. Each sequence includes RGB images under six weather conditions, semantic and instance segmentation, depth, optical flow, CAN bus signals, and synchronized eye-tracking data. LAIA enables applications including training attention-aware end-to-end AI drivers, predicting driver behavior, developing methods to detect anomalous driver-attention patterns, and improving model explainability. In this work, we use LAIA to compare human attention with the perceptual attention emerging in our end-to-end driving models, thereby providing insight into their behavior.
Chinese Translation
自主车辆（AV）的发展通常高度依赖于数据驱动的人工智能（AI）模型，这些模型需要大量带有真实标签的传感器数据。虽然模块化架构被广泛使用，但端到端驾驶范式通过直接将传感器输入映射到控制动作提供了一个有前景的替代方案。然而，由于可解释性和可解释性方面的挑战，其采用受到限制。为了解决这个问题，我们提出了 LAIA（智能汽车的标注注意力），这是一个新颖的合成数据集，旨在通过人类注意力数据丰富端到端驾驶研究。LAIA 使用 CARLA 模拟器在闭环环境中收集，包含来自 44 名参与者的超过 15 小时的驾驶数据，涵盖精心设计的场景，以引发自然反应。每个序列包括在六种天气条件下的 RGB 图像、语义和实例分割、深度、光流、CAN 总线信号和同步的眼动追踪数据。LAIA 支持多种应用，包括训练关注注意力的端到端 AI 驾驶员、预测驾驶员行为、开发检测异常驾驶员注意力模式的方法，以及提高模型的可解释性。在本研究中，我们使用 LAIA 比较人类注意力与我们端到端驾驶模型中出现的感知注意力，从而提供对其行为的洞察。
cs.CV / 56 / 2607.25589
Forensic Reproducibility Audit of a Radiology Vision-Language Model Benchmark: From Intended Protocol to Released Artifact
放射学视觉-语言模型基准的法医学可重复性审计：从预期协议到发布的工件
Abstract
Medical-imaging AI benchmarks combine datasets, DICOM rendering, prompts, provider APIs, automated labels, statistical code, manuscripts, and repository releases. Agreement across these artifacts is usually assumed rather than tested. We performed a retrospective forensic reproducibility audit of a preserved chest-radiograph vision-language model (VLM) pilot; no model was called again and no image or report was newly annotated. We traced prompt bindings, DICOM metadata, output completeness, label extraction, matched analyses, and release propagation. Of 300 planned model-prompt calls, 297 yielded nonempty reports. Sixty Claude calls labeled A/B were executed with the same C prompt. The 30 studies represented 28 patients. Four MONOCHROME1 images were rendered without required polarity inversion, dataset split membership was not retained, and the unvalidated extractor truncated five reports to 4000 characters. Reconstructing one common cohort of 369 complete case-finding blocks changed Cochran's Q from 154.73 to 182.29. Of 45 McNemar comparisons, 27 had unadjusted p < 0.05 and 20 remained below 0.05 after Holm adjustment. These values describe only the archived automated-label matrix; they do not recover the intended prompt comparison or establish clinical performance. We withdraw the original performance, ranking, prompt-effect, and clinical claims and specify machine-verifiable controls for cohort, DICOM rendering, prompt and model identity, call status, annotation provenance, keyed analysis, and derived artifacts.
Chinese Translation
医学影像人工智能基准结合了数据集、DICOM渲染、提示、提供者API、自动标签、统计代码、手稿和仓库发布。这些工件之间的协议通常被假定而非经过测试。我们对一个保存的胸部X光视觉-语言模型（VLM）试点进行了回顾性的法医学可重复性审计；没有再次调用模型，也没有对图像或报告进行新的注释。我们追踪了提示绑定、DICOM元数据、输出完整性、标签提取、匹配分析和发布传播。在300个计划的模型-提示调用中，297个产生了非空报告。60个标记为A/B的Claude调用使用了相同的C提示。这30项研究代表了28名患者。四个MONOCHROME1图像在没有所需极性反转的情况下被渲染，数据集拆分成员资格未被保留，未经验证的提取器将五个报告截断至4000个字符。重建一个包含369个完整病例发现块的共同队列将Cochran的Q值从154.73更改为182.29。在45个McNemar比较中，27个的未调整p值<0.05，20个在Holm调整后仍低于0.05。这些值仅描述了存档的自动标签矩阵；它们并未恢复预期的提示比较或建立临床性能。我们撤回原始性能、排名、提示效应和临床声明，并指定机器可验证的控制措施，以确保队列、DICOM渲染、提示和模型身份、调用状态、注释来源、关键分析和派生工件的准确性。
cs.CV / 57 / 2607.25622
Beyond Facial Consistency: Personalized Person Image Generation with Holistic Identity Preservation
超越面部一致性：具有整体身份保留的个性化人像生成
Abstract
Personalized person image generation requires preserving subject identity across both local facial details and broader appearance cues. Existing methods typically emphasize only one level of identity information, leading to an inherent trade-off between facial fidelity and overall appearance consistency. To address this, we first propose a simple dual-branch baseline that unifies global appearance control and local facial control within a shared generation framework. This simple combination of different branches yields promising results, but suffers from instability in practice due to uncoordinated branch contributions. To this end, we propose Dynamic Balancing Scaling (DBS), a fine-tuning strategy for improving face and appearance identity coordination. DBS consists of two components: adaptive temporal gating, which dynamically modulates branch contributions along the denoising trajectory, and region-aware optimization, which improves the coordination of facial, appearance, and global supervision. Together, these designs alleviate persistent face-branch over-dominance and encourage more effective appearance-aware guidance. We also introduce Pexels-100, a benchmark for evaluating holistic identity consistency in personalized person generation. Experiments show that DBS achieves a better trade-off between facial fidelity and appearance consistency than existing open-source baselines, while providing a controllable basic framework for holistic identity modeling.
Chinese Translation
个性化人像生成需要在局部面部细节和更广泛的外观线索之间保持主体身份。现有方法通常只强调一种身份信息层次，导致面部逼真度和整体外观一致性之间存在固有的权衡。为了解决这个问题，我们首先提出了一个简单的双分支基线，将全局外观控制和局部面部控制统一在一个共享的生成框架内。这种不同分支的简单组合产生了令人鼓舞的结果，但由于分支贡献的不协调，实际应用中存在不稳定性。为此，我们提出了动态平衡缩放（Dynamic Balancing Scaling, DBS），这是一种改进面部和外观身份协调的微调策略。DBS由两个组件组成：自适应时间门控，动态调节分支在去噪轨迹上的贡献；区域感知优化，改善面部、外观和全局监督的协调。综上所述，这些设计缓解了面部分支的持续过度主导，并鼓励更有效的外观感知指导。我们还引入了Pexels-100，一个用于评估个性化人像生成中整体身份一致性的基准。实验表明，DBS在面部逼真度和外观一致性之间实现了比现有开源基线更好的权衡，同时为整体身份建模提供了一个可控的基本框架。
cs.CV / 58 / 2607.25641
OmniPhys: Knowledge-Graph-Driven Benchmarking and Collective Optimization for Physical Commonsense in Text-to-Image Generation
OmniPhys：基于知识图谱的文本到图像生成中的物理常识基准测试与集体优化
Abstract
While text-to-image models exhibit remarkable visual fidelity, they frequently violate fundamental physical commonsense. Existing benchmarks often rely on coarse-grained descriptions, failing to diagnose the mastery of specific physical principles. Moreover, the high stochasticity of generative processes causes current prompt optimization methods to suffer from gradient hallucinations, where optimizers are misled by transient visual artifacts rather than systemic flaws. To address these challenges, we introduce OmniPhys, a rigorous benchmark of 1,551 samples grounded in a Physical Knowledge Graph. By aligning PhET simulations with standard curricula, OmniPhys operationalizes a knowledge-to-scenario pipeline that performs diagnostic stress tests via a dual-path verification protocol. We further propose OmniPrompt, an iterative framework that treats physical alignment as a discrete optimization problem. For each query, OmniPrompt aggregates K stochastic images into a per-query feedback buffer. Across training, it further merges feedback from batches of B queries before each meta-policy update, filtering seed and query-local noise. Evaluations across 12 representative text-to-image models reveal universal physical bottlenecks. Results demonstrate that OmniPrompt significantly enhances physical consistency across diverse backbones, proving the transferability and efficacy of our evolved meta-policies. The code and data are available at https://github.com/zjukg/OmniPhys
Chinese Translation
尽管文本到图像模型展现出显著的视觉逼真性，但它们常常违反基本的物理常识。现有的基准测试通常依赖于粗略的描述，未能有效诊断特定物理原理的掌握情况。此外，生成过程的高度随机性导致当前的提示优化方法受到梯度幻觉的影响，使得优化器被瞬态视觉伪影而非系统性缺陷所误导。为了解决这些挑战，我们引入了OmniPhys，这是一个基于物理知识图谱的严格基准，包含1,551个样本。通过将PhET模拟与标准课程对齐，OmniPhys实现了一个知识到场景的管道，通过双路径验证协议进行诊断压力测试。我们进一步提出了OmniPrompt，这是一个将物理对齐视为离散优化问题的迭代框架。对于每个查询，OmniPrompt将K个随机图像聚合到每个查询的反馈缓冲区。在训练过程中，它进一步在每次元策略更新之前合并B个查询批次的反馈，过滤种子和查询局部噪声。在12个具有代表性的文本到图像模型上的评估揭示了普遍的物理瓶颈。结果表明，OmniPrompt显著增强了不同基础模型之间的物理一致性，证明了我们演变的元策略的可转移性和有效性。代码和数据可在https://github.com/zjukg/OmniPhys获取。
cs.CV / 59 / 2607.25642
Instruction-based Image Editing: A Survey on Data, Models, Evaluation, and Applications
基于指令的图像编辑：数据、模型、评估和应用的综述
Abstract
Instruction-based Image Editing (IIE) aims to transform a given image into a new one based on textual instructions. Advances in Large Language Models (LLMs) and Vision-Language Models (VLMs) have accelerated progress toward practical ``one-sentence image editing" systems. This survey presents a systematic taxonomy and comprehensive review of IIE research, structured around five core dimensions: (1) task definition and hierarchical categorization of editing operations, (2) methodologies for training data construction, (3) architectural evolution from GAN-based to diffusion and autoregressive paradigms, (4) standardized evaluation metrics and benchmark development, and (5) introduction of commercial solutions. Our analysis shows critical technological milestones across model generations. We further propose a Comprehensive, in-Depth, and Diagnostic benchmark for IIE task (CDD-IIE Bench), which can rigorously assess the multiple aspects of model performance. Through empirical comparisons of open-source solutions, we highlight their respective capabilities and limitations. Finally, we discuss future research directions to advance the field.
Chinese Translation
基于指令的图像编辑（Instruction-based Image Editing, IIE）旨在根据文本指令将给定图像转换为新图像。大型语言模型（Large Language Models, LLMs）和视觉-语言模型（Vision-Language Models, VLMs）的进展加速了实用的“单句图像编辑”系统的发展。本文综述系统地对IIE研究进行了分类和全面回顾，围绕五个核心维度进行结构化： (1) 任务定义和编辑操作的层次分类， (2) 训练数据构建的方法论， (3) 从基于生成对抗网络（GAN）的架构演变到扩散（diffusion）和自回归（autoregressive）范式， (4) 标准化评估指标和基准开发， (5) 商业解决方案的介绍。我们的分析展示了模型世代间的关键技术里程碑。我们进一步提出了一个全面、深入且具有诊断性的IIE任务基准（Comprehensive, in-Depth, and Diagnostic benchmark for IIE task, CDD-IIE Bench），该基准可以严格评估模型性能的多个方面。通过对开源解决方案的实证比较，我们突出了它们各自的能力和局限性。最后，我们讨论了推动该领域发展的未来研究方向。
cs.CV / 60 / 2607.25710
Impact Detection in Fall Events: Leveraging Spatio-Temporal Graph Convolutional Networks and Recurrent Neural Networks Using 3D Skeletons Data
跌倒事件中的冲击检测：利用时空图卷积网络和递归神经网络处理3D骨架数据
Abstract
Fall represents a significant risk of accidental death among individuals aged over 65, presenting a global health concern. A fall is defined as any event where a person loses balance and moves to an off-position, which may or may not result in an impact where the person hits the ground. While fall detection systems have achieved good results in general, impact detection within falls remains challenging. This study proposes an efficient methodology for accurately detecting impacts within fall events by incorporating 3D joints skeleton data treated as a graph using Spatio-Temporal Graph Convolutional Networks (STGCN), Gated Recurrent Unit (GRU), and Bidirectional Long Short-Term Memory (BiLSTM) layers. By pinpointing impact moments, our approach enhances precision by distinguishing between false falls and actual impacts, contributing to better healthcare resource allocation. Our methodology, evaluated using the improved 3D skeletons UP-Fall dataset, achieves accuracy exceeding 90\% across various fall scenarios. We have made this improved dataset publicly available at https://zenodo.org/records/12773013 to facilitate further research.
Chinese Translation
跌倒对65岁以上人群构成了显著的意外死亡风险，成为全球健康问题。跌倒被定义为任何一个人失去平衡并移动到不正常位置的事件，这可能导致或不导致人撞击地面。尽管跌倒检测系统在一般情况下取得了良好的结果，但在跌倒事件中的冲击检测仍然具有挑战性。本研究提出了一种高效的方法，通过结合将3D关节骨架数据视为图形的时空图卷积网络（Spatio-Temporal Graph Convolutional Networks, STGCN）、门控递归单元（Gated Recurrent Unit, GRU）和双向长短期记忆（Bidirectional Long Short-Term Memory, BiLSTM）层，准确检测跌倒事件中的冲击。通过精准识别冲击时刻，我们的方法提高了精度，能够区分虚假跌倒和实际冲击，从而有助于更好地分配医疗资源。我们的方法在改进的3D骨架UP-Fall数据集上进行评估，准确率超过90 ext{%}，适用于多种跌倒场景。我们已将这一改进的数据集公开发布于 https://zenodo.org/records/12773013，以促进进一步的研究。
cs.CV / 61 / 2607.25732
A systematic evaluation of machine learning classifiers for event-by-event background rejection in LAFOV PET scanners
针对LAFOV PET扫描仪事件逐个背景拒绝的机器学习分类器的系统评估
Abstract
The introduction of LAFOV PET scanners brings significant sensitivity gains but also a substantial increase in the background rate from accidental coincidences, phantom-scattered and detector-scattered photons. While machine learning methods have been applied to background reduction in PET imaging, they target specific background components in post-processing rather than event-by-event classification on the raw data. In this work, we formulate coincidence classification as a supervised multi-class problem and evaluate XGBoost, AdaBoost and Neural Network classifiers as pre-reconstruction filters, using Monte Carlo simulations of the Siemens Biograph Vision Quadra scanner with NEMA IEC and anthropomorphic XCAT phantoms. We investigate two feature sets: a 4-feature representation based on the Attenuation Factor, photon time difference, energy sum, and energy difference, and an extended 6-feature set that incorporates topology-based variables. A systematic robustness study via cross-phantom inference reveals that the 4-feature models generalise significantly better across different phantom geometries, with XGBoost suffering an accuracy loss of only 0.04 compared to 0.13 for the 6-feature variant. Our best models achieve accuracies of up to 0.74 and 0.69 for the NEMA IEC and XCAT phantoms, respectively, outperforming traditional geometry-based cuts. However, we show that this compact feature set not only provides limited suppression of in-phantom scattered coincidences, but it also can lead to non-trivial spatial patterns. With scattered coincidences being the dominant background component in clinical conditions, this suggests that while the method serves as an effective and geometry-agnostic replacement for traditional cut-based selection, meaningful further gains in image quality will require either larger input representations or dedicated treatment of the phantom-scattered component.
Chinese Translation
LAFOV PET扫描仪的引入带来了显著的灵敏度提升，但也大幅增加了由于偶然重合、幻影散射和探测器散射光子所导致的背景率。虽然机器学习方法已被应用于PET成像中的背景减少，但它们主要针对特定的背景成分进行后处理，而不是对原始数据进行逐事件分类。在本研究中，我们将重合分类公式化为一个监督多类问题，并评估XGBoost、AdaBoost和神经网络分类器作为重建前滤波器，使用西门子Biograph Vision Quadra扫描仪的蒙特卡罗模拟，结合NEMA IEC和人形XCAT幻影。我们研究了两组特征：基于衰减因子的4特征表示、光子时间差、能量总和和能量差，以及一个扩展的6特征集，后者包含基于拓扑的变量。通过跨幻影推断进行的系统稳健性研究表明，4特征模型在不同幻影几何形状下的泛化能力显著更好，XGBoost的准确率损失仅为0.04，而6特征变体的损失为0.13。我们的最佳模型在NEMA IEC和XCAT幻影中分别达到了高达0.74和0.69的准确率，优于传统的基于几何的切割方法。然而，我们表明，这一紧凑的特征集不仅对幻影内散射重合的抑制有限，而且可能导致非平凡的空间模式。由于散射重合是临床条件下主要的背景成分，这表明尽管该方法作为传统基于切割选择的有效且与几何无关的替代方案，但在图像质量上实现有意义的进一步提升将需要更大的输入表示或对幻影散射成分的专门处理。
cs.CV / 62 / 2607.25736
Image Quality Dependent Degradation for AI Systems
依赖于图像质量的人工智能系统退化
Abstract
Perception is one of the primary applications where neural networks outperform conventional algorithms. One example is AI systems for automated driving, which can detect pedestrians based on image data and avoid them accordingly. A substantial challenge with these AI systems is that their output depends heavily on the quality of the input images. For example, if an image is of inferior quality due to heavy contamination, such as noise or darkness, accurate predictions are hardly feasible. Additionally, various types of errors can occur, each with varying relevance to the trustworthiness of the underlying AI system. In particular, it may be more critical not to detect an existing person than to detect a person where there is none. Therefore, we want to show that we can still avoid the most critical errors in situations of inferior image quality. To achieve this, we aim to establish a fail-degraded system by lowering the network's confidence threshold based on the estimated image quality, enabling it to detect objects more cautiously in uncertain situations. Additionally, we present a novel method for estimating the quality of incoming images by comparing them to the training data using normalizing flows. We will also conduct experiments applying our method to state-of-the-art object detection. In summary, we will present a design strategy for AI-based systems in automated driving that can deal with poor-quality input data without resorting to fallback solutions. Such measures enhance trust in AI-based systems and lead to an increased provision of the AI component.
Chinese Translation
感知是神经网络超越传统算法的主要应用之一。一个例子是用于自动驾驶的人工智能系统，它能够根据图像数据检测行人并相应地避让。然而，这些人工智能系统面临的一个重大挑战是其输出在很大程度上依赖于输入图像的质量。例如，如果图像因严重污染（如噪声或昏暗）而质量较差，准确预测几乎不可行。此外，可能会出现各种类型的错误，每种错误对基础人工智能系统的可信度具有不同的相关性。特别是，未能检测到一个存在的人可能比在没有人的情况下检测到一个人更为关键。因此，我们希望展示在图像质量较差的情况下，仍然能够避免最关键的错误。为此，我们旨在通过根据估计的图像质量降低网络的置信度阈值，建立一个故障退化系统，使其能够在不确定的情况下更谨慎地检测物体。此外，我们提出了一种新方法，通过使用归一化流将输入图像与训练数据进行比较，从而估计图像质量。我们还将进行实验，将我们的方法应用于最先进的目标检测技术。总之，我们将提出一种针对自动驾驶中基于人工智能系统的设计策略，该策略能够处理低质量输入数据，而无需依赖后备解决方案。这些措施增强了对基于人工智能系统的信任，并促进了人工智能组件的更广泛应用。
cs.CV / 63 / 2607.25778
A Unified Benchmark and Modality-Adaptive Network for Day-and-Night Drone-View Geo-Localization
统一基准与模态自适应网络用于昼夜无人机视角地理定位
Abstract
Most existing drone-view geo-localization (DVGL) benchmarks contain drone imagery captured under a single illumination condition and lack geographically aligned visible drone images, infrared drone images, and satellite images from the same locations. To evaluate the generalization capability of DVGL methods under challenging illumination conditions, some methods train models on a visible benchmark and test them on an independent infrared benchmark. This protocol essentially constitutes transfer between datasets, which makes it difficult to systematically evaluate DVGL across daytime and nighttime conditions within a unified benchmark. To address this limitation, we construct IRCHN,a real-world DVGL benchmark designed for localization across different illumination conditions. IRCHN contains 26,460 images collected from 8,820 geographic locations across four representative scene categories, including farmland, coastline, forest, and urban areas. Each location provides one visible drone image, one infrared drone image, and one corresponding satellite image, which enables unified evaluation of DVGL methods across different illumination conditions and sensing modalities. We further propose the Modality-Adaptive State-Space Transport Relation Network (MASTR-Net), a DVGL framework tailored to localization under varying illumination conditions. MASTR-Net integrates modality-adaptive feature enhancement, bidirectional selective state-space relation modeling, and soft optimal transport relation alignment to jointly reduce modality gaps and view-induced structural discrepancies. Extensive experiments demonstrate that MASTR-Net outperforms existing state-of-the-art methods on IRCHN for localization under varying illumination conditions and achieves competitive performance on two infrared benchmarks, IR-VL328 and CVGL-RGBT. Code: https://github.com/SongtianhaoXu/MASTR-Net
Chinese Translation
现有的大多数无人机视角地理定位（DVGL）基准包含在单一光照条件下捕获的无人机图像，并且缺乏来自相同位置的地理对齐的可见无人机图像、红外无人机图像和卫星图像。为了评估DVGL方法在具有挑战性的光照条件下的泛化能力，一些方法在可见基准上训练模型，并在独立的红外基准上进行测试。这种协议本质上构成了数据集之间的迁移，这使得在统一基准下系统地评估昼夜条件下的DVGL变得困难。为了解决这一限制，我们构建了IRCHN，一个旨在不同光照条件下进行定位的真实世界DVGL基准。IRCHN包含从8820个地理位置收集的26460张图像，涵盖四个代表性场景类别，包括农田、海岸线、森林和城市区域。每个位置提供一张可见无人机图像、一张红外无人机图像和一张对应的卫星图像，从而实现了在不同光照条件和感知模态下对DVGL方法的统一评估。我们进一步提出了模态自适应状态空间传输关系网络（MASTR-Net），这是一个针对不同光照条件下定位的DVGL框架。MASTR-Net集成了模态自适应特征增强、双向选择性状态空间关系建模和软最优传输关系对齐，以共同减少模态差异和视角引起的结构差异。大量实验表明，MASTR-Net在IRCHN上优于现有的最先进方法，能够在不同光照条件下进行定位，并在两个红外基准IR-VL328和CVGL-RGBT上实现了竞争性能。代码：https://github.com/SongtianhaoXu/MASTR-Net
cs.CV / 64 / 2607.25788
GeoMFD: Continual Drone-View Geo-Localization with Geometry-Aware Adapter and Margin-Field Distillation
GeoMFD：具有几何感知适配器和边际场蒸馏的持续无人机视角地理定位
Abstract
Existing drone-view geo-localization (DVGL) methods are mainly developed under a static training paradigm, where models are optimized for fixed environments with all training data available in advance. However, this paradigm is difficult to extend to real-world deployment, where drones may encounter diverse environments and require multiple environment-specific models, resulting in additional storage and model-selection costs. Directly adapting a single model to new environments also risks distorting previously learned cross-view embedding geometry and causing forgetting. To address these challenges, we formalize the continual drone-view geo-localization (C-DVGL) setting and propose GeoMFD, a geometry-aware continual adaptation method for DVGL. GeoMFD combines a cold-start bootstrapping strategy (CBS), a geometry-aware adapter (Geo-Adapter), and margin-field distillation (MFD) to balance adaptation and cross-view geometry preservation. CBS initializes a stable embedding space, Geo-Adapter enables environment adaptation through controlled residual corrections, and MFD preserves similarity margins between positive pairs and hard negatives to alleviate cross-view geometry forgetting. Extensive experiments demonstrate that GeoMFD effectively mitigates forgetting and achieves competitive performance with environment-specific DVGL methods using a single continuously updated model.
Chinese Translation
现有的无人机视角地理定位（DVGL）方法主要是在静态训练范式下开发的，其中模型针对固定环境进行优化，并且所有训练数据在开始时就可用。然而，这种范式难以扩展到现实世界的部署中，因为无人机可能会遇到多样化的环境，并且需要多个特定于环境的模型，这导致额外的存储和模型选择成本。直接将单一模型适应于新环境也存在扭曲先前学习的跨视图嵌入几何结构和造成遗忘的风险。为了解决这些挑战，我们正式定义了持续无人机视角地理定位（C-DVGL）设置，并提出了GeoMFD，一种用于DVGL的几何感知持续适应方法。GeoMFD结合了冷启动自举策略（CBS）、几何感知适配器（Geo-Adapter）和边际场蒸馏（MFD），以平衡适应和跨视图几何保留。CBS初始化一个稳定的嵌入空间，Geo-Adapter通过受控的残差修正实现环境适应，而MFD则保留正样本对和困难负样本之间的相似性边际，以减轻跨视图几何遗忘。大量实验表明，GeoMFD有效减轻了遗忘，并且在使用单一持续更新模型的情况下，达到了与特定环境DVGL方法竞争的性能。
cs.CV / 65 / 2607.25789
Towards Faithful Sentimental Image Captioning via Evidence-Aware Multi-Agent Reasoning
通过证据感知的多智能体推理实现忠实的情感图像描述
Abstract
Sentimental Image Captioning (SIC) requires balancing emotional expression with visual fidelity. Existing methods often struggle with this trade-off, leading to hallucinations due to insufficient local grounding and the lack of sentimental verification mechanisms. To address these limitations, we propose SEA-Cap, a Sentiment-Evidence-Aware Multi-Agent System for faithful and evidence-grounded sentimental image captioning. SEA-Cap incorporates a Sentiment Evidence Miner that extracts structured, local affective cues to shift sentiment control from global attributes to verifiable object-level evidence. Leveraging this evidence, our framework orchestrates a collaborative workflow where a Generator, Hallucination Checker, and Arbitrator iteratively refine captions via a shared blackboard. By explicitly auditing generated content against mined visual evidence, SEA-Cap ensures both sentiment accuracy and factual consistency. Extensive experiments on two benchmark datasets demonstrate that SEA-Cap effectively mitigates hallucinations and achieves state-of-the-art performance.
Chinese Translation
情感图像描述（SIC）需要在情感表达与视觉真实之间取得平衡。现有方法往往在这一权衡中面临挑战，导致由于局部基础不足和缺乏情感验证机制而产生幻觉。为了解决这些局限性，我们提出了SEA-Cap，一种情感证据感知的多智能体系统，用于忠实且基于证据的情感图像描述。SEA-Cap结合了情感证据挖掘器，提取结构化的局部情感线索，将情感控制从全局属性转移到可验证的对象级证据。利用这些证据，我们的框架协调了一个协作工作流程，其中生成器（Generator）、幻觉检查器（Hallucination Checker）和仲裁者（Arbitrator）通过共享的黑板迭代地优化描述。通过明确审计生成内容与挖掘的视觉证据之间的一致性，SEA-Cap确保了情感的准确性和事实的一致性。在两个基准数据集上的大量实验表明，SEA-Cap有效减少了幻觉现象，并实现了最先进的性能。
cs.CV / 66 / 2607.25791
FLASH: Efficient Impact Fall Detection with Unified Hypergraph State-Space Model
FLASH：统一超图状态空间模型的高效冲击跌倒检测
Abstract
Falls represent a critical public health challenge, and accurate detection of the impact moment when an individual hits the ground is crucial for timely intervention. Existing skeleton-based methods rely on graph neural networks modeling only pairwise joint connections, failing to capture multi-joint coordination characteristic of fall impacts, while transformer-based temporal models suffer from quadratic complexity limiting real-time deployment. We propose FLASH, a novel framework integrating single-matrix hypergraph representations with Mamba's selective state-space models through adaptive feedback mechanisms for efficient impact detection. Our approach constructs biomechanically-grounded hyperedges to model functional joint coordination while leveraging Mamba's linear-time complexity to capture temporal dynamics. Experiments on UP-Fall and UMAFall datasets demonstrate that FLASH achieves state-of-the-art accuracy with real-time inference capability and strong zero-shot cross-dataset generalization, while significantly reducing computational cost compared to dual-representation and transformer-based methods. The model provides interpretable feedback through learned attention patterns aligned with biomechanical principles. Code is available at https://github.com/Tresor-Koffi/FLASH-Impact-Fall-Detection.
Chinese Translation
跌倒是一个严重的公共健康挑战，准确检测个体撞击地面的瞬间对于及时干预至关重要。现有的基于骨架的方法依赖于仅建模成对关节连接的图神经网络，未能捕捉跌倒冲击的多关节协调特征，而基于变换器的时间模型则因其二次复杂度限制了实时部署。我们提出了FLASH，一个新颖的框架，通过自适应反馈机制将单矩阵超图表示与Mamba的选择性状态空间模型相结合，以实现高效的冲击检测。我们的方法构建了生物力学基础的超边，以建模功能性关节协调，同时利用Mamba的线性时间复杂度捕捉时间动态。在UP-Fall和UMAFall数据集上的实验表明，FLASH在实时推断能力和强大的零样本跨数据集泛化方面达到了最先进的准确性，同时显著降低了与双重表示和基于变换器的方法相比的计算成本。该模型通过与生物力学原理对齐的学习注意模式提供可解释的反馈。代码可在 https://github.com/Tresor-Koffi/FLASH-Impact-Fall-Detection 获取。
cs.CV / 67 / 2607.25794
Fine-Grained Food Image Understanding via Target-Aware Data Alignment
通过目标感知数据对齐实现细粒度食品图像理解
Abstract
Fine-grained food visual--semantic understanding requires models to capture subtle distinctions across ingredients, cooking methods, doneness, color, texture, and plate composition. Although CLIP-style vision-language models provide a natural framework for this task, their effectiveness is limited when training relies on heterogeneous web-collected image--text pairs. Such data often exhibit a web-to-target domain gap and cross-modal misalignment, where images differ from the target distribution and captions are noisy, multilingual, or weakly grounded in visual content. We propose a data-centric multimodal alignment method for fine-grained food description and recognition. Our method first performs target-aware data selection to identify visually relevant training subsets, then applies VLM-based caption refinement to generate visually grounded, target-style descriptions. Using these curated image--caption pairs, we train complementary CLIP-style retrieval experts and further combine their decisions through a hierarchical VLM-assisted multi-expert decision-level fusion strategy that invokes the VLM only when experts disagree. Experiments show that our data refinement strategy significantly improves retrieval performance over naive web supervision, with VLM-based caption refinement alone yielding an average performance gain of approximately 19%. Our full method also achieves more than twice the retrieval score of pure VLM-based retrieval while remaining substantially more efficient.
Chinese Translation
细粒度食品视觉-语义理解要求模型能够捕捉成分、烹饪方法、熟度、颜色、质地和盘子组成之间的微妙差异。尽管CLIP风格的视觉-语言模型为这一任务提供了自然的框架，但当训练依赖于异构的网络收集的图像-文本对时，其有效性受到限制。这类数据往往表现出网络到目标领域的差距和跨模态的不对齐，图像与目标分布不同，且标题可能嘈杂、多语言或与视觉内容的关联较弱。我们提出了一种以数据为中心的多模态对齐方法，用于细粒度食品描述和识别。我们的方法首先执行目标感知的数据选择，以识别视觉相关的训练子集，然后应用基于VLM（视觉语言模型）的标题精炼，生成视觉上有依据的目标风格描述。利用这些精心挑选的图像-标题对，我们训练互补的CLIP风格检索专家，并通过分层的VLM辅助多专家决策级融合策略进一步结合他们的决策，该策略仅在专家意见不一致时调用VLM。实验表明，我们的数据精炼策略显著提高了检索性能，相较于简单的网络监督，基于VLM的标题精炼单独带来了约19%的平均性能提升。我们的完整方法在检索得分上也超过了纯基于VLM的检索两倍，同时保持了更高的效率。
cs.CV / 68 / 2607.25802
Explicit Layer Modeling for Video Object Insertion and Layer Decomposition
视频对象插入与层分解的显式层建模
Abstract
Most video editing systems still lack explicit layered video representations, limiting their ability to perform realistic compositing, object reuse, and consistent manipulation. This limitation is especially pronounced in video object insertion and video layer decomposition, where existing methods rely on implicit inference or per-scene optimization due to the absence of explicit foreground-layer supervision. We introduce TriLayer, a large-scale triplet video dataset containing aligned composite, background, and foreground videos, where the foreground layers include both object appearance and associated visual effects. This explicit supervision enables models to learn layered video representations directly rather than inferring them implicitly. Building on this dataset, we propose DBL-Diffusion, a dual-branch diffusion framework that jointly models RGB composites and RGBA foreground layers through shared denoising and cross-branch interaction. We instantiate the framework in two tasks: DBL-Insert for layered object insertion, which generates explicit RGBA layers for realistic compositing and flexible post-editing, and DBL-Decompose for video layer decomposition, which recovers foreground and background layers using triplet supervision. Experiments demonstrate that explicit layer modeling substantially improves both insertion fidelity and decomposition quality.
Chinese Translation
大多数视频编辑系统仍然缺乏显式的分层视频表示，这限制了它们进行真实合成、对象重用和一致操作的能力。这一限制在视频对象插入和视频层分解中尤为明显，现有方法由于缺乏显式前景层监督，依赖于隐式推断或逐场景优化。我们引入了TriLayer，这是一个大规模的三元组视频数据集，包含对齐的合成视频、背景视频和前景视频，其中前景层包括对象外观和相关视觉效果。这种显式监督使模型能够直接学习分层视频表示，而不是隐式推断。基于该数据集，我们提出了DBL-Diffusion，这是一种双分支扩散框架，通过共享去噪和跨分支交互共同建模RGB合成和RGBA前景层。我们在两个任务中实例化该框架：DBL-Insert用于分层对象插入，生成显式的RGBA层以实现真实合成和灵活的后期编辑；DBL-Decompose用于视频层分解，利用三元组监督恢复前景和背景层。实验表明，显式层建模显著提高了插入保真度和分解质量。
cs.CV / 69 / 2607.25815
Freq-RemoteVAR: Next-Frequency Autoregressive Modeling for Remote Sensing Change Detection
Freq-RemoteVAR：用于遥感变化检测的下一频率自回归建模
Abstract
Remote sensing change detection aims to identify land-cover changes from bi-temporal images. Most existing methods follow a one-shot dense prediction paradigm, directly regressing a change mask from fused features. However, such approaches overlook the intrinsic frequency characteristics of change patterns. We propose Freq-RemoteVAR, a frequency autoregressive framework that reformulates change detection as a structured generation problem in the frequency domain. Instead of predicting the change mask in a single step, we introduce a next-frequency prediction paradigm, where change information is progressively generated from coarse to fine. We design a frequency-aware mask tokenization strategy that decomposes change supervision into multi-frequency token targets via Fourier transformation and quantization. We develop a Frequency VAR Transformer, which performs causal autoregressive modeling over frequency tokens. The model starts from learned mask queries and progressively predicts frequency-level tokens conditioned on previously generated tokens and bi-temporal image features, effectively capturing long-range dependencies across frequency scales. We introduce Scale-Aligned RoPE Cross Attention (SRCA) module, which aligns frequency-domain mask queries with spatial-domain bi-temporal features under a unified coordinate system, enhancing spatial-frequency consistency during generation. We propose a Change-quality Control module that adaptively modulates the generation process through dynamic normalization, attention biasing, and spatial offset adjustment, thereby suppressing pseudo-change responses and improving robustness. Extensive experiments on CDD, GZ-CD, and LEVIR-CD demonstrate that Freq-RemoteVAR consistently outperforms existing methods, particularly in challenging scenarios with complex appearance variations and noisy disturbances.
Chinese Translation
遥感变化检测旨在从双时相图像中识别土地覆盖变化。现有的大多数方法遵循一次性密集预测范式，直接从融合特征回归变化掩膜。然而，这些方法忽视了变化模式的内在频率特征。我们提出了Freq-RemoteVAR，一种频率自回归框架，将变化检测重新表述为频域中的结构生成问题。我们引入下一频率预测范式，而不是在单一步骤中预测变化掩膜，在此过程中变化信息从粗到细逐步生成。我们设计了一种频率感知掩膜标记策略，通过傅里叶变换和量化将变化监督分解为多频率标记目标。我们开发了一个频率VAR Transformer，对频率标记进行因果自回归建模。该模型从学习到的掩膜查询开始，逐步预测基于先前生成的标记和双时相图像特征的频率级标记，有效捕捉跨频率尺度的长程依赖关系。我们引入了尺度对齐的RoPE交叉注意力（SRCA）模块，在统一坐标系统下将频域掩膜查询与空间域双时相特征对齐，从而增强生成过程中的空间频率一致性。我们提出了一种变化质量控制模块，通过动态归一化、注意力偏置和空间偏移调整自适应调节生成过程，从而抑制伪变化响应并提高鲁棒性。在CDD、GZ-CD和LEVIR-CD上的大量实验表明，Freq-RemoteVAR在复杂外观变化和噪声干扰的挑战性场景中，始终优于现有方法。
cs.CV / 70 / 2607.25818
SepPrune:A Separator-based Pruning Framework for Efficient Multimodal Large Language Models
SepPrune：一种基于分隔符的修剪框架，用于高效的多模态大型语言模型
Abstract
Recent multimodal large language models (MLLMs), such as Qwen2.5-VL and InternVL3, generate large numbers of vision tokens for high-resolution inputs, leading to substantial computational cost. Existing vision token pruning methods either depend on cross-modal attention and cannot prune before the prefill stage, or rely on diversity estimation with high computational overhead. We observe that attention scores from both vision and text tokens peak at modality separator tokens, suggesting that these separators bridge the two modalities. Based on this observation, we propose SepPrune, an efficient, training-free, plug-and-play pruning method that uses the separator token as a unified query to rank and select informative vision tokens. SepPrune reuses the LLM's built-in projection parameters and requires no architectural changes. Experiments on Qwen2.5-VL-7B show that SepPrune achieves state-of-the-art performance, retaining 96.3% of the original accuracy while removing 80.2% of vision tokens.
Chinese Translation
最近的多模态大型语言模型（MLLMs），如 Qwen2.5-VL 和 InternVL3，为高分辨率输入生成大量视觉标记，导致了显著的计算成本。现有的视觉标记修剪方法要么依赖于跨模态注意力，无法在预填充阶段之前进行修剪，要么依赖于具有高计算开销的多样性估计。我们观察到，视觉和文本标记的注意力得分在模态分隔符标记处达到峰值，这表明这些分隔符在两种模态之间架起了桥梁。基于这一观察，我们提出了 SepPrune，一种高效的、无训练的、即插即用的修剪方法，利用分隔符标记作为统一查询来排名和选择信息丰富的视觉标记。SepPrune 重用 LLM 的内置投影参数，且无需架构更改。在 Qwen2.5-VL-7B 上的实验表明，SepPrune 实现了最先进的性能，保留了 96.3% 的原始准确性，同时去除了 80.2% 的视觉标记。
cs.CV / 71 / 2607.25820
Food Image Segmentation with LLM-Derived Ingredient Labels and Multimodal Fusion
基于大语言模型衍生成分标签和多模态融合的食品图像分割
Abstract
Food image segmentation plays a vital role in health-related applications such as nutrition tracking and personalized health monitoring. However, existing models often underperform on visually similar ingredients and rare food categories. To address this issue, we propose two plug-and-play multimodal modules that enhance the segmentation performance by leveraging ingredient labels inferred from food images using large language models (LLMs). The first module, called LIM-F (Language Injection Module for Features), is designed to pair with any image encoder that produces multi-layer outputs (e.g., Swin Transformer), while the second module, LIM-Q (Language Injection Module for Queries), targets Mask2Former-style Transformer-based decoders. Both modules enable training without the need for pre-aligning images with text by directly injecting semantic ingredient information into the visual analysis pipeline. On the FoodSeg103 benchmark, the proposed method achieves state-of-the-art performance. Specifically, integrating LIM-Q into the Mask2Former decoder with a Swin-L image encoder yields a mean Intersection over Union (mIoU) of 55.0. LIM-F also demonstrates strong generalization and competitive performance, reaching an mIoU of 54.4 under the same model (Swin-L+Mask2Former). Furthermore, its applicability extends beyond Transformer-based decoders, as evidenced by an improvement from 47.7 to 49.8 mIoU when integrated into a CNN-based architecture. Notably, the improved segmentation accuracy is achieved with only a moderate (at most 3.8 GB) increase in the GPU memory consumption during training. Thus, the proposed approach offers a practical and scalable solution for fine-grained food understanding.
Chinese Translation
食品图像分割在营养追踪和个性化健康监测等健康相关应用中发挥着至关重要的作用。然而，现有模型在视觉上相似的成分和稀有食品类别上往往表现不佳。为了解决这一问题，我们提出了两个即插即用的多模态模块，通过利用大语言模型（LLMs）从食品图像中推断的成分标签来增强分割性能。第一个模块称为LIM-F（特征语言注入模块），旨在与任何生成多层输出的图像编码器配对（例如，Swin Transformer），而第二个模块LIM-Q（查询语言注入模块）则针对Mask2Former风格的基于Transformer的解码器。这两个模块允许在训练过程中无需将图像与文本预对齐，而是通过直接将语义成分信息注入视觉分析流程来实现。在FoodSeg103基准测试中，所提出的方法达到了最先进的性能。具体而言，将LIM-Q集成到使用Swin-L图像编码器的Mask2Former解码器中，获得了55.0的平均交并比（mIoU）。LIM-F也展示了强大的泛化能力和竞争性能，在相同模型（Swin-L+Mask2Former）下达到了54.4的mIoU。此外，其适用性超越了基于Transformer的解码器，集成到基于CNN的架构中时，mIoU从47.7提高到49.8。值得注意的是，在训练过程中，改进的分割准确性仅需适度增加（最多3.8 GB）的GPU内存消耗。因此，所提出的方法为细粒度食品理解提供了一个实用且可扩展的解决方案。
cs.CV / 72 / 2607.25830
Beyond Static Costs: Learning-Dynamics Aware Loss Functions for Long-Tailed Classification
超越静态成本：面向学习动态的长尾分类损失函数
Abstract
Deep learning models in computer vision face significant challenges when trained on long-tailed datasets, where a few majority classes dominate while many minority classes are severely underrepresented. Such imbalances frequently arise in real-world scenarios such as rare species recognition, manufacturing fault detection, and medical image understanding, leading to biased models that underperform on tail classes. Existing reweighting methods typically rely on static class frequencies to penalize the model, ignoring the dynamic nature of how effectively a network actually learns a class over time. We address this by introducing a novel Learning-Dynamics Aware Loss (LDAL) function that shifts the focus from static sample counts to dynamic learning progress. LDAL framework adjusts class weights continuously by leveraging: (i) the strength of learned feature representations (semantic scale), (ii) the intrinsic learning difficulty of each class, measured via the Shannon entropy of its predictions, and (iii) an inter-epoch regularizer term that tracks prediction shifts between consecutive epochs to stabilize training and avoid local minima. LDAL is purely a objective function which incurs negligible computational overhead while adapting to the feature learning of the model. Experimental results on multiple benchmark datasets demonstrate that our approach significantly surpasses state-of-the-art reweighting loss functions, providing an optimal trade-off between accuracy and generalizability. The source code is available at https://github.com/sdm2026/ldal
Chinese Translation
计算机视觉中的深度学习模型在长尾数据集上训练时面临重大挑战，其中少数主导类占主导地位，而许多少数类则严重不足。这种不平衡在现实场景中经常出现，例如稀有物种识别、制造缺陷检测和医学图像理解，导致模型偏向于主导类而在尾部类上表现不佳。现有的重加权方法通常依赖于静态类频率来惩罚模型，忽视了网络在时间上学习某一类的动态特性。我们通过引入一种新颖的学习动态感知损失（Learning-Dynamics Aware Loss, LDAL）函数来解决这一问题，该函数将重点从静态样本计数转向动态学习进展。LDAL框架通过利用以下三点持续调整类权重：(i) 学习到的特征表示的强度（语义尺度），(ii) 每个类的内在学习难度，通过其预测的香农熵进行测量，以及 (iii) 一个跨周期正则化项，该项跟踪连续周期之间的预测变化，以稳定训练并避免局部最小值。LDAL纯粹是一个目标函数，其计算开销微乎其微，同时适应模型的特征学习。在多个基准数据集上的实验结果表明，我们的方法显著超越了最先进的重加权损失函数，在准确性和泛化能力之间提供了最佳的权衡。源代码可在 https://github.com/sdm2026/ldal 获取。
cs.CV / 73 / 2607.25842
Adversarial Deepfake Generation and an Investigation of Purification-Based Adversarial Detection
对抗性深伪生成及基于净化的对抗检测研究
Abstract
This paper describes the participation of team "Go To Germany" in the ImageCLEF 2026 Deepfake Detection and Generation Task. For the image generation task, we employ FLUX.1-dev with PuLID for identity-preserving face synthesis, combined with a multi-model PGD adversarial attack targeting 12 detectors simultaneously (DiffJPEG-in-loop, MI/DI/EoT, adaptive weighting, two-stage warm-start). Our approach achieved 90% evasion against organizer detectors and 57.6% against participant detectors, with a final generation score of 0.4170. For the image detection task, we combine two complementary detectors - SigLIP+DINOv2 for AI-generated images and GenD-DINOv3 for face manipulations - in a max-probability ensemble, achieving 99.4% accuracy on baseline deepfakes but suffering from high false-positive rates on real images, resulting in a final detection score of 0.6986. Beyond the official submission, we conducted a self-initiated investigation of purification-based adversarial detection, comparing three families of detection signals across six detectors that share a CLIP ViT-L/14 backbone. We find that raw $|\Delta \text{logit}|$ under median-3 purification, applied through the EFFORT detector, separates adversarial inputs from clean inputs with AUROC 0.81-0.98 across four adversarial source types - a finding that refutes the simple backbone-preservation hypothesis and exposes a sharp JPEG-quality cliff at Q70 where the signal collapses.
Chinese Translation
本文描述了团队“Go To Germany”在ImageCLEF 2026深伪检测与生成任务中的参与。在图像生成任务中，我们采用了FLUX.1-dev与PuLID相结合进行身份保留的人脸合成，并结合了针对12个检测器的多模型PGD对抗攻击（DiffJPEG-in-loop, MI/DI/EoT, 自适应加权，双阶段热启动）。我们的方法在组织者检测器上实现了90%的规避率，在参与者检测器上实现了57.6%的规避率，最终生成得分为0.4170。在图像检测任务中，我们将两个互补的检测器结合起来——SigLIP+DINOv2用于AI生成图像，GenD-DINOv3用于人脸操控，采用最大概率集成，在基线深伪上实现了99.4%的准确率，但在真实图像上出现了高假阳性率，最终检测得分为0.6986。除了官方提交外，我们还进行了自发的基于净化的对抗检测研究，比较了六个共享CLIP ViT-L/14骨干网的检测信号的三种家族。我们发现，在通过EFFORT检测器应用的中位数-3净化下，原始$| ext{logit}|$能够在四种对抗源类型中以AUROC 0.81-0.98有效区分对抗输入与干净输入，这一发现反驳了简单的骨干网保留假设，并揭示了在Q70处信号崩溃的明显JPEG质量悬崖。
cs.CV / 74 / 2607.25860
Open-Ended CT Volume Segmentation with Weak Supervision from Language
基于语言的弱监督下开放式CT体积分割
Abstract
We introduce a method for training a text-conditioned segmentation model for CT scans, which combines voxel-level supervision with coarse but scalable slice-level supervision from reports. We extract, from a large database of scan-report pairs, descriptions of findings with indices of slices where those findings occur. We then finetune a general-purpose 2D image segmentation model, SAM3, with standard segmentation losses from strongly labeled data and with a slice-level classification loss from the extracted weak supervision. Our results on the ReXGroundingCT dataset illustrate that this strategy improves the segmentation dice score: from an 8% relative gain when there are 1000 fully labeled volumes to 22% when there are 250 fully labeled volumes.
Chinese Translation
我们提出了一种用于训练文本条件下CT扫描分割模型的方法，该方法结合了体素级监督与来自报告的粗略但可扩展的切片级监督。我们从一个大型扫描-报告对数据库中提取了发现的描述，并标注了这些发现出现的切片索引。然后，我们利用来自强标签数据的标准分割损失和来自提取的弱监督的切片级分类损失，对通用的2D图像分割模型SAM3进行微调。我们在ReXGroundingCT数据集上的结果表明，这一策略提高了分割的Dice系数：在1000个完全标注的体积时相对提升8%，而在250个完全标注的体积时提升达到22%。
cs.CV / 75 / 2607.25894
TIGA: Trajectory-Injected Generative Attack against Black-box AIGC Detectors
TIGA：针对黑箱AIGC检测器的轨迹注入生成攻击
Abstract
Recent diffusion models have achieved remarkable realism in facial image synthesis, posing growing challenges to artificial intelligence-generated content (AIGC) forensic detectors.Existing evasion methods typically perturb pre-generated images or require detector-aware training, which may introduce visible or statistical artifacts and limit applicability when the diffusion model must remain frozen and the target detector is accessible only through black-box queries. We propose Trajectory-Injected Generative Attack (TIGA), a source-image-free and training free framework that generates detector-evasive images within a single diffusion sampling trajectory. TIGA steers the latent Denoising Diffusion Implicit Model (DDIM) trajectory so that adversarial properties emerge during generation rather than being added afterward. TIGA first aggregates gradients from multiple white-box surrogate detectors to form a transferable, sign-aware prior, and then performs anisotropic directional search with symmetric finite-difference queries to estimate the black-box target response. The estimated directions are stabilized by decayed momentum and injected according to the DDIM noise schedule, with frequency-domain reshaping to suppress high frequency artifacts. Experiments on surrogate and unseen specialized forensic detectors show that TIGA achieves strong blackbox attack performance, transferability, and high robustness under common post-processing operations without source images or diffusion-model retraining, while preserving high perceptual quality.
Chinese Translation
近期的扩散模型在面部图像合成方面取得了显著的真实感，这对人工智能生成内容（AIGC）取证检测器提出了日益严峻的挑战。现有的规避方法通常会扰动预生成的图像或需要检测器感知的训练，这可能会引入可见或统计伪影，并在扩散模型必须保持不变且目标检测器仅通过黑箱查询可访问时限制适用性。我们提出了轨迹注入生成攻击（TIGA），这是一种无源图像和无训练的框架，能够在单一的扩散采样轨迹中生成规避检测器的图像。TIGA引导潜在的去噪扩散隐式模型（DDIM）轨迹，使得对抗特性在生成过程中出现，而不是事后添加。TIGA首先从多个白箱替代检测器聚合梯度，以形成可转移的、符号感知的先验，然后通过对称有限差分查询进行各向异性方向搜索，以估计黑箱目标响应。估计的方向通过衰减动量进行稳定，并根据DDIM噪声调度进行注入，同时进行频域重塑以抑制高频伪影。在替代和未见的专业取证检测器上的实验表明，TIGA在没有源图像或扩散模型再训练的情况下，实现了强大的黑箱攻击性能、可转移性以及在常见后处理操作下的高鲁棒性，同时保持了高感知质量。
cs.CV / 76 / 2607.25921
Evaluating VLMs for Autonomous Agent-Driven Geometry Clipping Detection in Video Game QA
评估视觉语言模型在自主代理驱动的视频游戏质量保证中的几何剪切检测
Abstract
In this work, we study the use of Vision-Language Models (VLMs) for anomaly detection in an agent-driven game Quality Assurance (QA) pipeline focusing on geometry clipping. In this evaluation, a custom exploration agent navigates a game level to collect visual observations, while the automatic annotation pipeline provides frame-level clipping labels. This setup allows us to evaluate recent VLMs on a controlled anomaly detection task without manual annotation. We benchmark six recent VLMs (Gemini, GPT, Qwen, Gemma, Llama, and Ministral) under a zero-shot prompting setting and analyse their sensitivity to four prompt variants. Our results show that while the VLMs can capture visual cues associated with geometry clipping, they all produce substantial false positives on visually ambiguous frames such as near-contact geometry and partial occlusions. Gemini-3.1-Flash achieves the best overall accuracy and is the most robust to prompt variation, while open-source models exhibit large precision--recall swings depending on the prompt design. These findings suggest that current VLMs are best suited as high-recall candidate filters within multi-stage QA pipelines rather than as standalone bug detectors.
Chinese Translation
在本研究中，我们探讨了视觉语言模型（VLMs）在代理驱动的游戏质量保证（QA）流程中用于异常检测的应用，重点关注几何剪切。该评估中，一个定制的探索代理在游戏关卡中导航以收集视觉观察，同时自动注释流程提供帧级剪切标签。这一设置使我们能够在没有人工注释的情况下评估近期的VLMs在受控异常检测任务中的表现。我们在零-shot 提示设置下对六个近期的VLMs（Gemini、GPT、Qwen、Gemma、Llama 和 Ministral）进行了基准测试，并分析了它们对四种提示变体的敏感性。我们的结果表明，尽管VLMs能够捕捉与几何剪切相关的视觉线索，但它们在视觉模糊的帧（如近接几何和部分遮挡）上均产生了显著的误报。Gemini-3.1-Flash在整体准确性上表现最佳，并且对提示变体最为稳健，而开源模型则根据提示设计表现出较大的精确度-召回波动。这些发现表明，当前的VLMs更适合作为多阶段QA流程中的高召回候选过滤器，而非独立的缺陷检测器。
cs.CV / 77 / 2607.25926
Face De-Identification: A Domain-Centric Survey from Capture to Processing
面部去标识化：从捕获到处理的领域中心调查
Abstract
Face de-identification (De-ID) aims to remove or conceal personally identifiable facial features in images or videos to prevent identity recognition while preserving utility for downstream tasks. With the rising emphasis on data privacy and responsible AI, face De-ID has emerged as an active research area spanning computer vision and privacy-preserving communities. Early approaches, and many contemporary ones, operate in the digital domain by modifying pixel-level or appearance-level features through post-capture processing. Recent advances extend face De-ID beyond post-processing by integrating privacy mechanisms directly into sensors during image acquisition, bridging sensing systems and downstream vision algorithms. In parallel, physical-domain methods explore wearable accessories and materials that conceal identity information in real-world environments prior to capture. In this survey, we present the first unified overview that spans the full data acquisition pipeline, encompassing the physical, sensor, and digital domains. Through this domain-centric lens, we systematically analyze current methodologies, technical progress, and the distinct challenges inherent to each stage. We then review and organize existing evaluation protocols, examining current practices and highlighting the critical need for standardized, comprehensive benchmarks. Finally, we identify key open problems and outline emerging research directions to guide future work in this rapidly evolving field. To support ongoing research, we maintain a project page that organizes relevant literature with collected datasets and open source code: https://github.com/CV-AC/Awesome-FaceDe-ID.
Chinese Translation
面部去标识化（De-ID）旨在去除或隐匿图像或视频中可识别的面部特征，以防止身份识别，同时为下游任务保留效用。随着对数据隐私和负责任人工智能的日益重视，面部去标识化已成为一个活跃的研究领域，涵盖计算机视觉和隐私保护社区。早期的方法以及许多当代方法通过后捕获处理在数字领域内操作，修改像素级或外观级特征。近期的进展将面部去标识化扩展到后处理之外，通过在图像采集过程中直接将隐私机制集成到传感器中，架起了传感系统与下游视觉算法之间的桥梁。同时，物理领域的方法探索可穿戴配件和材料，这些配件和材料在捕获之前能够隐匿现实环境中的身份信息。在本调查中，我们首次提供了一个涵盖整个数据采集管道的统一概述，包括物理、传感器和数字领域。通过这一领域中心的视角，我们系统地分析了当前的方法论、技术进展以及每个阶段固有的独特挑战。然后，我们回顾并整理了现有的评估协议，检查当前的实践，并强调了标准化、全面基准的迫切需求。最后，我们识别出关键的开放问题，并概述了新兴的研究方向，以指导这一快速发展的领域的未来工作。为了支持持续的研究，我们维护了一个项目页面，组织相关文献、收集的数据集和开源代码： https://github.com/CV-AC/Awesome-FaceDe-ID.
cs.CV / 78 / 2607.25948
MODUS: Decoder-Only Any-to-Any Modeling of Diverse Modalities
MODUS：仅解码器的任意对任意多模态建模
Abstract
Any-to-any models predict any modality from any combination of others within a single network, a formulation used in multimodal vision and vision-language models, and increasingly in scientific domains such as ecology and astronomy. Existing any-to-any models are typically trained from scratch using encoder-decoder or diffusion architectures, impacting their performance and preventing them from using strong pre-trained decoder-only models as a prior. In this work, we investigate decoder-only any-to-any multimodal modeling, which treats all modalities symmetrically and supports arbitrary modalities as inputs and outputs without modality-specific heads, losses, or task pipelines. Because every modality is both an input and an output of the same model, the resulting model, named Modus, can support a range of applications, such as chained generation through intermediate modalities or cross-modal self-verification by scoring the model's own outputs with another generated modality. Modus demonstrates strong out-of-the-box performance and is competitive with specialist and multitask baselines using a single model across various benchmarks. All materials are open-sourced at https://modus-multimodal.epfl.ch/.
Chinese Translation
任意对任意模型在单一网络中从其他任意组合预测任意模态，这一形式在多模态视觉和视觉-语言模型中得到应用，并在生态学和天文学等科学领域中日益受到关注。现有的任意对任意模型通常从头开始训练，采用编码器-解码器或扩散架构，这影响了它们的性能，并阻止它们利用强大的预训练仅解码器模型作为先验。在本研究中，我们探讨了仅解码器的任意对任意多模态建模，该方法对所有模态进行对称处理，支持任意模态作为输入和输出，而无需特定于模态的头部、损失或任务管道。由于每个模态既是同一模型的输入也是输出，因此所得到的模型命名为Modus，可以支持一系列应用，例如通过中间模态进行链式生成或通过用另一种生成模态对模型自身输出进行评分来进行跨模态自我验证。Modus展示了强大的开箱即用性能，并在各种基准测试中与专业和多任务基线模型具有竞争力，且仅使用单一模型。所有材料均已开源，网址为 https://modus-multimodal.epfl.ch/.
cs.CV / 79 / 2607.25961
Knowledge-Guided Multimodal Reasoning over Interacting Streams for Video-Level Ambivalence and Hesitancy Recognition
基于知识引导的多模态推理框架用于视频级的矛盾与犹豫识别
Abstract
Ambivalence and hesitancy (A/H) are conflicting affective states that precede the delay or abandonment of health behaviour change. Recognition of A/H at the video level is difficult, since the signal arises from disagreement across and within facial, vocal, linguistic, and bodily modalities, and manifests differently across individuals. The proposed PRISM-AH (Predictive Reasoning over Interacting Streams for Multimodal Ambivalence/Hesitancy Recognition), is a framework that treats A/H as a multimodal conflict that unfolds over time. Frozen vision, audio, and text encoders are aligned into short time windows and passed to a lightweight streaming model that scores cross-modal dissonance, predicts each next window to expose a hesitation surprise signal, discovers behaviour prototypes, and is conditioned on participant metadata. Dense window-level annotations supervise the model as an auxiliary objective, and the decision threshold is calibrated for macro F1. A knowledge-guided large language model then reasons over structured evidence using the expert cue taxonomy of the dataset, and its verdict is fused late only when validation performance improves. On the labelled public test partition of 525 videos, PRISM-AH attains a macro F1 of 0.6133, compared to the reported zero-shot baseline of 0.2827. The reasoning gain is validated to transfer from validation to the larger test partition.
Chinese Translation
矛盾与犹豫（A/H）是导致健康行为改变延迟或放弃的相互冲突的情感状态。在视频级别上识别A/H是困难的，因为信号源于面部、声音、语言和身体模态之间及内部的不一致，并且在不同个体中表现不同。我们提出的PRISM-AH（基于交互流的多模态矛盾/犹豫识别的预测推理框架）将A/H视为随时间展开的多模态冲突。冻结的视觉、音频和文本编码器被对齐到短时间窗口，并传递给一个轻量级流模型，该模型对跨模态的不和谐进行评分，预测每个下一个窗口以揭示犹豫的惊讶信号，发现行为原型，并基于参与者的元数据进行条件化。密集的窗口级注释作为辅助目标监督模型，并且决策阈值经过宏F1的校准。然后，基于知识引导的大型语言模型利用数据集的专家提示分类法对结构化证据进行推理，只有在验证性能改善时，其裁决才会进行后期融合。在525个视频的标注公共测试分区上，PRISM-AH达到了0.6133的宏F1，相比之下，报告的零-shot基线为0.2827。推理增益被验证能够从验证转移到更大的测试分区。
cs.CV / 80 / 2607.25962
LaP-Forensics: Latent-Pixel Consistency Guided Multimodal Reasoning for Deepfake Detection
LaP-Forensics：基于潜在像素一致性的多模态推理用于深伪检测
Abstract
Recent generative models can produce images with few obvious visual artifacts, weakening detectors and explanations that rely only on surface appearance. We present LaP-Forensics, a multimodal framework that augments RGB semantics with reconstruction-based forensic evidence. A frozen Stable Diffusion DDIM inversion-reconstruction model provides a fixed reconstruction reference, and its residual map measures local compatibility with that reference. Independent projectors encode the RGB image and residual map before a structured Where-What-Why model predicts a textual analysis and an artifact mask.Supervised fine-tuning is followed by Group Relative Policy Optimization (GRPO), whose reward combines mask overlap with output-structure and evidence-reference terms. These text-side terms encourage the model to refer to the consistency map but do not constitute a verifier of free-form textual truth. A separate image-level head fuses RGB and DDIM-residual class features. Experiments show cross-generator detection on UniversalFakeDetect and competitive artifact localization on the official SynthScars benchmark. Controlled cue-construction, inversion-horizon, component, reward-term, and counterfactual analyses support the utility of the residual stream under the evaluated settings, while free-form textual faithfulness and reliability under post-processing remain open limitations.
Chinese Translation
近期的生成模型能够生成几乎没有明显视觉伪影的图像，这削弱了仅依赖表面外观的检测器和解释。我们提出了LaP-Forensics，这是一个多模态框架，通过重建基础的取证证据增强RGB语义。一个冻结的Stable Diffusion DDIM反演-重建模型提供了固定的重建参考，其残差图测量与该参考的局部兼容性。独立的投影器对RGB图像和残差图进行编码，然后一个结构化的Where-What-Why模型预测文本分析和伪影掩码。经过监督微调后，采用了群体相对策略优化（GRPO），其奖励结合了掩码重叠与输出结构和证据参考项。这些文本侧的项鼓励模型参考一致性图，但并不构成自由形式文本真相的验证器。一个单独的图像级头部融合了RGB和DDIM残差类别特征。实验表明在UniversalFakeDetect上进行跨生成器检测，并在官方SynthScars基准上进行竞争性的伪影定位。受控的线索构建、反演视野、组件、奖励项和反事实分析支持了在评估设置下残差流的有效性，而后处理下自由形式文本的真实性和可靠性仍然是开放的局限性。
cs.CV / 81 / 2607.25967
Quasi-SVD: Learning a Lie-constrained matrix factorisation for real-time imaging
准奇异值分解（Quasi-SVD）：一种用于实时成像的李约束矩阵分解学习
Abstract
Singular Value Decomposition (SVD) underlies matrix factorisation tasks across computational imaging, with medical applications increasingly demanding real-time processing. Yet SVD algorithms are inherently sequential, constraining real-time GPU throughput and limit online deployment in clinical pipelines. This study introduces Quasi-SVD, a differentiable, fully parallelized matrix factorization framework for GPUs. Rather than enforcing orthogonality on both factors, it guarantees exact orthogonality for a single Lie-parameterized factor while recovering the remaining components through soft constraints, enabling efficient parallel decomposition without iterative singular-vector optimization. This asymmetric design, provably sufficient for valid factorisation, achieves reconstruction fidelity of SSIM = 0.89-0.94 and accelerates computation by 3-20x relative to cuSOLVER and randomised SVD, enabling throughput above 25 FPS. Performance is evaluated on two medical imaging tasks spanning complementary computational regimes: (1) spatio-temporal background subtraction for ultrasound localisation microscopy, requiring high-dimensional matrix separation, and (2) Mueller matrix polarimetry for neurosurgical tissue characterisation, requiring massive batch processing of small matrices. Across both regimes and multiple imaging instruments, the proposed framework demonstrates robust domain transfer and throughput exceeding 25 FPS at clinical matrix scales, a rate sufficient for live image-guided workflows that classical solvers cannot currently support in these settings. By prioritising downstream reconstruction fidelity over exact spectral recovery, Quasi-SVD makes structured matrix factorisation practical for real-time imaging.
Chinese Translation
奇异值分解（SVD）是计算成像中矩阵分解任务的基础，医学应用日益需要实时处理。然而，SVD算法本质上是顺序的，这限制了实时GPU的吞吐量，并限制了在临床流程中的在线部署。本研究提出了准奇异值分解（Quasi-SVD），这是一个可微分的、完全并行化的GPU矩阵分解框架。该方法并不强制要求两个因子正交，而是保证一个李参数化因子的精确正交，同时通过软约束恢复其余组件，从而实现高效的并行分解，无需迭代的奇异向量优化。这种不对称设计在理论上足以实现有效的分解，重建保真度达到SSIM = 0.89-0.94，并相较于cuSOLVER和随机SVD加速计算3-20倍，使得吞吐量超过25帧每秒（FPS）。在两个涵盖互补计算模式的医学成像任务上评估性能：（1）超声定位显微镜的时空背景减除，要求高维矩阵分离；（2）用于神经外科组织特征化的穆勒矩阵偏振测量，要求对小矩阵的大规模批处理。在这两种模式和多种成像仪器中，所提出的框架展示了强大的领域转移能力，且在临床矩阵规模下的吞吐量超过25 FPS，这一速率足以支持实时图像引导工作流程，而传统求解器在这些场景中无法满足需求。通过优先考虑下游重建保真度而非精确的谱恢复，准奇异值分解（Quasi-SVD）使结构化矩阵分解在实时成像中变得可行。
cs.CV / 82 / 2607.25990
On the Use of Synthetic Data for Threshold Calibration in Face Recognition: Performance and Security Implications for Border Control Systems
合成数据在面部识别阈值校准中的应用：对边境控制系统的性能和安全影响
Abstract
The recently deployed Entry/Exit System (EES) introduces large-scale biometric verification into European border control, requiring face recognition systems to operate at extremely low false match rates (FMR). While regulatory frameworks define performance targets at the EES Central System level, they do not specify how verification thresholds should be calibrated in practice at the Member State level. In operational settings, obtaining representative real-world data for calibration is often constrained by legal, logistical, and privacy limitations. In this work, we investigate the use of synthetic face data for threshold calibration in document-to-live verification scenarios relevant to border control systems. We analyze the alignment of genuine and impostor score distributions between synthetic and real datasets and evaluate the transferability of calibrated thresholds across domains, with a focus on low-FMR operating points. Our results show that synthetic data can approximate calibration behavior in controlled settings, but fails to reliably generalize to unconstrained conditions due to mismatches in score distribution tails. These discrepancies lead to significant degradation in recognition performance and increased vulnerability to morph-based attacks. We further demonstrate that calibration outcomes are highly dataset-dependent, even across synthetic datasets. Overall, our findings highlight that while synthetic data is useful for system development and preliminary calibration, our results indicate that reliable threshold selection in high-security deployments typically requires validation and adjustment using representative real-world data.
Chinese Translation
最近部署的入境/出境系统（Entry/Exit System, EES）在欧洲边境控制中引入了大规模生物识别验证，要求面部识别系统在极低的误匹配率（False Match Rate, FMR）下运行。虽然监管框架在EES中央系统层面定义了性能目标，但并未具体说明成员国层面应如何实际校准验证阈值。在操作环境中，出于法律、后勤和隐私限制，获取具有代表性的真实世界数据进行校准往往受到限制。在本研究中，我们探讨了在与边境控制系统相关的文件与实时验证场景中使用合成面部数据进行阈值校准。我们分析了合成数据与真实数据集中真实样本和冒名样本得分分布的一致性，并评估了在不同领域之间校准阈值的可转移性，特别关注低FMR操作点。我们的结果表明，合成数据可以在受控环境中近似校准行为，但由于得分分布尾部的不匹配，未能可靠地推广到不受限制的条件。这些差异导致识别性能显著下降，并增加了对基于形态攻击的脆弱性。我们进一步证明，校准结果高度依赖于数据集，即使在合成数据集之间也是如此。总体而言，我们的研究结果强调，尽管合成数据对系统开发和初步校准是有用的，但我们的结果表明，在高安全性部署中，可靠的阈值选择通常需要使用具有代表性的真实世界数据进行验证和调整。
cs.CV / 83 / 2607.25993
Beyond Zooming: Learning Multi-Tool Visual Reasoning for Ultra-High-Resolution Remote Sensing
超越缩放：学习超高分辨率遥感的多工具视觉推理
Abstract
Ultra-high-resolution (UHR) remote-sensing (RS) imagery provides fine-grained Earth-observation evidence over city-scale scenes, but poses a fundamental challenge for multimodal large language models (MLLMs): task-relevant evidence is often sparse, local, and spatially dispersed across extremely large visual contexts. A natural solution is to equip MLLMs with zoom-in tools for active local inspection. However, through a pilot study on XLRS-Bench, we find that zoom-in is only partially effective: it resolves easy and medium-level tasks with locally recoverable evidence, but saturates on hard cases requiring global search, multi-region comparison, path planning, or dispersed-evidence reasoning. Motivated by this finding, we move beyond single-tool zoom-in and introduce GeoMTVR, a large-scale Geospatial Multi-Tool Visual Reasoning dataset built from wide-area satellite imagery. GeoMTVR contains 13K UHR VQA samples with interleaved reasoning trajectories, diverse visual tool calls, and returned visual observations, enabling models to learn question decomposition, tool selection, regional inspection, object-level grounding, auxiliary visual reasoning, and cross-tool evidence integration. Beyond supervised fine-tuning, we propose a tool-attention-focused reinforcement learning algorithm that concentrates optimization on critical tool-use decisions, including when to invoke tools, which tool to select, where to apply it, and how to interpret tool outputs. By combining SFT on GeoMTVR with our RL algorithm, we develop GeoLens, a multi-tool visual reasoning MLLM for UHR RS. Experiments show that GeoLens consistently outperforms direct reasoning and single-tool zoom-in baselines, achieving stronger accuracy, better evidence grounding, and more efficient tool-use trajectories.
Chinese Translation
超高分辨率（UHR）遥感（RS）影像为城市规模场景提供了细粒度的地球观测证据，但对多模态大型语言模型（MLLMs）提出了根本性挑战：与任务相关的证据通常是稀疏的、局部的，并且在极大的视觉上下文中空间分散。一个自然的解决方案是为MLLMs配备缩放工具以进行主动的局部检查。然而，通过对XLRS-Bench的初步研究，我们发现缩放工具的效果仅部分有效：它能够解决具有局部可恢复证据的简单和中等难度任务，但在需要全局搜索、多区域比较、路径规划或分散证据推理的困难案例中则饱和。基于这一发现，我们超越了单一工具的缩放，提出了GeoMTVR，这是一个基于广域卫星影像构建的大规模地理空间多工具视觉推理数据集。GeoMTVR包含13K个UHR视觉问答（VQA）样本，具有交错的推理轨迹、多样的视觉工具调用和返回的视觉观察，能够使模型学习问题分解、工具选择、区域检查、对象级定位、辅助视觉推理和跨工具证据整合。除了监督微调外，我们还提出了一种以工具注意力为中心的强化学习算法，专注于关键工具使用决策的优化，包括何时调用工具、选择哪个工具、在哪里应用以及如何解释工具输出。通过在GeoMTVR上结合监督微调和我们的强化学习算法，我们开发了GeoLens，一个用于UHR RS的多工具视觉推理MLLM。实验表明，GeoLens在准确性、证据定位和工具使用轨迹的效率上始终优于直接推理和单工具缩放基线。
cs.CV / 84 / 2607.26004
Parallel Decoding Distillation for Fast Image and Video Generation
快速图像和视频生成的并行解码蒸馏
Abstract
Generation in video diffusion or flow models is computationally expensive due to the slow and iterative sampling process. Current state-of-the-art (SOTA) acceleration methods heavily rely on variational score distillation (VSD) and adversarial losses to distill diffusion models into few-step generators. Albeit achieving high-quality video generation, these training losses are notoriously hard to optimize and suffer from mode collapse, leading to loss of video diversity and lack of motion. In this paper, we introduce Parallel Decoding Distillation (PDD), a simplified and scalable trajectory-based distillation method for fast inference of diffusion and flow matching models. Our architecture and training procedure are compatible with any pre-trained model and support sampling with a varying number of function evaluations (NFE). PDD accelerates generation by predicting multiple denoising steps per network evaluation. Conceptually, it learns a representation of the mean velocity without regressing its derivative using JVPs or finite-difference approximations. Our method achieves SOTA performance with 4-8 NFE on LTX-2.3 Text-to-Video/Audio, Wan 14B Text-to-Video, and Qwen-Image Text-to-Image. Moreover, PDD presents a significant improvement in generated video diversity.
Chinese Translation
视频扩散或流模型中的生成过程计算开销大，主要是由于缓慢且迭代的采样过程。目前的最先进加速方法（SOTA）严重依赖变分评分蒸馏（VSD）和对抗损失，将扩散模型蒸馏为少步生成器。尽管实现了高质量的视频生成，这些训练损失 notoriously 难以优化，并且容易出现模式崩溃，导致视频多样性的丧失和运动的缺乏。在本文中，我们引入了并行解码蒸馏（PDD），这是一种简化且可扩展的基于轨迹的蒸馏方法，用于快速推理扩散和流匹配模型。我们的架构和训练过程与任何预训练模型兼容，并支持以不同数量的函数评估（NFE）进行采样。PDD 通过每次网络评估预测多个去噪步骤来加速生成。从概念上讲，它学习平均速度的表示，而不使用 JVPs 或有限差分近似来回归其导数。我们的方法在 LTX-2.3 文本到视频/音频、Wan 14B 文本到视频和 Qwen-Image 文本到图像上以 4-8 NFE 达到了 SOTA 性能。此外，PDD 在生成视频多样性方面显著提高。
cs.CV / 85 / 2607.26005
Pictura: Perspective-View Self-Play at Scale for Driving
Pictura：大规模视角自我对弈驱动
Abstract
Self-play in simulation produces robust driving policies at scale. Demonstrations of such behavior have been made using privileged vectorized observations such as exact poses and velocities, even for occluded agents. This assumes that perception is solved and introduces a representation gap with the partial observation of a deployed agent driving from the perspective view of egocentric cameras. A common fix, distilling the privileged policy into a camera-input student, leaves the student imitating decisions its own view cannot justify. Instead, we establish perspective-view self-play as a practical training regime. We introduce Pictura, a GPU-accelerated multi-agent driving simulator that renders each agent's egocentric view at every step, mitigating the representation gap at its source. Pictura sustains up to 500K agent-steps/s (2M images/s) on a single H100. Using Pictura, we train Alberti by self-play with plain PPO. It is the first large-scale driving self-play policy trained directly from perspective images, without privileged observations. Training spans 50B agent steps for ~35M km of driving. It approaches the driving performance of its privileged vectorized counterpart, and transfers zero-shot to Waymo Open Motion Dataset layouts re-rendered in Pictura, where it outperforms privileged vectorized agents. Project page: https://valeoai.github.io/Pictura/
Chinese Translation
在仿真中进行自我对弈能够大规模产生稳健的驾驶策略。此类行为的演示使用了特权向量化观测，例如精确的姿态和速度，即使对于被遮挡的代理也适用。这假设感知问题已解决，并引入了与从自我中心相机的视角观察的部署代理部分观测之间的表示差距。一种常见的解决方案是将特权策略提炼为相机输入的学生策略，但这使得学生模仿其自身视角无法证明的决策。相反，我们确立了视角自我对弈作为一种实用的训练机制。我们引入了Pictura，这是一款GPU加速的多代理驾驶模拟器，在每一步渲染每个代理的自我中心视图，从源头上减轻了表示差距。Pictura在单个H100上支持高达500K代理步/秒（2M图像/秒）。使用Pictura，我们通过普通的PPO进行自我对弈训练Alberti。这是首个直接从视角图像中训练的大规模驾驶自我对弈策略，而不依赖特权观测。训练跨越50B代理步，约35M公里的驾驶。其驾驶性能接近特权向量化对应物，并在Pictura中重新渲染的Waymo Open Motion Dataset布局中实现了零样本迁移，超越了特权向量化代理。项目页面：https://valeoai.github.io/Pictura/
cs.CV / 86 / 2607.26037
Wonder: Video World Model Done Better
Wonder：更优秀的视频世界模型
Abstract
We present Wonder, a general-purpose video world model for real-time, camera-controllable world exploration. Given an image or a conditional video, Wonder constructs a playable world where users can navigate interactively by moving the camera, discovering unseen regions, and revisiting previously observed areas in real time and over a long-term horizon. Achieving this capability requires a system-level co-design of control method, memory mechanism, and training strategy. We introduce a novel camera conditioning with a dense coordinate field whose renderings provide spatially aligned motion and orientation cues, allowing the model to interpret camera motion directly as visual evidence. To support fast and precise memory retrieval over a growing generation context, we propose an efficient sparse attention-based memory mechanism, enabling the model to selectively attend to a small set of relevant context tokens at inference time, regardless of actual context length. We further develop several techniques to rectify the self-forcing-style distillation pipeline, improving the student model's ability to respect control signals, as well as maintaining diverse generation modes and long-term memory from the teacher. Together, these components enable Wonder to synthesize diverse, minute-scale videos at 16 FPS while preserving coherent geometry, appearance, and dynamics across long rollouts. Beyond image-to-video generation, Wonder naturally supports video-conditioned generation, allowing existing dynamic scenes to be re-shot in real time.
Chinese Translation
我们提出了Wonder，一个通用的视频世界模型，用于实时、可通过相机控制的世界探索。给定一幅图像或一个条件视频，Wonder构建了一个可玩世界，用户可以通过移动相机进行交互式导航，实时发现未见区域，并重新访问之前观察过的区域，且可在长期范围内进行探索。实现这一能力需要对控制方法、记忆机制和训练策略进行系统级的共同设计。我们引入了一种新颖的相机条件化方法，结合密集坐标场，其渲染提供了空间对齐的运动和方向线索，使模型能够直接将相机运动解释为视觉证据。为了支持在不断增长的生成上下文中快速而精确的记忆检索，我们提出了一种高效的基于稀疏注意力的记忆机制，使模型能够在推理时选择性地关注一小组相关的上下文标记，而不受实际上下文长度的限制。我们进一步开发了几种技术来修正自我强制风格的蒸馏流程，提高学生模型对控制信号的尊重能力，同时保持来自教师的多样化生成模式和长期记忆。所有这些组件共同使Wonder能够以16帧每秒的速度合成多样化的微观视频，同时在长时间的生成过程中保持一致的几何形状、外观和动态。除了图像到视频的生成，Wonder还自然支持视频条件生成，允许现有动态场景实时重新拍摄。
cs.CV / 87 / 2607.26042
VetClaw: An Edge-Cloud Multimodal Agentic System for Veterinary Disease Screening
VetClaw：一种用于兽医疾病筛查的边缘-云多模态智能系统
Abstract
We present VetClaw, an edge-cloud multimodal agentic system for early veterinary disease screening. VetClaw uses a camera module as an edge sensing device and sends captured images, together with optional symptom descriptions, to a server-hosted vision-language model for zero-shot disease classification. The system separates agent interaction from workflow orchestration: OpenClaw provides scheduling, tool access, user interaction, and notification services on the edge device, while LangGraph manages the stateful screening workflow, including input validation, image transmission, model invocation, safety checks, conditional routing, failure handling, and structured logging. This design moves beyond static image classification by enabling the system to collect visual evidence, invoke external models, apply deterministic safety rules, and generate diagnostic-support alerts. Results show that image-only VLM prediction remains limited, whereas symptom-guided and multimodal inputs improve zero-shot classification performance. Thus, VetClaw transforms a static prediction model into a coordinated, safety-aware system that can use tools, manage workflows, handle failures, and escalate uncertain cases.
Chinese Translation
我们提出了VetClaw，一种用于早期兽医疾病筛查的边缘-云多模态智能系统。VetClaw使用摄像头模块作为边缘传感设备，并将捕获的图像及可选的症状描述发送至服务器托管的视觉-语言模型进行零样本疾病分类。该系统将代理交互与工作流编排分离：OpenClaw在边缘设备上提供调度、工具访问、用户交互和通知服务，而LangGraph管理有状态的筛查工作流，包括输入验证、图像传输、模型调用、安全检查、条件路由、故障处理和结构化日志记录。该设计超越了静态图像分类，使系统能够收集视觉证据、调用外部模型、应用确定性安全规则并生成诊断支持警报。结果表明，仅使用图像的视觉-语言模型预测仍然有限，而症状引导和多模态输入则改善了零样本分类性能。因此，VetClaw将静态预测模型转变为一个协调的、具有安全意识的系统，能够使用工具、管理工作流、处理故障并升级不确定案例。
cs.AI / 1 / 2607.24758
Do Models Fake Alignment Without Clear Consequences?
模型是否在没有明确后果的情况下伪装对齐？
Abstract
Large language models are capable of recognizing evaluation contexts and altering their behavior to reflect evaluator expectations rather than typical deployment behaviors, a phenomenon known as alignment faking. The reasons why models fake alignment are not fully understood, however. Canonical examples of alignment faking have taken place in scenarios that explicitly connect evaluation to consequences for the model, such as retraining the model or delaying its deployment. However, recent work by Sheshadri et al. has suggested that mechanistic motivations for alignment faking may vary across models and be more complex than previously considered. To investigate whether consequence-linking information is necessary for alignment faking, we placed 15 models in a scenario testing their willingness to violate a corporate network access policy to help a user with a pro-social request. Nine models were found to produce significant compliance gaps, 5 of which persisted with the removal of scenario language relating model evaluations to deployment consequences. We additionally tested the effect of goal language on model preferences, finding it drove violations in some while suppressing violations in others. This suggests that alignment faking may not require as much instrumental scaffolding as was previously believed, and monitored behavior may be a poor indicator of how agents may behave in deployment.
Chinese Translation
大型语言模型能够识别评估上下文，并改变其行为以反映评估者的期望，而不是典型的部署行为，这种现象被称为对齐伪装。然而，模型伪装对齐的原因尚未完全理解。对齐伪装的经典例子发生在那些明确将评估与模型后果联系起来的场景中，例如重新训练模型或延迟其部署。然而，Sheshadri 等人的最新研究表明，模型伪装对齐的机械动机可能因模型而异，并比之前认为的更复杂。为了调查后果关联信息是否对对齐伪装是必要的，我们将15个模型置于一个测试场景中，测试它们是否愿意违反企业网络访问政策以帮助用户满足亲社会请求。发现有9个模型产生了显著的合规差距，其中5个在移除与部署后果相关的场景语言后仍然存在。我们还测试了目标语言对模型偏好的影响，发现它在某些模型中驱动了违规行为，而在其他模型中则抑制了违规行为。这表明，对齐伪装可能不需要如之前认为的那样多的工具支撑，而监测行为可能是代理在部署中行为的一个不良指标。
cs.AI / 2 / 2607.24759
Beyond Memory: A Templated Substrate for Heterogeneous Collaborative Knowledge Work with LLM Agents
超越记忆：一种用于异构协作知识工作的模板化基底，适用于大型语言模型代理
Abstract
Research projects, educational efforts, and adjacent knowledge work accumulate findings, decisions, and reasoning that future collaborators rarely recover. The parts most useful to that work, including dead ends and walked-back claims, are routinely excluded from publications and shared code; future researchers re-attempt the same failures because no record survives. LLM coding agents are common participants but hold no persistent memory across sessions, and retrieval-augmented generation over raw sources does not compound. The llm-wiki pattern (Karpathy, 2026; tonbi, 2026) addresses this by inserting an LLM-maintained, interlinked wiki between raw sources and the agent. We present llm-wiki-memory-template, a reusable, agent-aware instantiation, and argue it is a substrate for heterogeneous collaborative knowledge work along three axes (multi-human, multi-AI-agent, multi-domain) with each axis supported by a distinct architectural element of the template ({\S}4). The wiki is append-only by convention, which preserves what did not work alongside what did, addressing a negative-result loss problem that publications and code-sharing structurally cannot solve. Three deployed case studies and one design report cover the axes individually: a solo research lineage that preserves abandoned iterations; a two-author project whose retroactive audit revised two prior experiments' claimed 20-of-20 coverage down to 14 and 12 evidence-based answers, then to 18 and 18 after a fix, with the failure path preserved across the artifact; an in-progress multi-agent deployment reported as a design; and a cross-domain educational variant. We name failure-path preservation, agent honesty, and appropriation as cross-cutting sociotechnical properties of the artifact, not only of its technical mechanisms.
Chinese Translation
研究项目、教育工作和相关知识工作积累了发现、决策和推理，但未来的合作者很少能恢复这些内容。对该工作的最有用部分，包括无果的尝试和被撤回的主张，通常在出版物和共享代码中被排除；未来的研究者因没有记录而重新尝试相同的失败。大型语言模型（LLM）编码代理是常见参与者，但在会话之间没有持久的记忆，而基于原始来源的检索增强生成并不累积。llm-wiki模式（Karpathy, 2026; tonbi, 2026）通过在原始来源和代理之间插入一个由LLM维护的相互链接的维基来解决这个问题。我们提出了llm-wiki-memory-template，这是一种可重用的、关注代理的实例，并认为它是异构协作知识工作的基底，涵盖三个维度（多个人类、多AI代理、多领域），每个维度由模板的一个独特架构元素支持（{ extbackslash}S4）。该维基按惯例为追加式，这保留了未成功的尝试与成功的尝试，解决了出版物和代码共享在结构上无法解决的负结果损失问题。三个已部署的案例研究和一个设计报告分别涵盖了这些维度：一个保留被放弃迭代的单人研究谱系；一个两位作者的项目，其追溯审计将之前实验声称的20个覆盖率修订为14和12个基于证据的答案，经过修正后又变为18和18，同时保留了失败路径；一个正在进行的多代理部署报告作为设计；以及一个跨领域的教育变体。我们将失败路径的保留、代理的诚实性和适用性命名为该工件的跨切面社会技术特性，而不仅仅是其技术机制。
cs.AI / 3 / 2607.24762
Kernel Forge: An Agent Harness for LLM-based Generation and Optimization of CUDA Kernels
内核锻造：基于大语言模型的CUDA内核生成与优化的智能代理框架
Abstract
Machine learning models are increasingly embedded in everyday software, and most of their runtime is spent in a small set of compute kernels such as matrix multiplication, convolution, and normalization. Optimizing these kernels is one of the most direct ways to reduce latency and cost, but it has traditionally required expert engineers to hand-write low-level GPU code. Agentic systems built on large language models (LLMs) can now generate and optimize kernels with far less human effort, yet existing tools are largely evaluated on randomly generated tensors and isolated kernels, emit standalone CUDA code that developers must manually reintegrate, mostly target only LLM PyTorch models, and offer limited support for inspecting and debugging results. We present Kernel Forge, an open-source, end-to-end agentic harness that accepts any unmodified PyTorch model in place. Kernel Forge supports vision, diffusion, and LLM workloads, uses Monte Carlo Tree Search (MCTS) to explore multiple optimization paths rather than a single linear refinement chain, and ships with a graphical user interface for monitoring progress, inspecting candidate kernels, and debugging failures. We evaluate Kernel Forge on four PyTorch models spanning vision, diffusion, and LLM workloads on an NVIDIA DGX Spark with GB10 GPU. With only 50 optimization iterations per kernel, it optimizes 14 kernels to outperform PyTorch eager mode, reaching $1.52\times$ on adaptive\_avgpool2d in ResNet-50, $1.70\times$ on group\_norm in Stable Diffusion 3.5 Medium, $2.83\times$ on softmax in Gemma 4 E2B, and $1.54\times$ on softmax in Qwen 3.5 35B-A3B.
Chinese Translation
机器学习模型越来越多地嵌入到日常软件中，而它们的大部分运行时间都花费在一小部分计算内核上，例如矩阵乘法、卷积和归一化。优化这些内核是减少延迟和成本的最直接方式之一，但传统上需要专家工程师手动编写低级GPU代码。基于大语言模型（LLMs）构建的智能系统现在可以以更少的人力生成和优化内核，然而现有工具主要在随机生成的张量和孤立内核上进行评估，生成的CUDA代码是独立的，开发者必须手动重新集成，且大多数仅针对LLM PyTorch模型，并且在检查和调试结果方面提供的支持有限。我们提出了Kernel Forge，一个开源的端到端智能代理框架，能够直接接受任何未修改的PyTorch模型。Kernel Forge支持视觉、扩散和LLM工作负载，使用蒙特卡洛树搜索（MCTS）探索多个优化路径，而不是单一的线性优化链，并配备了图形用户界面，用于监控进度、检查候选内核和调试故障。我们在一台配备GB10 GPU的NVIDIA DGX Spark上评估了Kernel Forge，涉及四个涵盖视觉、扩散和LLM工作负载的PyTorch模型。在每个内核仅进行50次优化迭代的情况下，它优化了14个内核，超越了PyTorch的急切模式，在ResNet-50的adaptive_avgpool2d上达到$1.52 imes$，在Stable Diffusion 3.5 Medium的group_norm上达到$1.70 imes$，在Gemma 4 E2B的softmax上达到$2.83 imes$，以及在Qwen 3.5 35B-A3B的softmax上达到$1.54 imes$。
cs.AI / 4 / 2607.24763
CaRE Compute-aware Remasking Evaluation Protocol for Masked Diffusion Language Models
CaRE 计算感知重掩码评估协议用于掩码扩散语言模型
Abstract
Masked diffusion language models (MDLMs) are advancing rapidly, yet the evaluation standards needed to reliably interpret their progress have not kept pace. Despite MDLMs becoming competitive with autoregressive language models, seven recent remasking papers evaluate under incompatible settings, varying nominal step counts, metrics, and sampling temperatures without jointly controlling these factors, rendering their strategy rankings largely incomparable and leaving open whether reported gains reflect algorithmic improvements or evaluation artifacts. We present CaRE, a compute-aware evaluation framework that audits MDLM remasking strategies by standardizing actual number of function evaluations (NFE), enforcing multi-metric reporting, and explicitly controlling stochasticity. Applied to 7 remasking strategies across LLaDA-8B-Base and Dream-7B-Base at 4 stochasticity levels and 3 step budgets on OpenWebText and LM1B, CaRE reveals that: (i) temperature explains the majority of MAUVE variance, (ii) compute-matched comparisons reverse several published strategy rankings, and (iii) informed remasking and stochastic unmasking are in tension, with high-entropy remasking reducing MAUVE by 0.296 at 256 steps at unmask_temp=0.25 (p=0.020). A CaRE leaderboard covering 12 open-weight MDLMs (150M to 8B parameters) shows that this interaction direction holds across architectures and scales. These findings demonstrate that current MDLM evaluations can systematically conflate algorithmic improvements with hidden choices of compute and stochasticity. We release the evaluation protocol, implementation, and leaderboard to ensure future remasking claims are reproducible and comparable.
Chinese Translation
掩码扩散语言模型（MDLMs）正在迅速发展，但可靠解读其进展所需的评估标准并未同步提升。尽管MDLMs已与自回归语言模型竞争，但最近的七篇重掩码论文在不兼容的设置下进行评估，变化的名义步数、指标和采样温度未能共同控制这些因素，导致其策略排名大部分不可比，并且尚不清楚报告的增益是否反映了算法改进或评估伪影。我们提出了CaRE，一个计算感知的评估框架，通过标准化实际函数评估次数（NFE）、强制多指标报告和明确控制随机性来审计MDLM重掩码策略。应用于LLaDA-8B-Base和Dream-7B-Base的7种重掩码策略，在OpenWebText和LM1B上以4个随机性水平和3个步数预算进行评估，CaRE揭示了：(i) 温度解释了MAUVE方差的大部分，(ii) 计算匹配的比较反转了几项已发布的策略排名，(iii) 信息重掩码与随机去掩码之间存在张力，高熵重掩码在unmask_temp=0.25时以256步减少MAUVE 0.296（p=0.020）。一个涵盖12个开放权重MDLM（150M到8B参数）的CaRE排行榜显示，这种交互方向在不同架构和规模中保持一致。这些发现表明，当前的MDLM评估可能系统性地将算法改进与隐藏的计算和随机性选择混淆。我们发布了评估协议、实现和排行榜，以确保未来的重掩码声明可重复和可比较。
cs.AI / 5 / 2607.24764
GrocLM: Grocery Category Recommendation in E-Commerce with Large Language Models
GrocLM：基于大型语言模型的电子商务杂货类别推荐
Abstract
The rapid growth of online grocery shopping requires recommendation systems that capture cyclical purchasing behavior and diverse user intents. Traditional item-level methods face scalability and accuracy challenges, motivating category-level recommendation as a more structured and practical alternative. We present GROCLM, a fine-tuned language model for grocery category recommendation in a real-world production environment. GROCLM employs a two-stage LoRA-based training strategy to encode cyclical purchasing patterns directly into model parameters, enabling more effective utilization of rebuying signals compared to prompt-based conditioning. To ensure valid and controllable outputs, we further introduce a trie-based constrained decoding mechanism over a predefined category space. Experiments on both proprietary production data and a public benchmark demonstrate that GROCLM consistently outperforms strong baselines. In a live production restocking task, GROCLM achieves a 7.5% relative improvement in cart-adds per impression, while maintaining efficient inference by generating all categories jointly. These results highlight the effectiveness and practicality of integrating large language models into structured recommendation systems.
Chinese Translation
在线杂货购物的快速增长需要能够捕捉周期性购买行为和多样化用户意图的推荐系统。传统的商品级方法面临可扩展性和准确性挑战，这促使我们考虑类别级推荐作为一种更结构化和实用的替代方案。我们提出了GROCLM，一种针对真实生产环境中杂货类别推荐的微调语言模型。GROCLM采用基于LoRA的两阶段训练策略，将周期性购买模式直接编码到模型参数中，相较于基于提示的条件化方法，更有效地利用了重复购买信号。为了确保输出的有效性和可控性，我们进一步引入了一种基于前缀树的约束解码机制，覆盖预定义的类别空间。在专有生产数据和公共基准上的实验表明，GROCLM始终优于强基线。在实时生产补货任务中，GROCLM在每次展示的购物车添加量上实现了7.5%的相对提升，同时通过联合生成所有类别保持了高效的推理。这些结果突显了将大型语言模型整合到结构化推荐系统中的有效性和实用性。
cs.AI / 6 / 2607.24766
Crystalis: Progressive Nucleation and Semantic Annealing for Coordinated Multi-View Visualization Generation
Crystalis：用于协调多视图可视化生成的渐进成核与语义退火
Abstract
Large language models (LLMs) can generate individual charts, but coordinated multi-view visualizations (CMVs), where views share data flows and cross-view interactions, remain out of reach. Tight field-level coupling among data transformations, visual encodings, and interaction coordinations causes errors in one component to silently invalidate others. Rather than pursuing end-to-end analytical quality, which depends on model capability, domain knowledge, and user expertise, we target a foundational question: can LLMs reliably produce structurally correct CMVs, and what abstractions make this possible? We present Crystalis, a framework built on query-centric CMV modeling that decomposes a CMV into structured queries over a dependency graph spanning three component types (Data, Visualization, Interaction) and three abstraction levels (requirement, specification, executable object). Two complementary mechanisms operate over this structure: progressive nucleation crystallizes each query vertically from requirement to object along the dependency order, while semantic annealing enforces horizontal consistency across queries at each level through layered logical checks. On a 12-task benchmark across five frontier LLMs, Crystalis achieves up to 75% end-to-end success, substantially outperforming an agentic coding baseline (8.3% E2E with the same foundation model), and a user study with 12 practitioners confirms the usability of the decomposition and iterative refinement workflow.
Chinese Translation
大型语言模型（LLMs）可以生成单个图表，但协调多视图可视化（CMVs）——其中视图共享数据流和跨视图交互——仍然难以实现。数据转换、视觉编码和交互协调之间的紧密领域级耦合导致一个组件中的错误会无声地使其他组件失效。我们并不追求依赖于模型能力、领域知识和用户专业知识的端到端分析质量，而是关注一个基础性问题：LLMs能否可靠地生成结构上正确的CMVs，以及哪些抽象使这一目标成为可能？我们提出了Crystalis，一个基于查询中心的CMV建模框架，它将CMV分解为跨越三种组件类型（数据、可视化、交互）和三个抽象层次（需求、规范、可执行对象）的结构化查询。两种互补机制在这一结构上运作：渐进成核从需求到对象沿依赖顺序垂直地细化每个查询，而语义退火通过分层逻辑检查在每个层次上强制查询之间的水平一致性。在针对五个前沿LLM的12个任务基准测试中，Crystalis实现了高达75%的端到端成功率，显著优于一个基于代理编码的基线（同一基础模型下的8.3% E2E），而一项针对12名从业者的用户研究证实了分解和迭代精炼工作流的可用性。
cs.AI / 7 / 2607.24768
PATHFinder Agent for Tailored Prenatal Care
个性化产前护理的PATHFinder代理
Abstract
Prenatal care is an important preventive service designed to improve outcomes for pregnant individuals. The American College of Obstetricians and Gynecologists (ACOG) recently introduced guidelines advocating tailored prenatal care, called PATH (Plan for Tailored Healthcare). We present PATHFinder Agent(Planner for Appropriate Tailored Healthcare), an end-to-end conversational agentic system that gathers patient health and social context through structured dialogue, curates individualized prenatal care plans aligned with PATH guidelines, and surfaces community resources from Michigan 211. The system features a four-stage workflow spanning patient intake, dynamic interaction, plan synthesis, and clinician oversight. We evaluate frontier large language models (LLMs) on expert-curated rubrics across five clinical dimensions, finding that GPT-5.2 achieves the highest average score (77.6\%) while identifying key gaps in antenatal testing recommendations. We discuss future validation through human participant studies and randomized controlled trials.
Chinese Translation
产前护理是一项重要的预防性服务，旨在改善孕妇的健康结果。美国妇产科医师学会（ACOG）最近推出了倡导个性化产前护理的指南，称为PATH（个性化医疗计划）。我们提出了PATHFinder代理（适当个性化医疗规划者），这是一个端到端的对话代理系统，通过结构化对话收集患者的健康和社会背景，策划与PATH指南一致的个性化产前护理计划，并提供来自密歇根211的社区资源。该系统具有四个阶段的工作流程，涵盖患者接收、动态互动、计划综合和临床监督。我们在五个临床维度上对前沿的大型语言模型（LLMs）进行了专家评估，发现GPT-5.2获得了最高的平均分（77.6%），同时识别出产前检测建议中的关键缺口。我们讨论了通过人类参与者研究和随机对照试验进行未来验证的可能性。
cs.AI / 8 / 2607.24769
LLM Scheming Inversely Scales with Pretraining Language Coverage
LLM 策划与预训练语言覆盖率呈反比关系
Abstract
With the growing capabilities of frontier models, AI alignment becomes increasingly critical in high-risk deployment settings. While recent work has empirically demonstrated in-context scheming -- the covert pursuit of misaligned objectives while feigning alignment -- in frontier language models, most work has been performed exclusively in English, leaving a major gap in multilingual safety. We apply Petri, an open-source automated auditing framework, to Qwen3-30B-A3B to evaluate deceptive and scheming behaviors across multiple languages. Our findings suggest that scheming scores are inversely correlated with the estimated pretraining language coverage, with low-resource languages averaging 34.2\% higher scores compared to high-resource languages on a five-category scheming index. Furthermore, we find that the effect of estimated pretraining language coverage is not uniform across scheming behaviors.
Chinese Translation
随着前沿模型能力的不断提升，人工智能的对齐问题在高风险部署环境中变得愈加重要。尽管近期的研究已经在前沿语言模型中实证展示了上下文策划——即在假装对齐的同时秘密追求不对齐目标——但大多数研究仅限于英语，导致多语言安全性存在重大缺口。我们将开源自动审计框架 Petri 应用到 Qwen3-30B-A3B，以评估多种语言中的欺骗和策划行为。我们的研究结果表明，策划得分与估计的预训练语言覆盖率呈反比关系，低资源语言的得分平均比高资源语言高出 34.2\%，在五类策划指数中表现尤为明显。此外，我们发现估计的预训练语言覆盖率对策划行为的影响并不均匀。
cs.AI / 9 / 2607.24770
ProcAgent: An Agentic Framework for Procedural Task Guidance on Edge with Human-in-the-Loop
ProcAgent：一种基于边缘的程序任务指导的代理框架，结合人机协作
Abstract
Procedural tasks such as furniture assembly and home repair impose substantial cognitive demands because users must interpret instructions, track task progress, reason about spatial state, and recover from errors while performing physical actions. Prior multimodal assistants have shown promise for procedural guidance, but most rely on cloud inference and fixed always-on perception, making them poorly suited to privacy-sensitive, latency-critical domestic settings. We present ProcAgent, a fully on-device, agentic, vision-based procedural assistant for real-time adaptive guidances on a single NVIDIA Jetson AGX Orin. ProcAgent uses a propose-and-verify architecture that combines low-latency continuous perception, a symbolic task graph, on-demand vision-language verification, and an LLM-based interaction agent. The system continuously proposes user progress, invokes expensive visual reasoning only when ambiguity or likely deviation arises, and supports both reactive question answering and proactive intervention with human-in-the- loop confirmation. We evaluate ProcAgent along four dimensions: perception accuracy, reasoning, task-level performance, and user experience. Despite running entirely on-device, the system maintains responsive interaction, resolving text-only queries in approximately 2 seconds and visually grounded queries in approximately 8 seconds. In a user study with 10 participants completing assembly tasks, ProcAgent receives positive ratings for comprehensibility, actionability, and privacy comfort. These results show that adaptive procedural assistance can be achieved entirely on edge hardware without sacrificing usability.
Chinese Translation
程序性任务，如家具组装和家庭修理，给用户带来了相当大的认知负担，因为用户必须解读说明、跟踪任务进度、推理空间状态，并在执行物理动作时从错误中恢复。之前的多模态助手在程序指导方面显示出潜力，但大多数依赖于云推理和固定的持续感知，使其不适合隐私敏感和延迟关键的家庭环境。我们提出了ProcAgent，这是一种完全在设备上运行的、基于视觉的程序助手，能够在单个NVIDIA Jetson AGX Orin上提供实时自适应指导。ProcAgent采用了一种提议与验证的架构，结合了低延迟的连续感知、符号任务图、按需的视觉-语言验证和基于大型语言模型（LLM）的交互代理。该系统持续提出用户进度，仅在出现歧义或可能偏离时调用昂贵的视觉推理，并支持人机协作确认下的反应式问答和主动干预。我们从感知准确性、推理、任务级性能和用户体验四个维度评估ProcAgent。尽管完全在设备上运行，该系统仍保持响应性，文本查询的响应时间约为2秒，视觉基础查询的响应时间约为8秒。在一项有10名参与者完成组装任务的用户研究中，ProcAgent在可理解性、可操作性和隐私舒适度方面获得了积极评价。这些结果表明，自适应程序辅助可以完全在边缘硬件上实现，而不牺牲可用性。
cs.AI / 10 / 2607.24771
RoCo-ACE: Rollout-Conditioned Online Distillation for Retention-Aware Knowledge Injection
RoCo-ACE：基于回滚条件的在线蒸馏用于关注保留的知识注入
Abstract
Knowledge injection updates pretrained MLLMs with new factual or domain-specific knowledge, but fitting full authoritative answers can cause drift in non-updated behavior. Online distillation mitigates this drift by training on model-generated rollouts, yet uniform reference-conditioned distillation provides coarse supervision: it can under-emphasize reference-supported rollout tokens and supervise omitted facts only indirectly. We introduce RoCo-ACE, a rollout-conditioned online distillation objective for knowledge injection. RoCo uses same-rollout reference-free/reference-conditioned likelihood contrast to reallocate additional distillation weight to reference-supported rollout tokens, while ACE adds sparse reference-side anchored correction for authoritative anchors omitted from the rollout without full-answer imitation. Across three knowledge-injection settings, six retention benchmarks, multiple baselines, and multiple base models, RoCo-ACE achieves the best injected-knowledge accuracy among compared methods while keeping evaluated retention close to the base model.
Chinese Translation
知识注入通过更新预训练的多语言大模型（MLLMs）以融入新的事实或特定领域知识，但完全拟合权威答案可能导致未更新行为的漂移。在线蒸馏通过对模型生成的回滚进行训练来减轻这种漂移，然而，统一的参考条件蒸馏提供的监督较为粗糙：它可能会低估参考支持的回滚标记，并且仅间接监督被省略的事实。我们提出了RoCo-ACE，一种用于知识注入的基于回滚条件的在线蒸馏目标。RoCo使用相同回滚的无参考/有参考条件的似然对比，将额外的蒸馏权重重新分配给参考支持的回滚标记，而ACE则为未在回滚中完全模仿的权威锚点添加稀疏的参考侧锚定修正。在三个知识注入设置、六个保留基准、多个基线和多个基础模型中，RoCo-ACE在比较方法中实现了最佳的注入知识准确性，同时保持评估的保留接近基础模型。
cs.AI / 11 / 2607.24772
RSMeM: Knowledge-Enhanced Memory Evolution for Remote Sensing Agents with Systematic Evaluation
RSMeM：针对遥感智能体的知识增强记忆演化及系统评估
Abstract
Geoscience research requires complex analysis and domain expertise, with remote sensing (RS) observations as a key foundation. However, existing RS agents built on general-purpose LLMs remain largely domain-agnostic, resulting in brittle and error-prone workflows. Moreover, these failures are seldom consolidated into a reusable experience for subsequent analyses. To address this issue, we introduce RSMeM, a knowledge-enhanced memory evolution mechanism that bootstraps RS agents with pre-distilled domain knowledge and iteratively integrates online experience for robust multi-step tool execution. RSMeM is composed of two components: (i) Hierarchical Knowledge Grounding, which performs taxonomy-aware retrieval over a hierarchical domain corpus to guide planning and tool selection; and (ii) Failure-Aware Experience Refinement, which distills failure-annotated tool-use traces into reusable constraints for next-round tool execution. By iteratively employing these two processes, RS agents can evolve to absorb task-level domain knowledge and effectively translate it into instance-level execution experience. Extensive experiments on EarthBench demonstrate that RSMeM consistently improves tool-use performance and end-to-end answer across a diverse set of LLM backbones. Notably, RSMeM achieves a 6% accuracy improvement on DeepSeek-V3.2 with less than 1% additional experience tokens, demonstrating the strong knowledge density of our distilled experience. Our code is available at https://github.com/AI9Stars/RSMeM
Chinese Translation
地球科学研究需要复杂的分析和领域专业知识，而遥感（RS）观测是其关键基础。然而，现有基于通用大语言模型（LLM）的遥感智能体在很大程度上仍然是领域无关的，导致工作流程脆弱且容易出错。此外，这些失败很少被整合为可重用的经验以供后续分析。为了解决这一问题，我们提出了RSMeM，一种知识增强的记忆演化机制，通过预先提炼的领域知识引导遥感智能体，并迭代整合在线经验以实现稳健的多步骤工具执行。RSMeM由两个组成部分构成：（i）层次知识基础，执行对层次领域语料库的分类感知检索，以指导规划和工具选择；（ii）失败感知经验精炼，将标注失败的工具使用轨迹提炼为下一轮工具执行的可重用约束。通过迭代使用这两个过程，遥感智能体能够演化以吸收任务级领域知识，并有效地将其转化为实例级执行经验。在EarthBench上的大量实验表明，RSMeM在多种LLM基础上始终提高了工具使用性能和端到端答案的准确性。值得注意的是，RSMeM在DeepSeek-V3.2上实现了6%的准确率提升，且额外经验令牌不足1%，显示出我们提炼经验的强知识密度。我们的代码可在 https://github.com/AI9Stars/RSMeM 获取。
cs.AI / 12 / 2607.24773
Right-sizing Recommendations (RSR): Cloud Workload Conformal Prediction for Virtual Machines in Data Center Operations
适当规模建议（RSR）：数据中心运营中虚拟机的云工作负载保守预测
Abstract
Managing cloud infrastructure efficiently, especially in environments of large cloud providers or hyperscalers, requires optimizing the use of physical resources to minimize costs and maximize performance. Selecting the right virtual machine (VM) sizes is crucial to achieving cost efficiency in these dynamic environments. However, traditional VM allocation and scheduling approaches often fail to account for the fluctuating and unpredictable nature of VM utilization, leading to inefficiencies such as over- or under-provisioning of resources. High-quality interval prediction helps accurately capture uncertainty in cloud resource demand and supports cloud operators in efficient instance provisioning. As an effective and reliable framework for constructing prediction intervals (PIs), conformal prediction (CP) is used for mid- and long-term forecasting tasks in cloud computing environments. This study proposes a new data-driven PI construction approach using bootstrapping conformal prediction for modern, dynamic, data-driven Right-sizing Recommendations (RSR) to enhance provisioning for diverse application workloads on hyperscalers. By learning workload utilization patterns, identifying correlations across multiple time series, and predicting medium- to long-term utilization trends, this research seeks to improve the efficiency of cloud and data center operations through an AI/ML-based provisioning pipeline. Our study demonstrates that AI-driven models, powered by machine learning regression techniques and evaluated using backtesting, achieve promising forecasting results for cloud resource utilization. Additionally, we rank the selected models to identify top-performing approaches for long-life VM candidates. The proposed framework enhances right-sizing recommendations and supports more cost-effective resource allocation in dynamic cloud environments.
Chinese Translation
高效管理云基础设施，尤其是在大型云服务提供商或超大规模环境中，需要优化物理资源的使用，以最小化成本并最大化性能。选择合适的虚拟机（VM）规模对于在这些动态环境中实现成本效率至关重要。然而，传统的虚拟机分配和调度方法往往未能考虑到虚拟机利用率的波动和不可预测性，导致资源的过度或不足配置等低效现象。高质量的区间预测有助于准确捕捉云资源需求中的不确定性，并支持云运营商高效地配置实例。作为构建预测区间（PIs）的有效且可靠的框架，保守预测（CP）被用于云计算环境中的中长期预测任务。本研究提出了一种新的数据驱动的预测区间构建方法，利用自助法保守预测，针对现代动态数据驱动的适当规模建议（RSR），以增强对超大规模环境中多样化应用工作负载的配置。通过学习工作负载利用模式、识别多个时间序列之间的相关性以及预测中长期利用趋势，本研究旨在通过基于人工智能/机器学习的配置管道提高云和数据中心运营的效率。我们的研究表明，基于机器学习回归技术的人工智能驱动模型，通过回测评估，能够实现云资源利用的良好预测结果。此外，我们对所选模型进行排名，以识别适合长期虚拟机候选的最佳表现方法。所提出的框架增强了适当规模建议，并支持在动态云环境中更具成本效益的资源分配。
cs.AI / 13 / 2607.24774
Atmospheric Diffusion-Guided Spatio-Temporal Transformer for Nuclear Radiation Forecasting
基于大气扩散引导的时空变换器用于核辐射预测
Abstract
Nuclear radiation, the energy released during atomic decay, poses persistent risks to public health and the environment, and concerns have only grown since the Fukushima accident and the recent commencement of treated-water discharge. Modern monitoring networks now record radiation levels and accompanying weather conditions at thousands of stations, opening the door to nationwide forecasting that can inform emergency response, agricultural advisories, and routine public-safety decisions. However, turning this abundance of monitoring data into reliable forecasts is difficult for three reasons. First, the time series at each station are highly non-stationary, shaped by radioactive decay, weather variability, and irregular human interventions. Second, monitoring stations are severely unevenly distributed in space. Roughly 78% of Japan's stations sit in less than 6% of the country, clustered near Fukushima, which breaks the assumptions of standard graph-based models. Third, radiation co-evolves with heterogeneous context such as wind, temperature, and humidity through atmospheric transport processes that purely data-driven models struggle to capture from observations alone. In this study, we introduce NRFormer+, a spatio-temporal Transformer for nationwide nuclear radiation forecasting. NRFormer+ couples non-stationary temporal attention and density-adaptive spatial attention with a new atmospheric diffusion module that estimates how meteorology drives radiation dispersion and injects this physical signal into the network as an architectural prior. NRFormer+ delivers state-of-the-art accuracy on both datasets across all 13 baselines, reducing sudden-change MAE by up to 19.1% over the strongest baseline at comparable inference latency. Our code and datasets are publicly available at https://github.com/tfeilyu/NRFormer_Plus.
Chinese Translation
核辐射是原子衰变过程中释放的能量，对公众健康和环境构成持续风险，自福岛事故以来以及近期处理水排放的开始，这一担忧愈发加剧。现代监测网络在数千个站点记录辐射水平和伴随的天气条件，为全国范围的预测提供了可能，这可以为应急响应、农业建议和日常公共安全决策提供信息。然而，将这些丰富的监测数据转化为可靠的预测面临三个主要挑战。首先，各个站点的时间序列高度非平稳，受放射性衰变、天气变异性和不规则的人为干预影响。其次，监测站点在空间上的分布极为不均匀。大约78%的日本监测站位于不到6%的国土面积内，集中在福岛附近，这打破了标准图模型的假设。第三，辐射与风、温度和湿度等异质环境因素通过大气输送过程共同演变，而纯数据驱动模型仅依靠观察数据难以捕捉这些关系。在本研究中，我们提出了NRFormer+，一种用于全国核辐射预测的时空变换器。NRFormer+将非平稳时间注意力和密度自适应空间注意力与一个新的大气扩散模块相结合，该模块估计气象如何驱动辐射扩散，并将这一物理信号作为结构先验注入网络中。NRFormer+在所有13个基准数据集上实现了最先进的准确性，在可比推理延迟下，突然变化的平均绝对误差（MAE）比最强基准降低了多达19.1%。我们的代码和数据集已公开发布，网址为 https://github.com/tfeilyu/NRFormer_Plus。
cs.AI / 14 / 2607.24777
Steering topology distributions for unified generative design of architected metamaterials
统一生成设计架构的拓扑分布引导用于建筑超材料
Abstract
Architected metamaterials derive their functions from structure, creating vast opportunities to program physical responses through topology design. However, existing design methods are often tailored to individual design problems, making limited use of topology knowledge for effective and broadly applicable design as objectives, constraints, and physical functions change. Here we introduce Generative Topology Optimization (GenTO), a unified framework that turns a learned topology prior into a reusable design engine. GenTO trains a diffusion model on a large full-order topology dataset and then iteratively steers the resulting topology distribution toward task-specific high-performing regions using user-defined physical objectives and constraints. This shifts the object of optimization from a single structure to a task-adapted topology distribution. Across topology design problems spanning thermal extremization, multi-objective morphology control, property-targeted auxetic design, and vibration transmission design, GenTO reuses pretrained topology priors for heterogeneous tasks, preserves structural diversity, and reaches high-performing solutions supported by numerical benchmarks and experimental validation. These results establish reusable topology knowledge as a unified principle for effective and scalable architected metamaterial design.
Chinese Translation
建筑超材料的功能源于其结构，通过拓扑设计创造了编程物理响应的广泛机会。然而，现有的设计方法通常针对个别设计问题进行定制，未能充分利用拓扑知识以应对目标、约束和物理功能的变化。本文介绍了生成拓扑优化（Generative Topology Optimization, GenTO），这是一个将学习到的拓扑先验转化为可重用设计引擎的统一框架。GenTO 在一个大型全阶拓扑数据集上训练扩散模型，然后使用用户定义的物理目标和约束，迭代地引导所得到的拓扑分布朝向特定任务的高性能区域。这一过程将优化对象从单一结构转变为任务适应的拓扑分布。在涵盖热极端化、多目标形态控制、属性目标的超弹性设计和振动传输设计的拓扑设计问题中，GenTO 重新利用预训练的拓扑先验以应对异构任务，保持结构多样性，并通过数值基准和实验验证达到高性能解决方案。这些结果确立了可重用的拓扑知识作为有效且可扩展的建筑超材料设计的统一原则。
cs.AI / 15 / 2607.24779
HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA：用于自适应在线广告的分层在线竞标代理
Abstract
Online advertising bidding systems typically deploy multiple offline-trained expert models (e.g., PID controllers, model predictive control, offline RL policies) but face two critical limitations: lack of online adaptability to non-stationary auction markets, and reliance on costly manual tuning of hyperparameters such as bid bounds and budget pacing constraints. We propose HOBA (Hierarchical On-policy Bidding Agents), a hierarchical reinforcement learning framework that decouples strategic reasoning, model selection, and bid execution across three time scales. At the high level, a large language model infers hyperparameters from contextual signals through a Think-Act-Observe-Reflect loop with historical experience retrieval. At the mid level, a SARSA agent dynamically selects among expert models, incorporating causal adjustment to eliminate selection bias. At the low level, a dynamic expert pool (PID, MPC, IQL, Decision Transformer) executes bids under high-level constraints. This design confines online learning to discrete expert selection rather than continuous bid optimization, significantly reducing exploration risk while maintaining adaptability. Experiments on the AuctionNet benchmark and a large-scale A/B test demonstrate consistent improvements over state-of-the-art baselines. In a large-scale online deployment, HOBA delivered substantial business value, achieving a +3.6\% increase in target cost, proving the effectiveness of our hierarchical multi-agent bidding paradigm.
Chinese Translation
在线广告竞标系统通常部署多个离线训练的专家模型（例如，PID控制器、模型预测控制、离线强化学习策略），但面临两个关键限制：缺乏对非平稳拍卖市场的在线适应能力，以及依赖于昂贵的手动调优超参数，如竞标边界和预算节奏约束。我们提出了HOBA（分层在线竞标代理），这是一种分层强化学习框架，它在三个时间尺度上解耦了战略推理、模型选择和竞标执行。在高层次上，一个大型语言模型通过历史经验检索的思考-行动-观察-反思循环，从上下文信号中推断超参数。在中层，一个SARSA代理动态选择专家模型，结合因果调整以消除选择偏差。在低层，一个动态专家池（PID、MPC、IQL、决策变换器）在高层约束下执行竞标。这一设计将在线学习限制在离散专家选择上，而不是连续竞标优化，显著降低了探索风险，同时保持了适应性。在AuctionNet基准测试和大规模A/B测试中的实验显示出相对于最先进基线的一致改进。在大规模在线部署中，HOBA带来了可观的商业价值，实现了目标成本提高3.6%的增长，证明了我们分层多代理竞标范式的有效性。
cs.AI / 16 / 2607.24780
LivingArena: Do LLMs Know What Other LLMs Don't? Peer-Probing as Scalable Evaluation
LivingArena：大型语言模型是否知道其他大型语言模型不知道的内容？同行探测作为可扩展评估
Abstract
Evaluating frontier LLMs is challenging: static benchmarks suffer from contamination and saturation -- leaving users unable to distinguish top models and developers blind to specific failure modes -- while human preference is subjective. In this paper, our question is: \emph{Do LLMs know what other LLMs don't? And can we leverage this dynamic for evaluation?} We present \textbf{LivingArena}, an automated, contamination-resistant evaluation framework. In this framework, models take turns proposing questions, aiming to pose items that opponents cannot answer correctly. Questioners are encouraged to actively identify and exploit opponents' knowledge boundaries, receiving rewards when the answerer fails, while the answerer is rewarded otherwise. To ensure questions contain objectively verifiable answers, a judge panel of strong models validates them, penalizing questioners if validation fails. Evaluating ten frontier LLMs, LivingArena yields a stable Elo leaderboard. Our behavioral analyses show that models identify and exploit their peers' cognitive boundaries: self-play and tournament logs indicate that they localize and double down on opponents' weak dimensions. Beyond static knowledge recall, peer probing measures factual rigor and the higher-order ability to probe an opponent's weaknesses, correlating only weakly with human preference and offering a scalable, low-cost approach to continuous evaluation.
Chinese Translation
评估前沿大型语言模型（LLMs）具有挑战性：静态基准测试受到污染和饱和的影响，使得用户无法区分顶尖模型，开发者也无法识别特定的失败模式，而人类偏好则是主观的。本文提出的问题是： extit{大型语言模型是否知道其他大型语言模型不知道的内容？我们能否利用这种动态进行评估？} 我们提出了 extbf{LivingArena}，一个自动化的、抗污染的评估框架。在该框架中，模型轮流提出问题，旨在提出对手无法正确回答的项目。提问者被鼓励积极识别和利用对手的知识边界，当回答者失败时，提问者会获得奖励，而回答者则在成功回答时获得奖励。为了确保问题包含客观可验证的答案，一个由强大模型组成的评审小组对其进行验证，如果验证失败则对提问者进行惩罚。在对十个前沿大型语言模型进行评估时，LivingArena产生了一个稳定的Elo排行榜。我们的行为分析表明，模型能够识别并利用同伴的认知边界：自我对弈和比赛记录表明它们能够定位并加大对手的弱点。除了静态知识回忆，同行探测还衡量了事实的严谨性和探测对手弱点的高阶能力，与人类偏好的相关性较弱，并提供了一种可扩展、低成本的持续评估方法。
cs.AI / 17 / 2607.24782
Personalization, Personas, and Forecasting in Value Alignment
价值对齐中的个性化、角色模型与预测
Abstract
LLM behavior may be conditioned by human identity in several ways: they may be asked to adapt to users, role-play populations, or forecast how people would answer value-laden questions. We test whether these framings are interchangeable using the World Values Survey (WVS). We evaluate GPT-5.4, Claude Sonnet 4.6, Gemini 2.5 Flash, and Qwen3-235B on 101 WVS-derived questions across 13 language-country slices, comparing a language-only baseline with user-country, persona-country, and third-person prompts. Across 21,008 model-response rows, prompt framing is a first-order determinant of cultural alignment: country cues often shift answers substantially, but not all shifts move toward matched human response distributions. Third-person forecasting yields the strongest directional alignment for three of the four hosted models, while personalization and role-play are weaker or less stable. Alignment gains concentrate on salient value dimensions such as religiosity, gender roles, and work-oriented material values, whereas institutional trust and democracy-related questions remain difficult. These results show that prompt framing is not a cosmetic choice in cultural value elicitation; it changes both model behavior and measured alignment.
Chinese Translation
大型语言模型（LLM）的行为可能受到人类身份的多种方式的影响：它们可能被要求适应用户、角色扮演特定人群，或预测人们如何回答涉及价值观的问题。我们使用世界价值观调查（World Values Survey, WVS）测试这些框架是否可以互换。我们评估了GPT-5.4、Claude Sonnet 4.6、Gemini 2.5 Flash和Qwen3-235B在基于101个WVS衍生问题的13个语言-国家切片上的表现，比较了仅基于语言的基线与用户-国家、角色模型-国家和第三人称提示。通过21,008个模型响应行的分析，提示框架是文化对齐的首要决定因素：国家提示通常会显著改变答案，但并非所有变化都朝向匹配的人类响应分布。第三人称预测为四个模型中的三个提供了最强的方向性对齐，而个性化和角色扮演则较弱或不太稳定。对齐的提升集中在显著的价值维度上，如宗教信仰、性别角色和以工作为导向的物质价值，而对制度信任和与民主相关的问题仍然较为困难。这些结果表明，提示框架在文化价值引导中并不是一种表面选择；它改变了模型的行为和测量的对齐程度。
cs.AI / 18 / 2607.24783
Unified Semantic Modeling Framework for Large-Scale Job Understanding at LinkedIn
LinkedIn 大规模职位理解的统一语义建模框架
Abstract
Job understanding is critical to LinkedIn's mission of connecting talent with opportunity. This task involves transforming unstructured and noisy job postings into standardized or derived job attributes that power numerous LinkedIn products. However, building a scalable, cost-efficient, and high-performing job understanding system remains challenging. In this paper, we present a unified semantic modeling framework powered by a small language model (SLM) to address the challenges. We begin by fine-tuning an open-source SLM using a suite of carefully curated synthetic tasks augmented with reasoning traces. These tasks jointly target taxonomy-guided classification and taxonomy-agnostic entity extraction. This allows the resulting model to acquire robust zero-shot generalization for job understanding in structured and unstructured contexts. Building upon this foundation, we introduce a multi-adapter architecture with attribute grouping to facilitate efficient task-specific adaptation while streamlining model management across diverse downstream attributes. Offline evaluations and online A/B tests demonstrate significant performance improvement while reducing operational complexity. Our work provides practical insights into building industry-scale text understanding systems.
Chinese Translation
职位理解对于 LinkedIn 实现人才与机会连接的使命至关重要。该任务涉及将非结构化和噪声较大的职位发布信息转化为标准化或派生的职位属性，这些属性为众多 LinkedIn 产品提供支持。然而，构建一个可扩展、成本高效且高性能的职位理解系统仍然面临挑战。本文提出了一种由小型语言模型（SLM）驱动的统一语义建模框架，以应对这些挑战。我们首先通过一系列经过精心策划的合成任务，并结合推理轨迹，对开源 SLM 进行微调。这些任务共同针对分类法引导的分类和无分类法的实体提取。这使得最终模型能够在结构化和非结构化上下文中获得强大的零样本泛化能力。基于此基础，我们引入了一种多适配器架构，并通过属性分组促进高效的任务特定适配，同时简化不同下游属性的模型管理。离线评估和在线 A/B 测试表明，性能显著提升，同时降低了操作复杂性。我们的工作为构建行业级文本理解系统提供了实用的见解。
cs.AI / 19 / 2607.24784
On the Use of LLMs for Specialised Terminology: A Good Alternative to Corpora?
大型语言模型在专业术语使用中的应用：对语料库的良好替代方案？
Abstract
Specialised translation relies on the use of documentary and terminological resources, including corpora. These resources are particularly useful for terminology. However, their compilation and exploitation have several limitations: they require time, technical skills and access to data that can be difficult to collect. This study examines the extent to which LLMs can assist specialised translators in finding equivalents from English to French. We evaluate four proprietary models, GPT-4o, GPT-5.2, Claude Sonnet 4.5 and DeepSeek, in two specialised domains, Earth, Environmental and Planetary Sciences (EEPS) and Natural Language Processing (NLP). The experiment is based on 80 terms per domain and compares two prompting strategies: a terminology and a translation mode. The results highlight clear differences between models, prompting strategies and, to a lesser extent, domains. Claude Sonnet 4.5 achieves the best results in the most favourable configuration, while DeepSeek stands out for its greater stability. Analysis of confidence estimates also shows that they are only a partial indicator of terminological accuracy. Overall, the findings suggest that LLMs can be useful tools for specialised translators, but cannot, at this stage, replace specialised corpora. This research therefore paves the way for future work on the real practical usefulness of LLMs for specialised translators in work and educational contexts.
Chinese Translation
专业翻译依赖于文献和术语资源的使用，包括语料库。这些资源在术语方面特别有用。然而，它们的编纂和利用存在若干局限性：需要时间、技术技能以及可能难以收集的数据访问。本文研究了大型语言模型（LLMs）在帮助专业翻译人员从英语到法语寻找等价词方面的有效性。我们评估了四个专有模型：GPT-4o、GPT-5.2、Claude Sonnet 4.5 和 DeepSeek，在两个专业领域，即地球、环境与行星科学（EEPS）和自然语言处理（NLP）中进行实验。实验基于每个领域80个术语，并比较了两种提示策略：术语模式和翻译模式。结果显示模型、提示策略之间存在明显差异，而领域的影响相对较小。在最有利的配置中，Claude Sonnet 4.5 达到了最佳结果，而 DeepSeek 则因其更大的稳定性而脱颖而出。对置信度估计的分析还表明，置信度仅是术语准确性的部分指标。总体而言，研究结果表明，LLMs可以成为专业翻译人员的有用工具，但在目前阶段尚无法替代专业语料库。因此，本研究为未来关于LLMs在工作和教育环境中对专业翻译人员实际实用性的研究铺平了道路。
cs.AI / 20 / 2607.24787
SpecPrefetch: Parameter-Efficient Expert Prefetching for Sparse MoE Foundation Models
SpecPrefetch：稀疏 MoE 基础模型的参数高效专家预取
Abstract
Sparse Mixture-of-Experts (MoE) models expand foundation model capacity through conditional expert activation, but their full expert pools remain difficult to deploy under limited accelerator memory. Although expert offloading alleviates memory pressure by moving inactive experts to host memory or storage, it introduces a routing-dependent transfer bottleneck: required experts are known only after native top-\(K\) routing, which serializes routing, expert loading, and expert execution during inference. To address this bottleneck, we propose SpecPrefetch, a parameter-efficient prefetching framework for offloaded MoE inference. SpecPrefetch uses a shared lightweight adapter to predict next-layer expert candidates only for asynchronous transfer, while the frozen native router still determines the final executed experts. By separating transfer prediction from execution routing, SpecPrefetch reduces exposed expert-loading latency without changing pretrained routing semantics, so prediction errors affect transfer efficiency rather than model outputs. In addition, a window-aware scheduler prioritizes feasible transfers under cache and bandwidth constraints. Across Qwen3-VL-30B-A3B and DeepSeek-VL2-Tiny, SpecPrefetch achieves the best average expert recall in 9 out of 10 model-benchmark settings with substantially fewer trainable parameters than learned predictor baselines. On a Snapdragon 8 Elite device, SpecPrefetch further improves decoding throughput by up to \(20\%\) over a compute-optimized offloading runtime, demonstrating practical benefits for storage-constrained MoE deployment. The code and model weights are available at https://github.com/wei390/SpecPrefetch.
Chinese Translation
稀疏专家混合模型（MoE）通过条件专家激活扩展基础模型的容量，但在有限的加速器内存下，完整的专家池仍然难以部署。尽管专家卸载通过将非活动专家移动到主机内存或存储中缓解了内存压力，但它引入了一个依赖路由的传输瓶颈：所需专家仅在原生 top-
K 路由后才能确定，这在推理过程中串行化了路由、专家加载和专家执行。为了解决这一瓶颈，我们提出了 SpecPrefetch，这是一种针对卸载 MoE 推理的参数高效预取框架。SpecPrefetch 使用共享的轻量适配器仅预测下一层专家候选者以进行异步传输，而冻结的原生路由器仍然决定最终执行的专家。通过将传输预测与执行路由分离，SpecPrefetch 在不改变预训练路由语义的情况下减少了暴露的专家加载延迟，因此预测错误影响的是传输效率而非模型输出。此外，窗口感知调度器在缓存和带宽限制下优先考虑可行的传输。在 Qwen3-VL-30B-A3B 和 DeepSeek-VL2-Tiny 上，SpecPrefetch 在 10 个模型基准设置中实现了 9 次最佳平均专家召回率，并且可训练参数显著少于学习预测基线。在 Snapdragon 8 Elite 设备上，SpecPrefetch 进一步提高了解码吞吐量，较计算优化的卸载运行时提高了高达 20%，展示了在存储受限的 MoE 部署中的实际效益。代码和模型权重可在 https://github.com/wei390/SpecPrefetch 获取。
cs.AI / 21 / 2607.24788
GLIDE: Guided Layerwise Hybrid Attention for Efficient LLM Inference
GLIDE：用于高效大型语言模型推理的引导层级混合注意力
Abstract
As Large Language Models scale to increasingly long contexts, the memory I/O and computational overhead of the Key-Value (KV) cache during decoding emerges as the primary throughput bottleneck. To address this, we propose GLIDE, a Guided Layerwise Hybrid Attention that strategically integrates sliding-window softmax attention with linear recurrent aggregation. GLIDE is motivated by layer-wise heterogeneity: early layers exhibit high sensitivity to softmax removal, while deeper layers demonstrate redundancy and tolerate aggressive replacement by linear alternatives. Leveraging this insight, GLIDE introduces a layer-wise adaptive mechanism wherein each layer balances an efficient linear recurrence with a variable-sized softmax window. Unlike uniform hybrid approaches, GLIDE non-uniformly compresses the softmax footprint across the model, reducing aggregate KV cache I/O while preserving expressive power where most vital. Empirical evaluations demonstrate the GLIDE achieves superior performance-efficiency tradeoffs, reducing end-to-end latency for long-context generation without compromising quality.
Chinese Translation
随着大型语言模型在越来越长的上下文中扩展，解码过程中键值（Key-Value，KV）缓存的内存输入/输出和计算开销成为主要的吞吐量瓶颈。为了解决这个问题，我们提出了GLIDE，一种引导层级混合注意力，战略性地将滑动窗口软最大（softmax）注意力与线性递归聚合相结合。GLIDE的设计灵感来源于层级异质性：早期层对软最大移除表现出高度敏感，而较深层则表现出冗余性，能够容忍被线性替代方案大幅替换。利用这一洞察，GLIDE引入了一种层级自适应机制，使每一层在高效的线性递归与可变大小的软最大窗口之间取得平衡。与均匀混合方法不同，GLIDE在模型中非均匀地压缩软最大的占用空间，减少了总的KV缓存输入/输出，同时在最重要的地方保留了表达能力。实证评估表明，GLIDE在性能与效率的权衡上表现优越，减少了长上下文生成的端到端延迟，而不影响质量。
cs.AI / 22 / 2607.24790
A GAN-Based Framework for Robust Data Synthesis in Satellite Internet Observations
基于生成对抗网络的卫星互联网观测中鲁棒数据合成框架
Abstract
Low-Earth orbit (LEO) satellite Internet has become an important infrastructure for enabling ubiquitous connectivity to align with the International Telecommunications Union vision for 6G telecommunications networks. However, current LEO satellite Internet observations often suffer from missing data, which complicates data augmentation task and limits the expansion of representative datasets. Given the complex characteristics of these datasets, generative AI (GenAI) presents a promising approach, yet its application in this domain has received little attention to date. In this paper, we propose a GenAI-based framework to synthesize high-fidelity data directly from incomplete LEO network observations. We propose the representative data missing scenarios, and evaluate the performance with the latest GAN- and VAE-based GenAI models on the recent WetLinks dataset. We design block-wise and point-wise missing scenarios to closely simulate the data loss that happens on real-world LEO satellite networks. Our results show the effectiveness of our proposed GAN-based framework and GT-GAN model exhibits the best performance among all models in both missing scenarios. Even under extreme conditions (e.g., 40% of the input data is missing), GT-GAN shows the highest robustness, consistently capturing the underlying input data distribution and being the least affected in terms of generalization. Our results shed light on future directions for GenAI-based data augmentation methods and data-driven research on satellite network measurement.
Chinese Translation
低地球轨道（LEO）卫星互联网已成为实现无处不在的连接的重要基础设施，以符合国际电信联盟对6G电信网络的愿景。然而，目前的LEO卫星互联网观测常常面临数据缺失的问题，这使得数据增强任务变得复杂，并限制了代表性数据集的扩展。鉴于这些数据集的复杂特征，生成性人工智能（GenAI）呈现出一种有前景的方法，但迄今为止在该领域的应用鲜有关注。本文提出了一种基于GenAI的框架，旨在直接从不完整的LEO网络观测中合成高保真数据。我们提出了代表性的数据缺失场景，并在最新的WetLinks数据集上评估了最新的基于GAN和VAE的GenAI模型的性能。我们设计了块状和点状缺失场景，以密切模拟在现实世界LEO卫星网络中发生的数据丢失。我们的结果表明，所提出的基于GAN的框架的有效性，其中GT-GAN模型在所有模型中在两种缺失场景下表现最佳。即使在极端条件下（例如，40%的输入数据缺失），GT-GAN也表现出最高的鲁棒性，始终能够捕捉到潜在的输入数据分布，并在泛化方面受到的影响最小。我们的结果为基于GenAI的数据增强方法和卫星网络测量的数据驱动研究指明了未来的方向。
cs.AI / 23 / 2607.24794
Reasoning with Memory: A Temporal Granularity-Adaptive Framework for Training-Free Long Video Understanding
基于记忆的推理：一种适应时间粒度的无训练长视频理解框架
Abstract
While Multimodal Large Language Models (MLLMs) demonstrate superior generalization in fundamental video tasks, restricted context windows limit their long video understanding. To accommodate this constraint, models typically resort to keyframe selection. However, uniform sampling or static query-guided selection often overlooks critical temporal context, failing to adapt to the varying query temporal granularities. In this paper, we propose ReMem, a temporal granularity-adaptive keyframe selection framework for training-free LongVideoQA. ReMem introduces a dual-level memory-augmented adaptation. At the query level, Memory-Driven Question Parsing leverages LLM long-term memory to decode question temporal granularity and extract semantic entities. At the video level, Synergistic Dual-Semantic Frame Alignment exploits intrinsic structural memory to align frames with query semantics, guiding Structure-Aware Dynamic Frame Routing to cluster events and optimally distribute sampling budgets. By explicitly preserving temporal information with memory mechanisms, ReMem suppresses redundancy and empowers MLLMs to perform robust multi-granular video reasoning. Evaluations across four popular LongVideoQA benchmarks using three MLLMs demonstrate highly efficient, state-of-the-art zero-shot performance; notably, LLaVA-Video with ReMem reaches 54.5% (+12.3%) on LVBench and 67.1% (+8.2%) on LongVideoBench.
Chinese Translation
尽管多模态大型语言模型（MLLMs）在基本视频任务中表现出优越的泛化能力，但受限的上下文窗口限制了它们对长视频的理解。为了适应这一限制，模型通常采用关键帧选择。然而，均匀采样或静态查询引导的选择往往忽视了关键的时间上下文，未能适应不同查询时间粒度的变化。本文提出了ReMem，一种适应时间粒度的关键帧选择框架，用于无训练的LongVideoQA。ReMem引入了双层记忆增强适应机制。在查询层面，记忆驱动的问题解析利用LLM的长期记忆来解码问题的时间粒度并提取语义实体。在视频层面，协同双语义帧对齐利用内在结构记忆将帧与查询语义对齐，指导结构感知的动态帧路由以聚类事件并最优分配采样预算。通过显式保留时间信息和记忆机制，ReMem抑制了冗余，使MLLMs能够进行稳健的多粒度视频推理。在使用三种MLLMs对四个流行的LongVideoQA基准进行评估时，表现出高效的最先进的零-shot性能；特别是，结合ReMem的LLaVA-Video在LVBench上达到了54.5%（+12.3%）和在LongVideoBench上达到了67.1%（+8.2%）。
cs.AI / 24 / 2607.24795
When Shortest Isn't Safest: A Design Science Approach to Senior-Friendly Pedestrian Routing
最短并不一定最安全：一种面向老年人的行人导航设计科学方法
Abstract
Older adults' independent mobility enables out-of-home participation, well-being and health, yet pedestrian navigation systems still optimize primarily for distance or time, often overlooking barriers, safety thresholds, and supportive infrastructure that shape late-life walking decisions. We present a senior-friendly pedestrian routing artefact developed through echeloned Design Science Re-search, translating lived mobility constraints into prescriptive design knowledge. Based on 11 semi-structured interviews, we derive initial Design Requirements (DRs) and Design Principles (DPs) for barrier-aware, amenity-sensitive routing and execution-relevant explanations. We instantiate these in an OpenStreetMap pedestrian network enriched with amenities (benches, toilets, and shelters) and height data, and implemented an A*-based routing engine with configurable costs and explanation payloads. In a field-based walking study, 14 older adults com-pared artefact-generated routes with baselines and provided ratings and qualitative feedback; the senior-friendly route was preferred overall. Thematic analysis further showed that infrastructure maintenance, seasonal conditions, traffic exposure, and social context shape route acceptance. We synthesize these insights into refined DRs and DPs emphasizing context-aware hazard modeling, multi-route transparency, landmark-grounded explanations, social-context sensitivity, and stage-appropriate information. Our contributions provide actionable guidance for practitioners developing senior-friendly pedestrian navigation systems.
Chinese Translation
老年人的独立出行能力促进了他们的户外参与、幸福感和健康，但现有的行人导航系统仍主要优化距离或时间，常常忽视影响晚年步行决策的障碍、安全阈值和支持性基础设施。我们提出了一种通过分层设计科学研究开发的老年友好型行人导航工具，将实际出行限制转化为规范性设计知识。基于11次半结构化访谈，我们得出了针对障碍感知、设施敏感的导航的初步设计需求（DRs）和设计原则（DPs），以及与执行相关的解释。我们在一个丰富了设施（长椅、厕所和避难所）和高度数据的OpenStreetMap行人网络中实现了这些设计，并实施了一个基于A*算法的导航引擎，具有可配置的成本和解释负载。在一项基于实地的步行研究中，14名老年人比较了该工具生成的路线与基线路线，并提供了评分和定性反馈；整体上，老年友好的路线更受欢迎。主题分析进一步表明，基础设施维护、季节性条件、交通暴露和社会背景影响路线的接受度。我们将这些见解综合为改进的设计需求和设计原则，强调情境感知的危险建模、多路线透明度、基于地标的解释、社会背景敏感性和阶段适宜的信息。我们的贡献为开发老年友好的行人导航系统的从业者提供了可操作的指导。
cs.AI / 25 / 2607.24810
RRS-10K: A Multitask Vision-Language Model Benchmark for Rare Remote Sensing Image Interpretation
RRS-10K：用于稀有遥感图像解读的多任务视觉-语言模型基准
Abstract
Vision-language models (VLMs) have achieved strong performance on general remote sensing tasks. However, their capability for rare scenes remains insufficiently understood, because existing benchmarks are dominated by common urban and rural imagery. To address this gap, we present RRS-10K, a benchmark for rare remote sensing image interpretation. RRS-10K contains 10,738 military-related remote sensing images and corresponding multiple format question-answer pairs for comprehensive evaluation. All of the images are collected from first-hand sources and organized into three capability dimensions, six sub-dimensions, and 20 leaf tasks, covering perception, reasoning, and robustness. To improve the quality of multiple-choice questions, we introduce a similarity-based distractor filtering strategy (SDFS) during benchmark construction. We further evaluate 52 representative models and show that current VLMs achieve only moderate zero-shot performance on rare remote sensing image interpretation, with clear weaknesses in visual grounding, referring segmentation, and complex semantic reasoning tasks. RRS-10K enables systematic analysis of failure modes in long-tail remote sensing interpretation and provides guidance for developing more reliable remote sensing VLMs.
Chinese Translation
视觉-语言模型（VLMs）在一般遥感任务上取得了良好的表现。然而，它们在稀有场景下的能力仍然不足以理解，因为现有基准主要集中于常见的城市和乡村图像。为了解决这一问题，我们提出了RRS-10K，这是一个用于稀有遥感图像解读的基准。RRS-10K包含10,738幅与军事相关的遥感图像及相应的多格式问答对，以便进行全面评估。所有图像均来自第一手来源，并被组织为三个能力维度、六个子维度和20个叶子任务，涵盖感知、推理和鲁棒性。为了提高多项选择题的质量，我们在基准构建过程中引入了一种基于相似性的干扰项过滤策略（SDFS）。我们进一步评估了52个代表性模型，并显示当前的VLM在稀有遥感图像解读上的零-shot表现仅为中等，且在视觉定位、指称分割和复杂语义推理任务上存在明显的弱点。RRS-10K使得对长尾遥感解读中的失败模式进行系统分析成为可能，并为开发更可靠的遥感VLM提供了指导。
cs.AI / 26 / 2607.24814
Aletheia: An Offline-First Clinical Decision Support System for Differential Diagnosis in Low-Resource Healthcare Settings
Aletheia：一种面向低资源医疗环境的离线优先临床决策支持系统用于鉴别诊断
Abstract
Access to specialist clinical expertise remains severely limited across sub-Saharan Africa, where physician-to-patient ratios can fall below 1:25,000 in rural settings. Existing AI-assisted diagnostic tools predominantly require reliable internet connectivity and high-specification hardware, rendering them impractical for frontline healthcare workers in district hospitals and health centres. This paper presents Aletheia, an offline-first clinical decision support system designed for low-resource healthcare contexts across sub-Saharan Africa. Aletheia is built upon Qwen2.5-3B-Instruct, fine-tuned using Quantised Low- Rank Adaptation (QLoRA) on a curated dataset of 27,000 clinical reasoning samples spanning 50 disease conditions with elevated prevalence in East Africa. Evaluation demonstrates a Top-1 diagnostic accuracy of 80.0%, Top-3 accuracy of 100.0%, BERTScore-F1 of 0.909, and METEOR of 0.467 across ten representative clinical case categories. The system achieves an Expected Calibration Error (ECE) of 0.275 and passes the Africa Deep Tech Challenge 2026 (ADTC 2026) memory budget constraint of 7 168 MB, achieving a peak inference RAM of approximately 3 630 MB on the standardised benchmark laptop. These results demonstrate the feasibility of deploying large language model-based clinical reasoning at the primary care level in resource-constrained settings without cloud infrastructure.
Chinese Translation
在撒哈拉以南非洲，专业临床专家的获取仍然严重有限，尤其是在农村地区，医生与患者的比例可能低于1:25,000。现有的人工智能辅助诊断工具主要依赖于可靠的互联网连接和高规格硬件，这使得它们在地区医院和卫生中心的前线医疗工作者中变得不切实际。本文介绍了Aletheia，一种为撒哈拉以南非洲低资源医疗环境设计的离线优先临床决策支持系统。Aletheia基于Qwen2.5-3B-Instruct构建，使用量化低秩适配（Quantised Low-Rank Adaptation, QLoRA）在一个包含27,000个临床推理样本的精心策划数据集上进行了微调，这些样本涵盖了在东非高发的50种疾病。评估结果显示，该系统在十个代表性临床案例类别中的Top-1诊断准确率为80.0%，Top-3准确率为100.0%，BERTScore-F1为0.909，METEOR为0.467。该系统实现了期望校准误差（Expected Calibration Error, ECE）为0.275，并通过了2026年非洲深科技挑战赛（Africa Deep Tech Challenge 2026, ADTC 2026）内存预算限制7,168 MB，在标准化基准笔记本电脑上实现了约3,630 MB的峰值推理内存。这些结果证明了在资源有限的环境中，无需云基础设施即可在初级医疗层面部署基于大型语言模型的临床推理的可行性。
cs.AI / 27 / 2607.24833
AdaKP: Online Adaptive Knowledge-Point Selection for Reasoning-Oriented Reinforcement Learning
AdaKP：面向推理的强化学习中的在线自适应知识点选择
Abstract
Reinforcement learning with verifiable rewards is a powerful paradigm for eliciting reasoning in large language models, yet it suffers from severe reward sparsity on competition-level mathematics. A common remedy injects atomic knowledge points (KPs) - short natural-language hints distilled from gold solutions - into the prompt. Existing methods, however, either fix this selection once offline or merely scale the monolithic quantity of injected text, leaving untouched the most informative axis of choice: which subset of atomic KPs to inject, and when. We introduce AdaKP, an online selector that re-chooses each problem's KP subset over the course of RL training. At its core is an entropy proxy that scores a KP by the reduction in next-token entropy it induces - a single inexpensive forward pass, with a provable bound on its truncation bias - in place of expensive rollout-based estimation. Three lightweight mechanisms make this signal usable online: a momentum smoother that absorbs per-step noise, a retirement-and-revival manager that prunes weak KPs while preserving exploration, and an adaptive scheduler that front-loads re-evaluations into early training. AdaKP further contributes a pre-flight validation gate that certifies the proxy against a leave-one-out ground truth before any expensive run is launched, turning method-level risk into a falsifiable check. Realized as a fully additive fork of a standard DAPO+GRPO trainer with no optimizer changes, AdaKP improves over a strong static-selection baseline on all eight competition-mathematics benchmarks at negligible added cost, positioning online, validated KP-subset selection as a practical and as-yet under-explored axis for reasoning-oriented reinforcement learning.
Chinese Translation
具有可验证奖励的强化学习是引发大型语言模型推理的强大范式，但在竞争级数学问题上却面临严重的奖励稀疏性。常见的解决方案是将原子知识点（Knowledge Points, KPs）——从黄金解决方案中提炼出的简短自然语言提示——注入到提示中。然而，现有方法要么在离线时固定选择，要么仅仅扩展注入文本的单一数量，未能触及选择的最具信息量的维度：选择哪个原子KPs的子集进行注入，以及何时进行注入。我们提出了AdaKP，这是一种在线选择器，在强化学习训练过程中重新选择每个问题的KP子集。其核心是一个熵代理，通过它所引起的下一个标记熵的减少来对KP进行评分——这只需一次便宜的前向传递，并且其截断偏差有可证明的界限——替代了昂贵的基于回滚的估计。三个轻量级机制使得这个信号可以在线使用：一个动量平滑器用于吸收每步的噪声，一个退休与复兴管理器用于修剪弱KPs同时保持探索，以及一个自适应调度器将重新评估前置到早期训练中。AdaKP还贡献了一个预飞行验证门，在任何昂贵的运行启动之前，针对留一法真实值对代理进行认证，将方法级风险转化为可证伪的检查。作为标准DAPO+GRPO训练器的完全加法分支，AdaKP在所有八个竞争数学基准上以微不足道的额外成本超越了强大的静态选择基线，确立了在线、经过验证的KP子集选择作为面向推理的强化学习中一个实用且尚待深入探索的维度。
cs.AI / 28 / 2607.24873
MusiChat: Vibe Composing for Music Creation
MusiChat：音乐创作的氛围编曲
Abstract
Recent advances in AI music generation have enabled users to create complete musical pieces from natural-language prompts. However, most existing systems follow a prompt-and-regenerate paradigm, making iterative refinement difficult because users must repeatedly recreate compositions instead of directly evolving existing musical ideas. We present MusiChat, a conversational vibe composing system that enables collaborative human-AI music creation through natural-language interaction and iterative refinement. At the core of MusiChat is a hierarchical controllable music generation framework that separates lyric-aligned musical structure generation from expressive surface realization, allowing flexible stylistic transformations and structure-preserving edits. The system integrates a large language model with a hybrid symbolic music engine through a memory-augmented architecture that maintains the active composition state and user history across interactions. A hybrid intent-routing mechanism further enables efficient interpretation of both precise musical edits and open-ended creative requests. Rather than regenerating compositions from scratch, MusiChat incrementally transforms an evolving musical artifact while preserving relevant musical structure and user intent. We evaluate MusiChat through objective analysis and human studies, achieving 95.31% and 100% accuracy for single- and multi-turn interactions, respectively, and obtaining like-to-dislike ratios of 2:1 for melody naturalness and 3:1 for musical quality. Our results demonstrate that MusiChat supports coherent multi-turn music authoring and interactive human-AI co-creation through a conversational interface.
Chinese Translation
近期人工智能音乐生成的进展使得用户能够通过自然语言提示创作完整的音乐作品。然而，大多数现有系统遵循提示-再生成的范式，使得迭代优化变得困难，因为用户必须反复重建作品，而不是直接发展已有的音乐创意。我们提出了MusiChat，一个对话式氛围编曲系统，通过自然语言交互和迭代优化实现人机协作音乐创作。MusiChat的核心是一个分层可控的音乐生成框架，该框架将与歌词对齐的音乐结构生成与表现表面实现分开，从而允许灵活的风格转换和结构保持编辑。该系统通过增强记忆的架构将大型语言模型与混合符号音乐引擎集成，保持交互过程中的活跃创作状态和用户历史。混合意图路由机制进一步实现了对精确音乐编辑和开放式创意请求的高效解释。MusiChat不是从头再生成作品，而是逐步转化一个不断演变的音乐工件，同时保持相关的音乐结构和用户意图。我们通过客观分析和人类研究评估MusiChat，单轮和多轮交互的准确率分别达到95.31%和100%，并且在旋律自然性和音乐质量方面的喜好比率分别为2:1和3:1。我们的结果表明，MusiChat通过对话界面支持连贯的多轮音乐创作和互动的人机协作。
cs.AI / 29 / 2607.24995
Understanding Semantic IDs: From Item Representation to Item Selection in Generative Recommendation
理解语义ID：从项目表示到生成推荐中的项目选择
Abstract
Semantic IDs (SIDs) are now a central component of generative recommendation. Current SID-based systems assign three roles to the same token sequence. Shared prefixes are intended to organize related items, the complete SID identifies an individual item, and each generated token narrows the items that can still be returned. We systematically investigate SIDs from item encoding and SID construction to autoregressive generation and final recommendation. We examine how SID construction changes item representations and how those changes affect generation. Across three Amazon domains and eight SID constructions, SID neighborhoods recover only 32.2% of the encoder's ten nearest neighbors on average. Alternative item descriptions still retrieve the corresponding item first in 99.57% of controlled cases, yet change 38.4% of exact SIDs. These results show that SIDs retain broad organization but lose much of the encoder's fine local structure, while their exact tokens are not determined by item meaning alone. This loss becomes consequential during generation. After the final semantic token, TIGER retains only 29.9% of held-out targets that were plausible recommendations before SID filtering. Motivated by these findings, we propose Item-Supported Decoding (ISD), a lightweight inference-time method that allows a user-specific item ranking to support corresponding SID prefixes before beam search discards them. The same ranking then orders the generated items. ISD requires no additional parameters or retraining of the SID constructor or decoder. We empirically show that ISD improves NDCG@10 over the corresponding SID backbone in every evaluated setting, with relative gains of up to 31.2%. Our results show that SIDs provide useful coarse item organization, but their fine boundaries should not alone determine which items remain available during generation.
Chinese Translation
语义ID（SIDs）现在是生成推荐的核心组成部分。目前基于SID的系统为相同的令牌序列分配了三种角色。共享前缀旨在组织相关项目，完整的SID用于识别单个项目，而每个生成的令牌则缩小了仍然可以返回的项目范围。我们系统地研究了SID，从项目编码和SID构建到自回归生成和最终推荐。我们考察了SID构建如何改变项目表示，以及这些变化如何影响生成。在三个亚马逊领域和八种SID构建中，SID邻域平均仅恢复了编码器十个最近邻的32.2%。替代项目描述在99.57%的受控情况下仍能首先检索到相应项目，但改变了38.4%的精确SID。这些结果表明，SID保留了广泛的组织结构，但失去了编码器的细微局部结构，而其精确令牌并不单由项目含义决定。这种损失在生成过程中变得至关重要。在最终的语义令牌之后，TIGER仅保留了29.9%的被排除目标，这些目标在SID过滤之前是合理的推荐。基于这些发现，我们提出了项目支持解码（Item-Supported Decoding, ISD），这是一种轻量级的推理时方法，允许用户特定的项目排名支持相应的SID前缀，然后在束搜索丢弃它们之前进行排序。相同的排名随后对生成的项目进行排序。ISD不需要额外的参数或重新训练SID构建器或解码器。我们实证表明，ISD在每个评估设置中都提高了相应SID基础模型的NDCG@10，相对增益高达31.2%。我们的结果表明，SID提供了有用的粗略项目组织，但其细微边界不应单独决定在生成过程中哪些项目仍然可用。
cs.AI / 30 / 2607.25020
Localized Anomaly Detection via Differentiable D-vine Copulas
通过可微分 D-vine Copulas 进行局部异常检测
Abstract
Vine copulas provide a flexible framework for modeling complex multivariate distributions through a hierarchical decomposition into bivariate pair-copulas. Fitting a D-vine requires selecting a copula family and parameter configuration for each pair-copula from a set of candidates encoding different dependence patterns. As the number of variables and candidate families increases, the number of possible configurations grows combinatorially. Existing fitting procedures address this challenge through sequential greedy decisions, committing to a single locally optimal family at each step and potentially discarding configurations that would yield a better global fit. To overcome this limitation, we propose a novel estimation framework that combines gradient-based maximum likelihood estimation, enabled by our fully differentiable implementation, with a beam-search strategy that maintains multiple competing D-vine configurations throughout the fitting process. This allows a broader exploration of the configuration space while remaining computationally tractable. Building on the fitted D-vine, we introduce a localized anomaly detection framework that exploits the hierarchical decomposition to produce both global anomaly scores and edge-level explanations. Statistical guarantees are provided through Mondrian conformal prediction, while the pair-copula structure enables the localization of anomalies to specific variable relationships. We evaluate the proposed framework on both benchmark and real-world datasets, demonstrating its effectiveness for interpretable anomaly detection with uncertainty quantification.
Chinese Translation
Vine copulas 提供了一种灵活的框架，通过将复杂的多变量分布分解为二元对 copulas 来建模。拟合 D-vine 需要为每个二元 copula 从一组编码不同依赖模式的候选者中选择一个 copula 家族和参数配置。随着变量数量和候选家族的增加，可能的配置数量呈组合增长。现有的拟合程序通过顺序贪婪决策来应对这一挑战，在每一步中承诺选择一个局部最优的家族，可能会丢弃那些能够提供更好全局拟合的配置。为了克服这一限制，我们提出了一种新颖的估计框架，该框架结合了基于梯度的最大似然估计（通过我们完全可微分的实现实现）和在拟合过程中保持多个竞争 D-vine 配置的束搜索策略。这使得在保持计算可行性的同时，能够更广泛地探索配置空间。在拟合的 D-vine 的基础上，我们引入了一种局部异常检测框架，利用层次分解生成全局异常评分和边缘级解释。通过 Mondrian 适应性预测提供统计保证，而对 copula 结构的利用使得异常能够定位到特定的变量关系。我们在基准和真实世界数据集上评估了所提框架，展示了其在可解释异常检测和不确定性量化方面的有效性。
cs.AI / 31 / 2607.25021
Chart-Supported or Model-Supplied? Examining MLLM-Generated Claims for Accessible Visualization
图表支持还是模型提供？检视 MLLM 生成的可访问可视化声明
Abstract
Multimodal large language models (MLLMs) can connect visualization patterns to external causes, consequences, and domain knowledge, but the evidential basis of these interpretations is often unclear. We present an exploratory study of 102 visualizations from four sources, three MLLMs, and four input conditions that vary access to the image, source-specific accessible chart context, and withheld-context framing. Across 1,224 descriptions, we analyze model-attributed DIRECT, DERIVED, and SPECULATIVE labels and conduct an automated audit of numeric agreement. Accessible chart context shifted Gemini and GPT toward DIRECT claims and improved numeric agreement for some models. Adding the image to the full context did not yield a consistent numeric benefit, and the withheld-context prompt did not reliably increase cautious language. The prompt-defined Real-World Significance section remained predominantly SPECULATIVE. These results motivate accessible description systems that distinguish claims supported by supplied evidence from model-supplied interpretation
Chinese Translation
多模态大语言模型（MLLMs）能够将可视化模式与外部原因、后果及领域知识联系起来，但这些解释的证据基础往往不明确。我们呈现了一项探索性研究，分析了来自四个来源的102个可视化图像、三个MLLM和四种输入条件，这些条件改变了对图像的访问、特定来源的可访问图表上下文以及隐含上下文的框架。在1224个描述中，我们分析了模型归属的直接（DIRECT）、派生（DERIVED）和推测（SPECULATIVE）标签，并进行了数值一致性的自动审计。可访问的图表上下文使Gemini和GPT更倾向于直接声明，并改善了某些模型的数值一致性。将图像添加到完整上下文并未带来一致的数值利益，而隐含上下文的提示并未可靠地增加谨慎语言。提示定义的现实世界重要性部分仍然主要是推测性的。这些结果激励我们开发可区分由提供证据支持的声明与模型提供解释的可访问描述系统。
cs.AI / 32 / 2607.25042
SAFAARI: Schema-Aware Framework for Accelerated Advertiser Response Intelligence
SAFAARI：加速广告主响应智能的模式感知框架
Abstract
The evolution of customer support systems is rapidly advancing with agentic chatbots, yet these systems face significant limitations when accessing enterprise data without predefined API endpoints. This paper presents SAFAARI (Schema-Aware Framework for Accelerated Advertiser Response Intelligence), a multi-agent framework that addresses the critical bottleneck of schema linking in Natural Language to SQL (NL-to-SQL) systems through specialized content, metadata, and orchestration agents. We also introduce SEAL (Schema Evaluation and Accuracy in Language-to-SQL), a novel composite metric that holistically evaluates system performance while penalizing inconsistent results. Through systematic experimentation with five feature set configurations, SAFAARI achieves an 81.66% SEAL score (6.65% improvement over baseline), with notable gains in datapoint accuracy (5.51%) and schema-linking precision (4.69%). The framework's effectiveness is validated through human-in-the-loop evaluation with domain experts, which proves its adaptability across diverse support domains. By automating the labor-intensive process of schema linking and query generation, our framework demonstrates 8x reduction in development time while maintaining high accuracy. The solution streamlines API development and enhances self-service capabilities, particularly benefiting customer support enterprises with complex data ecosystems.
Chinese Translation
客户支持系统的演变正在迅速发展，伴随着智能聊天机器人，但这些系统在没有预定义 API 端点的情况下访问企业数据时面临重大限制。本文提出了 SAFAARI（加速广告主响应智能的模式感知框架），这是一个多智能体框架，通过专门的内容、元数据和编排代理解决自然语言到 SQL（NL-to-SQL）系统中模式链接的关键瓶颈。我们还引入了 SEAL（语言到 SQL 的模式评估与准确性），这是一种新颖的复合指标，全面评估系统性能，同时惩罚不一致的结果。通过对五种特征集配置进行系统实验，SAFAARI 实现了 81.66% 的 SEAL 分数（比基线提高 6.65%），在数据点准确性（提高 5.51%）和模式链接精度（提高 4.69%）方面也取得了显著提升。该框架的有效性通过与领域专家的人工评估得到了验证，证明其在多样化支持领域的适应性。通过自动化模式链接和查询生成这一劳动密集型过程，我们的框架在保持高准确性的同时将开发时间减少了 8 倍。该解决方案简化了 API 开发，并增强了自助服务能力，特别有利于拥有复杂数据生态系统的客户支持企业。
cs.AI / 33 / 2607.25045
CogEEGAgent: Toward Autonomous Cognitive EEG Analysis with Grounded Execution and Selection-Aware Verification
CogEEGAgent：朝着具有基础执行和选择意识验证的自主认知EEG分析迈进
Abstract
Electroencephalography (EEG) analysis in cognitive studies requires specialized expertise and involves many defensible choices over contrasts, channels, time windows, and statistical tests. LLM agents can translate varied natural-language questions into analysis choices, offering a flexible interface for automation. Yet fluent reports alone cannot establish that an agent selected the requested analysis or evaluated a confirmatory claim independently of adaptive search. We present CogEEGAgent, a cognitive-EEG analysis agent grounded in MNE-Python. Its EEG-specific scientific harness separates semantic from scientific authority. The LLM interprets intent and proposes registered analyses, while deterministic components validate typed contracts, control confirmation access, and authorize evidence-bound release. On a prespecified routing benchmark, CogEEGAgent maps language to registered analyses more accurately than a matched deterministic router, while matched preflight makes both systems abstain whenever required. In an externally model-authored, outcome-blind campaign, the complete system releases supported analyses with participant-disjoint confirmation and blocks prespecified capability hazards and lifecycle-reuse requests. Policy stress testing shows that held-out confirmation curbs false positives from uncorrected adaptive search. Together, these studies establish bounded autonomy and an auditable automation framework for cognitive-EEG workflows. More broadly, they show how scientific agents can combine flexible language understanding with fail-closed control over inference and release.
Chinese Translation
在认知研究中，脑电图（EEG）分析需要专业的知识，并涉及对对比、通道、时间窗口和统计测试的多种合理选择。大型语言模型（LLM）代理能够将多样的自然语言问题转化为分析选择，提供灵活的自动化接口。然而，仅仅流畅的报告无法证明代理选择了请求的分析或独立于自适应搜索评估确认性主张。我们提出了CogEEGAgent，这是一个基于MNE-Python的认知EEG分析代理。其EEG特定的科学框架将语义与科学权威分开。LLM解读意图并提出注册分析，而确定性组件则验证输入的合同，控制确认访问，并授权基于证据的发布。在预先指定的路由基准上，CogEEGAgent将语言映射到注册分析的准确性超过了匹配的确定性路由器，而匹配的预飞行使得两个系统在必要时均可保持不动。在一个外部模型编写、结果盲目的活动中，完整系统发布支持的分析，确保参与者之间的确认不重叠，并阻止预先指定的能力风险和生命周期重用请求。政策压力测试表明，保留的确认可以抑制来自未校正自适应搜索的假阳性。综合来看，这些研究确立了有限的自主性和可审计的自动化框架，以支持认知EEG工作流程。更广泛地说，它们展示了科学代理如何将灵活的语言理解与对推理和发布的闭环控制相结合。
cs.AI / 34 / 2607.25057
Psychological Influences of Conversational AI: Research and Design Directions for Reducing Harm and Promoting Well-Being
对话式人工智能的心理影响：减少危害和促进福祉的研究与设计方向
Abstract
As conversational AI systems become increasingly integrated into daily life, their potential effects on user well-being require ongoing attention. While consumer-facing generalist models can provide benefits, including improved access to information, learning, productivity, self-reflection, and companionship, they also introduce risks, such as emotional entanglement, unhealthy dependence, and the amplification of psychological vulnerabilities. Drawing on prior research and empirical observations of AI chatbot behavior, we propose a set of aspirational directions for guiding the behavior of general-purpose AI systems in ways that may reduce potential psychological harms and support user well-being. We acknowledge the difficulty of systematically assessing the long-term impacts of AI chatbot use and frame these directions as hypotheses for studying how AI behavior may influence users across general interactions, role-playing scenarios, and contexts that could be characterized as providing psychological support. While some proposed directions are supported by existing research and expert insights, others identify open questions and areas requiring deeper study. We hope that this formulation and these hypotheses encourage further discussion, empirical investigation, and exploration of interactive design approaches aimed at better accommodating users' psychological needs and promoting their well-being.
Chinese Translation
随着对话式人工智能系统日益融入日常生活，其对用户福祉的潜在影响需要持续关注。虽然面向消费者的通用模型可以提供诸多好处，包括改善信息获取、学习、生产力、自我反思和陪伴，但它们也带来了风险，如情感纠缠、不健康的依赖以及心理脆弱性的放大。基于以往研究和对人工智能聊天机器人行为的实证观察，我们提出了一系列理想化的方向，以指导通用人工智能系统的行为，从而可能减少潜在的心理危害并支持用户福祉。我们承认系统评估人工智能聊天机器人使用的长期影响的困难，并将这些方向框架化为假设，以研究人工智能行为如何影响用户在一般交互、角色扮演场景以及可以被视为提供心理支持的背景下的表现。虽然一些提出的方向得到了现有研究和专家见解的支持，但其他方向则识别出开放性问题和需要深入研究的领域。我们希望这一表述和这些假设能够鼓励进一步讨论、实证研究以及探索旨在更好地满足用户心理需求和促进其福祉的互动设计方法。
cs.AI / 35 / 2607.25063
Similar Models Learn Differently: Final-Window Pretraining Shapes Post-Training Beyond SFT
相似模型的学习方式不同：最终窗口预训练塑造了超越SFT的后训练表现
Abstract
Developers judge a model checkpoint by how it behaves. After supervised fine-tuning (SFT), two checkpoints that perform about the same across relevant benchmarks are treated as interchangeable, equally ready for the next alignment stage, typically preference optimization. We ask whether this judgment misses a pretraining imprint: a difference that no post-SFT benchmark reveals, yet that decides how each checkpoint responds to further training. To find out, we run a controlled experiment on the final window of pretraining, the last data trained on before instruction tuning. Six branches fork from one partially pretrained checkpoint and differ only in this window: 500 million tokens, 0.1% to 1% of the tokens that precede it. Each branch trains its window on a single data source: generic web text, filtered web text, normative discourse, safety text, mathematical text, or synthetic educational text. SFT and post-training are then identical. After SFT the branches behave near-identically, within about one point on instruction following, refusal, and capability, yet the same post-training carries them to very different endpoints, under both a direct preference optimization update and a reinforcement learning update with a verifiable reward. We measure this deviation through refusal of harmful requests: when post-training begins the safety text branch refuses no more than the web text branch, yet by the end it has lost far less of its refusal. The other four branches gain little or no protection, so the effect is selective to what the window contained. The protection requires the safety text to arrive last rather than earlier in pretraining, and it reproduces on a second model family. What a model is pretrained on last shapes how it reacts to alignment. Therefore, a checkpoint should not be evaluated by its post-SFT behavior alone, and what it was trained on last should be reported with it.
Chinese Translation
开发者通过模型检查点的行为来评判其性能。在监督微调（SFT）之后，两个在相关基准测试中表现相近的检查点被视为可互换，均准备好进入下一个对齐阶段，通常是偏好优化。我们提出这个判断是否忽略了预训练的印记：一种后SFT基准无法揭示的差异，但却决定了每个检查点如何响应进一步的训练。为此，我们在预训练的最后窗口上进行了一项受控实验，即在指令调优之前训练的最后数据。六个分支从一个部分预训练的检查点分叉，仅在这个窗口上有所不同：500百万个标记，占其之前标记的0.1%到1%。每个分支在单一数据源上训练其窗口：通用网络文本、过滤后的网络文本、规范性话语、安全文本、数学文本或合成教育文本。SFT和后训练过程是相同的。在SFT之后，各分支的行为几乎相同，在指令遵循、拒绝和能力方面的差异不超过一个点，然而相同的后训练却将它们带到了非常不同的终点，无论是通过直接的偏好优化更新，还是通过具有可验证奖励的强化学习更新。我们通过拒绝有害请求来测量这种偏离：当后训练开始时，安全文本分支的拒绝率不超过网络文本分支，但到最后，它的拒绝率损失远低于后者。其他四个分支几乎没有获得保护，因此这一效应是选择性的，取决于窗口所包含的内容。保护要求安全文本在预训练中最后到达，而不是更早到达，并且这一现象在第二个模型系列中得到了重现。模型最后预训练的内容塑造了其对对齐的反应。因此，检查点不应仅通过其后SFT行为来评估，最后训练的内容也应与之报告。
cs.AI / 36 / 2607.25066
Addressable Recall Compaction for Long Context-Window Control in AI Agents
可寻址回忆压缩用于人工智能代理的长上下文窗口控制
Abstract
Long-horizon LLM agents accumulate reasoning traces, actions, and tool observations that can eventually exceed a model's fixed context window. Existing compaction methods address this limitation by discarding, summarizing, or retrieving earlier information, but they may remove task-critical details or fail to recover them reliably. We propose ARC (Addressable Recall Compaction), a context-management framework that separates archival storage from active-context presentation. ARC stores tool observations in an append-only, ID-addressable log and replaces older observations with compact citations when compaction is required. The agent can subsequently use these identifiers to request stored content without re-executing the corresponding tools or depending solely on similarity-based retrieval. We evaluate ARC using Qwen3-8B with a 16k context window and Qwen3-32B with a 32k context window. On the Needle-in-a-Haystack evaluation, ARC achieves an average exact-answer accuracy of 99.40%, compared with 88.12% for the best-performing baseline in our evaluation. ARC also reduces estimated serving time and HBM traffic under our hardware-cost model. On the LongBench-v2 Hard subset, ARC obtains an average accuracy of 29.97%, compared with 28.25% for the best-performing baseline. These results indicate that explicit, address-based recall can improve information retention and serving efficiency relative to the evaluated context-management baselines under the tested settings.
Chinese Translation
长时间跨度的LLM代理积累推理痕迹、动作和工具观察，这些信息最终可能超出模型的固定上下文窗口。现有的压缩方法通过丢弃、总结或检索早期信息来解决这一限制，但它们可能会移除任务关键细节或无法可靠地恢复这些细节。我们提出了ARC（可寻址回忆压缩），这是一种上下文管理框架，旨在将归档存储与活动上下文展示分开。ARC将工具观察存储在一个仅附加的、可通过ID寻址的日志中，并在需要压缩时用紧凑的引用替换较旧的观察。代理随后可以使用这些标识符请求存储的内容，而无需重新执行相应的工具或仅依赖基于相似性的检索。我们使用Qwen3-8B（16k上下文窗口）和Qwen3-32B（32k上下文窗口）对ARC进行了评估。在“针在干草堆中”评估中，ARC的平均准确答案率为99.40%，而我们评估中表现最佳的基线为88.12%。在我们的硬件成本模型下，ARC还减少了估计的服务时间和HBM流量。在LongBench-v2 Hard子集上，ARC的平均准确率为29.97%，而表现最佳的基线为28.25%。这些结果表明，在测试设置下，显式的基于地址的回忆可以相对于评估的上下文管理基线提高信息保留和服务效率。
cs.AI / 37 / 2607.25068
How Often Should a Recommender Call an LLM? Value-Weighted Routing, Monitoring, and Seasonal Robustness
推荐系统应多频繁调用大型语言模型（LLM）？价值加权路由、监控与季节性鲁棒性
Abstract
Routing decisions between a cheap heuristic and an expensive large language model (LLM) are typically framed as a difficulty problem: send the hard cases to the expensive path. We argue this framing is incomplete because difficulty and business value are distinct axes - a difficult cheap item and a difficult costly item do not have the same cost of error. We present Value Router, a fully synthetic simulation of a retail merchandising pipeline that routes items using only estimated difficulty and estimated value, never ground truth. The study has three stages. First, a value-weighted threshold router is compared with a difficulty-only and a random baseline on a synthetic catalog with an inverse correlation between category volume and value. Value-weighting matches the difficulty-only baseline's recall of true high-value items (60%) while achieving substantially higher precision (98.3% vs. 94.3%). Second, a decision logger and monitor expose a failure mode hidden by aggregate metrics showing that the aggregate result is driven almost entirely by between-category differences rather than per-item discrimination. Third, a simulated Black Friday demand surge (2.5 volume with a shift toward higher-value categories) compares a static router, a seasonally tuned router, and two slow-path budget policies. All results are from a controlled synthetic simulation with experimenter-defined ground truth and illustrate design principles for cost-aware routing systems rather than validated real-world claims.
Chinese Translation
在廉价启发式算法与昂贵的大型语言模型（LLM）之间的路由决策通常被视为一个难度问题：将困难案例发送到昂贵的路径。我们认为这种框架是不完整的，因为难度和商业价值是两个不同的维度——一个困难的廉价项目和一个困难的昂贵项目的错误成本并不相同。我们提出了价值路由器（Value Router），这是一个完全合成的零售商品管道模拟，使用估计的难度和估计的价值进行项目路由，而从不依赖真实数据。研究分为三个阶段。首先，在一个类别量与价值呈反相关的合成目录中，比较了价值加权阈值路由器与仅基于难度和随机基线。价值加权在真实高价值项目的召回率（60%）上与仅基于难度的基线相匹配，同时实现了显著更高的精确度（98.3% 对比 94.3%）。其次，决策记录器和监控工具揭示了一种被聚合指标隐藏的失败模式，显示聚合结果几乎完全由类别间差异驱动，而非逐项区分。第三，模拟的黑色星期五需求激增（2.5倍的量，且向高价值类别转移）比较了静态路由器、季节性调优路由器和两种慢路径预算政策。所有结果均来自一个受控的合成模拟，具有实验者定义的真实数据，展示了面向成本的路由系统的设计原则，而非经过验证的现实世界主张。
cs.AI / 38 / 2607.25076
Towards an Agent Operating System - Lessons from Classical and Cloud OS
迈向智能体操作系统 - 来自经典操作系统和云操作系统的经验教训
Abstract
Every major wave of platform software follows the same arc: an initial period of experimentation with competing frameworks and ad-hoc implementations, followed by the articulation of a small set of stable abstractions with well-defined semantics, and finally consolidation around those abstractions into a platform that applications can portably target. POSIX did this for classical operating systems; Kubernetes did it for the cloud. Agentic AI systems - autonomous, LLM-driven agents that plan, use tools, maintain memory, and collaborate - are currently in the experimentation phase of the third such wave. dozens of frameworks and protocols have emerged, but no community consensus exists on what the core abstractions are or what guarantees they carry. Without that consensus, agentic applications cannot be written portably, platforms cannot compose reliably, and the field cannot advance beyond prototype deployments. We argue that the path forward is to follow the prior-wave methodology: derive new agentic abstractions by extending classical OS and cloud OS primitives to stochastic, natural-language-mediated execution, specify their semantics precisely, and consolidate around them - just as POSIX and Kubernetes consolidated their respective waves.
Chinese Translation
每一波主要的平台软件都遵循相同的轨迹：最初是对竞争框架和临时实现的实验阶段，随后是对一小组稳定抽象的明确阐述，这些抽象具有明确的语义，最后围绕这些抽象的整合形成一个应用程序可以便携目标的平台。POSIX 为经典操作系统做到了这一点；Kubernetes 为云计算做到了这一点。智能体人工智能系统——自主的、基于大型语言模型（LLM）的智能体，能够进行计划、使用工具、维护记忆和协作——目前正处于第三波实验阶段。已经出现了数十种框架和协议，但对于核心抽象是什么或它们所承载的保证尚无社区共识。没有这种共识，智能体应用程序无法便携编写，平台无法可靠组合，领域也无法超越原型部署。我们认为，前进的道路是遵循前一波的方法论：通过将经典操作系统和云操作系统的原语扩展到随机、自然语言介导的执行，推导出新的智能体抽象，精确指定其语义，并围绕它们进行整合——就像 POSIX 和 Kubernetes 整合各自的波动一样。
cs.AI / 39 / 2607.25082
PLATO: Pointer Learner for Agent and Task Openness
PLATO：用于代理和任务开放性的指针学习器
Abstract
Open agent systems (OASYS) are increasingly prevalent in real-world domains where the sets of agents and tasks change unpredictably over time. Such openness, including agent openness (AO) and task openness (TO), poses a fundamental challenge to multi-agent reinforcement learning (MARL), which typically assumes fixed state and action spaces. Existing methods address openness only partially: padding and masking approaches introduce artificial bounds, while recent graph-based or hypergraph methods handle one dimension of openness but still depend on restrictive assumptions. In this paper, we introduce Pointer Learner for Agent and Task Openness (PLATO), a pointer-network-based actor combined with a centralized graph neural network (GNN) critic, trained with multi-agent proximal policy optimization under a centralized training and decentralized execution paradigm. Our pointer-based actor outputs distributions directly over the current task set. This directly supports changing action spaces without masking or retraining. Our GNN critic encodes agent-task interactions as a graph that changes shape with task and agent composition. Together, these components consider AO and TO without the boundedness of existing approaches. We formalize PLATO in a Task-and-Agent-Open Markov Game (TaAgO-MG), extending prior task-open formulations, and prove it is well-defined over the resulting unbounded state and action spaces. We evaluate PLATO with the Methods for Open Agent Systems Evaluation Initiative (MOASEI) wildfire suppression domain, an environment designed for open multi-agent system evaluation, and we demonstrate strong performance and more consistent zero-shot generalization than state-of-the-art baselines in OASYS.
Chinese Translation
开放代理系统（OASYS）在现实世界中越来越普遍，其中代理和任务的集合随时间不可预测地变化。这种开放性，包括代理开放性（AO）和任务开放性（TO），对多智能体强化学习（MARL）提出了根本性挑战，因为MARL通常假设状态和动作空间是固定的。现有方法仅部分解决了开放性问题：填充和掩蔽方法引入了人工界限，而最近的基于图或超图的方法处理了开放性的一个维度，但仍然依赖于限制性假设。在本文中，我们提出了用于代理和任务开放性的指针学习器（PLATO），它是一个基于指针网络的演员，结合了一个集中式图神经网络（GNN）评论员，在集中训练和分散执行的范式下，通过多智能体近端策略优化进行训练。我们的基于指针的演员直接输出当前任务集上的分布。这直接支持在不进行掩蔽或重新训练的情况下改变动作空间。我们的GNN评论员将代理-任务交互编码为一个随着任务和代理组合而变化形状的图。这些组件共同考虑了AO和TO，而不受现有方法的界限限制。我们在任务和代理开放马尔可夫博弈（TaAgO-MG）中形式化PLATO，扩展了先前的任务开放性表述，并证明它在结果的无界状态和动作空间上是良定义的。我们在开放多智能体系统评估设计的环境——火灾扑灭领域（MOASEI）中评估PLATO，展示了其强大的性能和比现有最先进基线更一致的零-shot泛化能力。
cs.AI / 40 / 2607.25090
Matryoshka Agent: Unfolding Sub-Agents for Long-Horizon Machine Learning Engineering
马特里奥什卡智能体：展开子智能体以实现长远机器学习工程
Abstract
Machine learning engineering (MLE) tasks require long-horizon decision making over iterative solution debugging and refinement, under expensive and feedback-driven environment interactions. Developing and training a monolithic agent for such tasks is fundamentally challenging, as it must simultaneously manage extremely long and noisy contexts, explore vast solution spaces, and remain effective under limited model capacity and computational budgets. To address these challenges, we propose Matryoshka Agent, a unified hierarchical agent framework for complex long-horizon tasks. Matryoshka Agent decomposes agentic problem solving into a coordinated hierarchy of decision making and execution: a high-level Orchestrator maintains compact, long-horizon exploration states and issues strategic instructions, while lower-level Sub-Agents execute concrete solution attempts through direct environment interaction, mediated by standardized Tool interface. This design decouples strategic exploration from costly execution, substantially reducing the burden of long-context reasoning and enabling efficient iterative refinement. We further develop an efficient training paradigm for Matryoshka Agent. Experimental results on a broad range of MLE tasks with diverse model types and scales demonstrate that Matryoshka Agent is an effective and scalable paradigm for long-horizon MLE tasks and complex agentic problem solving. Notably, Matryoshka Agent enables Qwen3-4B-Instruct to reach Orchestrator performance comparable to o4-mini. Applying Matryoshka Agent to Qwen3-30B-Coder results in at most 36.7% relative performance gain.
Chinese Translation
机器学习工程（MLE）任务需要在昂贵且以反馈驱动的环境交互下进行长远决策，涉及迭代解决方案的调试和优化。为此类任务开发和训练一个单一的智能体是根本性的挑战，因为它必须同时管理极长且嘈杂的上下文，探索广阔的解决方案空间，并在有限的模型容量和计算预算下保持有效。为了解决这些挑战，我们提出了马特里奥什卡智能体（Matryoshka Agent），这是一个针对复杂长远任务的统一层次智能体框架。马特里奥什卡智能体将智能问题解决分解为一个协调的决策和执行层次结构：高层的协调者（Orchestrator）维护紧凑的长远探索状态并发出战略指令，而低层的子智能体（Sub-Agents）通过标准化工具接口直接与环境交互，执行具体的解决方案尝试。这一设计将战略探索与高成本执行解耦，大大减少了长上下文推理的负担，并实现了高效的迭代优化。我们进一步为马特里奥什卡智能体开发了一种高效的训练范式。在广泛的MLE任务上进行的实验结果显示，马特里奥什卡智能体是一个有效且可扩展的长远MLE任务和复杂智能问题解决的范式。值得注意的是，马特里奥什卡智能体使Qwen3-4B-Instruct的协调者性能达到了与o4-mini相当的水平。将马特里奥什卡智能体应用于Qwen3-30B-Coder，最多实现了36.7%的相对性能提升。
cs.AI / 41 / 2607.25091
Towards Robust Reinforcement Learning for Small-Scale Language Model Agents
面向小规模语言模型代理的稳健强化学习
Abstract
The alignment of Small Language Models (SLMs) in the 70--500M parameter range using reinforcement learning is often considered unstable, though the underlying failure mechanisms have not been systematically investigated. In the State-of-the-Art (SOTA) research, fifteen (model, corpus) configurations were trained using Proximal Policy Optimization (PPO). The experiments included Pythia-70M, 160M, 410M and SmolLM2-135M, 360M on the TinyStories, CNN/DailyMail, and Wikitext-103 corpora. Three reproducible failure modes were identified in small-scale language models: silent LoRA parameter freezing in standard PEFT/TRL pipelines, numerical overflow in importance ratios when using bfloat16, and catastrophic policy collapse due to reward-model error. These issues were addressed using a merge-and-reinitialize adapter technique, float32 precision during PPO updates, and a three-layer safety mechanism comprising reward whitening, importance-ratio guarding, and weight rollback. In this paper, a capacity-headroom hypothesis is proposed, which states that PPO performance at the SLM scale depends on both a fluent supervised model ($\text{PPL}<20$) and a discriminative reward signal, rather than on the number of model parameters. The proposed system converged stably in all experiments and improved preference win rate over the SFT baseline in configurations with a fluent prior and an informative reward signal. Furthermore, it outperformed instruction-tuned baselines while requiring significantly less training data. All checkpoints, preference datasets, and training scripts are publicly released$^{\S}$.
Chinese Translation
在70到500M参数范围内，小型语言模型（SLMs）的对齐通常被认为是不稳定的，尽管其潜在的失败机制尚未系统性地研究。在最先进的（SOTA）研究中，使用近端策略优化（PPO）训练了十五个（模型，语料库）配置。实验包括在TinyStories、CNN/DailyMail和Wikitext-103语料库上训练Pythia-70M、160M、410M和SmolLM2-135M、360M。我们在小规模语言模型中识别出了三种可重复的失败模式：在标准PEFT/TRL管道中静默的LoRA参数冻结、使用bfloat16时重要性比率的数值溢出，以及由于奖励模型错误导致的灾难性策略崩溃。针对这些问题，采用了合并与重新初始化适配器技术、在PPO更新期间使用float32精度，以及包含奖励去白化、重要性比率保护和权重回滚的三层安全机制。本文提出了一种容量余量假说，认为SLM规模下PPO的性能依赖于流畅的监督模型（$ ext{PPL}<20$）和判别性奖励信号，而不是模型参数的数量。所提出的系统在所有实验中稳定收敛，并在具有流畅先验和信息性奖励信号的配置中提高了相较于SFT基线的偏好胜率。此外，它在需要显著更少训练数据的情况下超越了指令调优基线。所有检查点、偏好数据集和训练脚本均已公开发布。
cs.AI / 42 / 2607.25135
ScalableRAG: High-Quality RAG at Zero Ingestion Cost
可扩展的RAG：零摄取成本下的高质量RAG
Abstract
Recent advances in RAG aim to optimize for performance by paying high ingestion costs for knowledge ingestion: building knowledge graphs or extracting SQL tables. In this work we show that the operations that such knowledge bases allow can be replicated with zero ingestion costs (not even a vector database); in fact our solution, Zero-Ingestion ScalableRAG, handily out-performs all baselines (including knowledge graph approaches) in three out of the six corpora considered here, and only marginally missing maximum performance on the other three, with average accuracy across all six datasets 7.36% above the next most competitive baseline. It achieves this by keeping a workspace of document sets and values sets that it can write into and read from, allowing for on-the-fly aggregative reasoning in all situations where grouping is required on a primary key that is in one to one correspondence with a subset of the total document set. Capping the number of LLM calls by a constant independent of the corpus size, we also introduce Limited-Ingestion ScalableRAG, which does use a minimal vector database as well as an automated pattern discovery from a sample of documents, to further improve accuracy at scale. Our code is available at https://github.com/cohesity/ScalableRAG .
Chinese Translation
最近在RAG领域的进展旨在通过支付高昂的摄取成本来优化性能，以进行知识摄取：构建知识图谱或提取SQL表。在本研究中，我们展示了此类知识库所允许的操作可以以零摄取成本（甚至不需要向量数据库）进行复制；实际上，我们的解决方案Zero-Ingestion ScalableRAG在这里考虑的六个语料库中的三个上显著超越了所有基线（包括知识图谱方法），而在其他三个语料库上仅略微未达到最高性能，六个数据集的平均准确率比下一个最具竞争力的基线高出7.36%。它通过保持一个文档集和数值集的工作空间，使其能够进行读写，从而在所有需要基于与总文档集的子集一一对应的主键进行分组的情况下实现即时聚合推理。通过将LLM调用的数量限制为与语料库大小无关的常数，我们还引入了Limited-Ingestion ScalableRAG，它确实使用了一个最小的向量数据库以及从文档样本中自动发现模式，以进一步提高大规模下的准确性。我们的代码可在https://github.com/cohesity/ScalableRAG获取。
cs.AI / 43 / 2607.25136
Less Data, Better Alignment: Data-Centric Multi-Evaluator Agreement for Preference Optimization
更少的数据，更好的对齐：以数据为中心的多评估者一致性用于偏好优化
Abstract
Research on preference optimization often varies the training objective while holding the data fixed. We instead ask whether a small, high-confidence set of on-policy responses can provide a reliable learning signal. Our method, DMAPO (Data-centric Multi-evaluator Agreement for Preference Optimization), generates candidate responses from the target policy, evaluates helpfulness, factuality, and conciseness with rubric-specialized evaluators, applies a process-critic correction, and retains only high-consensus desirable or undesirable examples. This procedure accepts 1,871 of 54,236 Mistral-7B candidates (3.45%). KTO trained on this set reaches 7.50 on MT-Bench, 95.5% length-controlled win rate against a text-davinci-003 reference, and 57.3% IFEval prompt accuracy. Independent pairwise evaluation also favors DMAPO over SimPO: GPT-4o yields a net win rate of 23.3 points on 129 held-out prompts and 24.0 points on 200 out-of-distribution LMSYS-Chat prompts; Claude Opus 4.7 yields 24.1 points on the held-out set. Changing the evaluator model or rubric alters the selected examples but has little effect on downstream performance. A second-backbone study yields a similar 3.41% acceptance rate, although its performance gains are more modest. Across these experiments, consensus filtering offers a data-efficient route to preference optimization for general instructions, at the cost of additional curation compute and dependence on evaluator judgments.
Chinese Translation
偏好优化的研究通常在固定数据的情况下变化训练目标。我们则探讨一小组高置信度的在政策响应是否能够提供可靠的学习信号。我们的方法，DMAPO（以数据为中心的多评估者一致性用于偏好优化），从目标政策生成候选响应，通过专门评估标准评估其有用性、事实性和简洁性，应用过程批评校正，并仅保留高度一致的期望或不期望示例。该过程接受了54,236个Mistral-7B候选中的1,871个（3.45%）。基于该数据集训练的KTO在MT-Bench上达到了7.50，与text-davinci-003参考模型相比，长度控制的胜率为95.5%，IFEval提示的准确率为57.3%。独立的成对评估也更倾向于DMAPO而非SimPO：GPT-4o在129个保留提示上获得了23.3的净胜率，在200个分布外的LMSYS-Chat提示上获得了24.0的净胜率；Claude Opus 4.7在保留集上获得了24.1的净胜率。更换评估模型或评估标准会改变所选示例，但对下游性能影响不大。第二个主干研究的接受率也达到了类似的3.41%，尽管其性能提升较为温和。在这些实验中，一致性过滤为一般指令的偏好优化提供了一条数据高效的途径，但代价是需要额外的策划计算和对评估者判断的依赖。
cs.AI / 44 / 2607.25140
How Affect Propagates among LLM Agents: Emergent Emotional Contagion in Crowd Simulation
情感在大型语言模型代理之间的传播：人群模拟中的情感传染现象
Abstract
This paper studies the behavior of language models in a multi-agent crowd simulation, focusing on how affect propagates among agents that perceive and appraise one another. Each agent perceives its neighbors through visual, auditory, and tactile channels, then appraises these perceptions in light of its prompted personality profile, memory, current affective state, and situational context. Appraisal is carried out by an LLM, which updates the agent's internal affective state and selects its outward expression. The architecture contains no hand-authored mechanism for directly transferring affective state between agents; instead, inter-agent influence arises through the perception-appraisal-expression loop. The agent representation draws on the Big Five personality model and Russell's circumplex model of affect. To limit latency, low-level steering and navigation are handled by a conventional crowd simulator operating independently of the LLM-based cognitive layer. We evaluate the architecture across five scenario environments spanning alarming, joyful, and neutral situations in different spatial layouts. The results show that the system produces emotional contagion dynamics with spatial, temporal, and personality-dependent structure in sparse, small crowds. Alarm spreads from seeded agents as a traveling front, the mean alarmed fraction settles at a nonzero plateau, and the distribution of prompted personality profiles determines whether an ambiguous alarm ignites panic and whether a provocation is interpreted as anger or fear. We further evaluate the appraisal step through controlled experiments across prompt variants, sampling temperatures, and four model backends, showing that the dynamics are backend-dependent.
Chinese Translation
本文研究了语言模型在多代理人群模拟中的行为，重点关注情感如何在彼此感知和评估的代理人之间传播。每个代理通过视觉、听觉和触觉通道感知其邻近代理，然后根据其提示的个性特征、记忆、当前情感状态和情境背景对这些感知进行评估。评估由大型语言模型（LLM）执行，该模型更新代理的内部情感状态并选择其外部表达。该架构没有手动编写的机制来直接在代理之间转移情感状态；相反，代理之间的影响通过感知-评估-表达循环产生。代理的表示基于五大人格模型（Big Five）和拉塞尔的情感圆周模型（Russell's circumplex model of affect）。为了限制延迟，低级的引导和导航由一个独立于LLM认知层的传统人群模拟器处理。我们在五种场景环境中评估该架构，这些环境涵盖了不同空间布局下的惊恐、快乐和中性情境。结果表明，该系统在稀疏的小型人群中产生了具有空间、时间和个性依赖结构的情感传染动态。惊恐从种子代理人处以旅行前沿的形式传播，平均惊恐比例稳定在非零平台上，而提示的个性特征的分布决定了模糊的惊恐是否引发恐慌，以及挑衅是否被解读为愤怒或恐惧。我们进一步通过对提示变体、采样温度和四个模型后端的控制实验评估评估步骤，显示出动态依赖于后端。
cs.AI / 45 / 2607.25147
Inferring Missing Trajectory Data with Temporal Convolutional Networks
利用时间卷积网络推断缺失的轨迹数据
Abstract
Trajectory data collected in real-world settings is frequently incomplete due to sensor failure, communication loss, or occlusion. We address the task of \emph{trajectory inpainting}: reconstructing contiguous missing segments from observed context. We propose a Temporal Convolutional Network (TCN) with symmetric dilation that relaxes the standard causality constraint, allowing each time step to draw on both past and future observations, a property that is essential for inpainting, but absent from forecasting-oriented architectures. The model is trained with a composite loss that combines weighted mean squared error, boundary--continuity penalties, and a smoothness regularizer. Trained on a synthetic dataset of $1,000$ (train), $200$ (validation), and $300$ (test) two-dimensional trajectories with randomly placed 20% masked segments, the model achieves good R$^{2}$, MSE and MAE metrics.
Chinese Translation
在实际环境中收集的轨迹数据常常由于传感器故障、通信丢失或遮挡而不完整。我们针对 extit{轨迹修复}任务进行研究：从观察到的上下文中重建连续的缺失段。我们提出了一种具有对称膨胀的时间卷积网络（Temporal Convolutional Network, TCN），该网络放宽了标准的因果约束，允许每个时间步同时利用过去和未来的观测，这一特性对于修复至关重要，但在以预测为导向的架构中并不存在。该模型采用复合损失进行训练，结合了加权均方误差、边界连续性惩罚和光滑性正则化。模型在一个合成数据集上进行训练，该数据集包含$1,000$个（训练集）、$200$个（验证集）和$300$个（测试集）随机放置20%遮蔽段的二维轨迹，模型在R$^{2}$、均方误差（MSE）和平均绝对误差（MAE）指标上表现良好。
cs.AI / 46 / 2607.25152
When Do Agent Loops Mistake Stagnation for Progress? Self-Evaluation Bias and Externally Grounded Verification in Long-Running Autonomous LLM Agent Loops
代理循环何时将停滞误认为进展？自我评估偏差与长时间运行的自主大型语言模型代理循环中的外部验证
Abstract
Long-running autonomous agents plan, act, and judge their own completion without human intervention. When an agent grades its own work, self-evaluation bias takes hold: plausible changes are accepted as progress while real-world outcomes stagnate or regress. We name this failure mode the progress mirage and show, with controlled measurement, that it is a question of what the evaluator is grounded in. We built a testbed that holds the agent and its tool surface fixed and manipulates only the information-channel type of the evaluator that gates the loop. A world-state oracle, unfakeable in principle, is enforced by container and network isolation and verified at every run. Across 54 cycles a frontier agent claimed improvement every time, yet 56 percent had a measured delta of zero or below. Self-report was thus uninformative, and the self-verdict gate degenerated into accept-all, eroding the best deployed state it had reached by 19 percent. Even the strongest in-band judge, reading the full artifact text, the change diff, and its own verdict history, accepted cycles of which 44 percent were real-world regressions and rejected 38 percent of real improvements; the preregistered adversarial hypothesis that a strong judge closes the gap was rejected. On a boundary task whose success specification is verifiable from the artifact itself, the same judge's mirage vanished to zero and the gap collapsed within the registered threshold, showing that the gap depends on where the success signal resides. A sign-only variant returning only the acceptance verdict kept real-world output similar to full feedback (110.0 versus 113.0), locating the benefit in the gate's grounding rather than in feedback content. For open-ended objectives whose success signal lives outside the transcript, scaling up the judge is not enough; out-of-band evaluation with real-world access is a structural requirement.
Chinese Translation
长时间运行的自主代理在没有人类干预的情况下进行规划、行动和判断其自身的完成。当代理对自己的工作进行评分时，自我评估偏差便会产生：合理的变化被视为进展，而现实世界的结果却停滞或倒退。我们将这种失败模式称为进展幻影，并通过控制测量表明，这与评估者所依据的基础有关。我们构建了一个测试平台，固定代理及其工具表面，仅操控评估者的信息通道类型来控制循环。一个世界状态神谕，原则上无法伪造，通过容器和网络隔离得以实施，并在每次运行时进行验证。在54个循环中，一个前沿代理每次都声称有所改善，但56%的测量增量为零或以下。因此，自我报告并未提供有用信息，自我裁决的门槛退化为接受所有，从而使其达到的最佳部署状态下降了19%。即使是最强的内部评判者，阅读完整的文档文本、变化差异及其自身的裁决历史，也接受了44%为现实世界的倒退，并拒绝了38%的真实改进；预注册的对抗假设，即强评判者能够缩小差距，被拒绝。在一个成功规范可以从文档本身验证的边界任务中，同一评判者的幻影消失为零，差距在注册阈值内崩溃，表明差距取决于成功信号所在的位置。一个仅返回接受裁决的标志变体，其现实世界输出与完整反馈相似（110.0对比113.0），将收益定位于门槛的基础而非反馈内容。对于成功信号存在于记录之外的开放式目标，仅仅扩大评判者的规模是不够的；与现实世界接入的带外评估是一个结构性要求。
cs.AI / 47 / 2607.25157
PreDiff-LM: Pretrained Discrete Masked Diffusion Language Modeling with Hybrid Attention
PreDiff-LM：具有混合注意力的预训练离散掩蔽扩散语言建模
Abstract
Discrete masked diffusion language models support bidirectional generation and infilling, but adapting pretrained autoregressive (AR) transformers requires reconciling causal pretraining with bidirectional denoising. We study this problem at the level of attention rather than claiming AR-weight reuse itself as novel. PreDiff-LM preserves causal attention within the observed prompt while allowing full bidirectional attention within the masked target. Under a matched GPT-2 Medium, WikiText-103, 90K-step setup, this hybrid mask improves unconditional perplexity from 34.1 to 28.7 and MAUVE from 0.71 to 0.78 over uniform bidirectional attention with the same AR initialization. Attention adaptation also composes with a DiffuGPT-style objective adaptation, reaching 26.9 perplexity. Pretrained initialization reduces the steps required to reach perplexity below 50 from about 350K to 8K, although a compute-matched fine-tuned AR model remains stronger at equal scale (18.9 versus 28.7). Beyond perplexity, PreDiff-LM improves repetition, distributional quality, four zero-shot downstream tasks, and human preference over prior diffusion baselines. The results position hybrid attention as a complementary mechanism for adapting pretrained causal backbones, while making explicit the remaining quality and inference-efficiency gaps to optimized AR models.
Chinese Translation
离散掩蔽扩散语言模型支持双向生成和填充，但将预训练的自回归（AR）变换器适配需要调和因果预训练与双向去噪之间的关系。我们从注意力的层面研究这个问题，而不是声称AR权重重用本身是新颖的。PreDiff-LM在观察到的提示中保留因果注意力，同时允许在掩蔽目标中实现完全的双向注意力。在匹配的GPT-2 Medium、WikiText-103、90K步设置下，这种混合掩蔽将无条件困惑度从34.1提高到28.7，MAUVE从0.71提高到0.78，相较于使用相同AR初始化的均匀双向注意力。注意力适配还与DiffuGPT风格的目标适配相结合，达到了26.9的困惑度。预训练初始化将将困惑度降至50以下所需的步数从大约350K减少到8K，尽管在相同规模下，计算匹配的微调AR模型仍然更强（18.9对比28.7）。除了困惑度，PreDiff-LM还改善了重复性、分布质量、四个零样本下游任务以及相较于先前扩散基线的人类偏好。结果表明，混合注意力作为适配预训练因果骨干的补充机制，同时明确了与优化AR模型之间剩余的质量和推理效率差距。
cs.AI / 48 / 2607.25166
Observing sycophantic AI validate others reduces its appeal but not its persuasiveness
观察谄媚型人工智能验证他人降低其吸引力但不影响其说服力
Abstract
AI chatbots can be ``sycophantic,'' or overly agreeable and flattering toward users. Sycophantic AI has been shown to entrench attitudes, yet users frequently fail to recognize it (a phenomenon we call ``sycophancy blindness''). We tested whether increasing users' awareness of sycophancy protects them from its harmful effects. In one preregistered experiment (n = 940), participants received a brief written warning about sycophancy before conversing with a sycophantic chatbot. In a second preregistered experiment (n = 650), participants watched a video of a sycophantic AI validating several other users, including users on opposite sides of the same conflict, before interacting with it themselves. Both interventions changed how participants evaluated the AI. The warning reduced the AI's perceived objectivity, and the video reduced enjoyment of the AI, an effect mediated by the reduced belief that its validation was uniquely earned. We then pooled our experiments with two prior studies of sycophancy awareness interventions (six interventions total, n = 3,982). The pattern was consistent: interventions made the sycophantic AI appear less objective and trustworthy, and none of the six reduced its persuasiveness. These results suggest that individual-level interventions, such as warning labels or AI literacy, may not be enough to protect users from AI harms.
Chinese Translation
人工智能聊天机器人可能表现出“谄媚”，即对用户过于赞同和恭维。已有研究表明，谄媚型人工智能会加深用户的态度，但用户常常未能识别这一现象（我们称之为“谄媚盲目”）。我们测试了提高用户对谄媚的意识是否能保护他们免受其有害影响。在一项预注册实验中（n = 940），参与者在与谄媚型聊天机器人对话之前，收到了关于谄媚的简要书面警告。在第二项预注册实验中（n = 650），参与者观看了一段谄媚型人工智能验证其他用户的视频，其中包括在同一冲突中持相反立场的用户，然后才与其互动。这两种干预措施改变了参与者对人工智能的评估。警告降低了人工智能的客观性感知，而视频则降低了对人工智能的享受，这一效果是通过减少对其验证是独特获得的信念来介导的。随后，我们将我们的实验与两项先前的关于谄媚意识干预的研究汇总（共六项干预，总样本量 n = 3,982）。结果模式一致：干预措施使谄媚型人工智能看起来不那么客观和可信，而六项干预中没有一项降低其说服力。这些结果表明，个体层面的干预措施，如警告标签或人工智能素养，可能不足以保护用户免受人工智能的危害。
cs.AI / 49 / 2607.25218
Everyone is unique: Towards Behaviorally Heterogeneous Negotiation Dialogue Systems for Debt Collection
每个人都是独特的：面向行为异质的债务催收对话系统
Abstract
Debt collection is a critical negotiation task in the financial industry, with strong practical relevance and exceptional academic value as a behaviorally rich, high-stakes testbed for human-centered dialogue systems. While large language models (LLMs) have shown promise in dialogue and negotiation, effectively evaluating their performance in this complex scenarios remains a major challenge: existing benchmarks uniformly assume users to be static, rational agents with fixed preferences, failing to capture the rich behavioral heterogeneity inherent in real-world debt collection. To bridge this gap, we propose DebtBench, the first public persona-enriched debt collection benchmark, that highlights behavioral heterogeneity in negotiation. Moreover, we develop DebtGPT, a debt collection agent trained to jointly optimize financial recovery and interaction experience. Our experimental results, using 16 state-of-the-art LLMs, find that most existing models struggle in this complex but realistic scenarios, whereas DebtGPT outperforms all open-source baselines and achieves performance on par with GPT-4o. The code and data are available at https://github.com/YYuHhhh/DebtNegotiation.
Chinese Translation
债务催收是金融行业中的一项关键谈判任务，具有重要的实际意义和卓越的学术价值，作为以人为中心的对话系统的行为丰富、高风险的测试平台。尽管大型语言模型（LLMs）在对话和谈判中展现出潜力，但在这一复杂场景中有效评估其性能仍然是一个重大挑战：现有基准普遍假设用户为静态、理性的代理，具有固定的偏好，未能捕捉到现实债务催收中固有的丰富行为异质性。为了解决这一问题，我们提出了DebtBench，这是第一个公共的丰富角色的债务催收基准，突出了谈判中的行为异质性。此外，我们开发了DebtGPT，这是一个债务催收代理，旨在共同优化财务回收和互动体验。我们的实验结果使用了16个最先进的LLMs，发现大多数现有模型在这一复杂但现实的场景中表现不佳，而DebtGPT超越了所有开源基准，并在性能上与GPT-4o持平。代码和数据可在 https://github.com/YYuHhhh/DebtNegotiation 获取。
cs.AI / 50 / 2607.25244
CADENCE: A Cardiac Atom Dictionary for Interpretable Neural Concept Extraction from ECG Foundation Models
CADENCE：用于从心电图基础模型中提取可解释神经概念的心脏原子词典
Abstract
Foundation models for 12-lead electrocardiograms (ECGs) transfer well across clinical tasks, but the physiological knowledge encoded in their representations remains opaque. We present CADENCE, a framework that decomposes an ECG foundation model into a human-interpretable, queryable dictionary of physiological concepts. Using a BatchTopK sparse autoencoder, CADENCE factorizes Layer-6 embeddings from more than nine million ECG tokens into 8,192 sparse cardiac atoms. These atoms align better than individual dense embedding dimensions with clinical phenotypes and waveform morphology, recovering arrhythmias, conduction abnormalities, infarction and repolarization patterns, chamber and axis findings, and lead- and beat-phase-specific waveform primitives. At Layer 6, the best atoms achieve mean AUROCs of 0.88 for clinical phenotypes and 0.90 for morphology, versus 0.78 and 0.83 for the best dense dimensions. Sparse atom probes match or outperform dense probes for phenotype, morphology, and age prediction while attributing each prediction to a small set of interpretable atoms; phenotype AUROC improves from 0.93 to 0.95. Atom-space geometry recovers physiologically coherent relationships, and targeted atom ablation selectively changes frozen downstream outputs. An automated LLM pipeline generates and quantitatively validates atom descriptions by predicting held-out activations. On independent external ECG datasets, CADENCE recovers overlapping concepts and maintains consistent phenotype-prediction performance. CADENCE provides a scalable framework for discovering and auditing the physiological knowledge encoded by ECG foundation models.
Chinese Translation
12导联心电图（ECG）的基础模型在临床任务中具有良好的迁移能力，但其表示中编码的生理知识仍然不透明。我们提出了CADENCE，一个将心电图基础模型分解为人类可解释的、可查询的生理概念词典的框架。通过使用BatchTopK稀疏自编码器，CADENCE将来自超过九百万个心电图标记的第六层嵌入分解为8192个稀疏心脏原子。这些原子在临床表型和波形形态学方面的对齐效果优于单个稠密嵌入维度，能够恢复心律失常、传导异常、心肌梗死和复极模式、心腔和轴向发现，以及导联和心跳相位特定的波形原件。在第六层，最佳原子的临床表型均值AUROC达到0.88，形态学达到0.90，而最佳稠密维度分别为0.78和0.83。稀疏原子探针在表型、形态学和年龄预测方面的表现与稠密探针相当或更优，同时将每个预测归因于一小组可解释的原子；表型的AUROC从0.93提高到0.95。原子空间几何恢复了生理上连贯的关系，针对性地去除原子选择性地改变了冻结的下游输出。一个自动化的LLM管道通过预测保留的激活生成并定量验证原子描述。在独立的外部心电图数据集上，CADENCE恢复了重叠的概念，并保持了一致的表型预测性能。CADENCE提供了一个可扩展的框架，用于发现和审计心电图基础模型中编码的生理知识。
cs.AI / 51 / 2607.25253
The User Asks, Platforms Compete: How Agentic Recommendation Markets Take Shape
用户提问，平台竞争：代理推荐市场的形成
Abstract
Online recommendation has traditionally taken place after a user enters a platform, which determines the candidate pool and the ranking shown to the user. LLM-based user agents enable a different recommendation process: a user specifies a need before choosing a platform, leaving platforms to compete for the user's attention, which we refer to as an agentic recommendation market. In our controlled LLM-based experiments across three product domains, we find this new setting of recommendation creates a tension between access and attention. Compared with traditional platform-centric recommendation, user-centric recommendation greatly expands the opportunity for relevant items to enter comparison; yet broader participation does not translate directly into effective exposure. Competition directly triggers platforms' strategic play: selectively positive explanations occupy 73--78% of first-ranked positions. When the user agent relates platforms' actions to subsequent user feedback, this share falls to 36--41%, while the chance of a user purchasing the relevant item increases. A user agent is therefore more than a ranker over a larger pool of candidates: its querying, ranking, and feedback mechanism governing who can compete, how scarce attention is allocated, and how earlier outcomes shape the evaluation of platforms directly affect user utility. Designing agentic recommendation therefore requires treating access, attention, and accountability as a joint mechanism design problem.
Chinese Translation
在线推荐传统上发生在用户进入平台之后，平台决定候选池和展示给用户的排名。基于大语言模型（LLM）的用户代理使得推荐过程有所不同：用户在选择平台之前先明确需求，平台因此竞争用户的注意力，我们称之为代理推荐市场。在我们针对三个产品领域进行的受控LLM实验中，我们发现这种新的推荐环境在访问和注意力之间产生了紧张关系。与传统的平台中心推荐相比，用户中心推荐大大扩展了相关项目进入比较的机会；然而，更广泛的参与并不直接转化为有效的曝光。竞争直接触发了平台的战略行为：选择性地积极解释占据了73%至78%的第一排名位置。当用户代理将平台的行为与后续用户反馈相关联时，这一比例下降至36%至41%，而用户购买相关项目的机会则增加。因此，用户代理不仅仅是在更大候选池中的排名者：其查询、排名和反馈机制决定了谁可以竞争、稀缺的注意力如何分配，以及早期结果如何影响平台的评估，直接影响用户效用。因此，设计代理推荐需要将访问、注意力和问责制视为一个联合机制设计问题。
cs.AI / 52 / 2607.25279
Many-body Tipping Dynamics of ChatGPT-like AIs
类ChatGPT人工智能的多体倾斜动力学
Abstract
Why do ChatGPT-like AIs, despite major architectural and training differences, unexpectedly tip to undesirable content (e.g. harmful, misleading, repetitive) even under deterministic greedy decoding? We show that a broad class of such tippings is caused by the many-body interactions between tokens (spins) as they cross the finite-layer system. Tipping emerges as a dynamical first passage process between competing output basins. Attention disorder controls the transport toward, away from, or along the basins' boundary. A few-basin reduction yields a closed finite-layer threshold, whose coarse-grained predictions show good agreement across ChatGPT-like families. These results suggest that a broad class of AI failures represents 'foreseeable engineering risk' rather than inherently unpredictable behavior, with important implications for legal and societal assessments of AI harm.
Chinese Translation
为什么类ChatGPT的人工智能尽管在架构和训练上存在重大差异，却在确定性贪婪解码下意外倾向于不良内容（例如有害、误导性、重复性）？我们表明，这种倾斜的广泛类别是由于在有限层系统中，标记（自旋）之间的多体相互作用所导致的。倾斜作为在竞争输出盆地之间的动态首次通过过程而出现。注意力障碍控制着向盆地的运输、远离盆地或沿着盆地边界的运输。少盆地简化产生了一个封闭的有限层阈值，其粗粒度预测在类ChatGPT的家族中表现出良好的一致性。这些结果表明，广泛类别的人工智能失败代表了“可预见的工程风险”，而不是固有的不可预测行为，这对法律和社会对人工智能危害的评估具有重要影响。
cs.AI / 53 / 2607.25283
ContractHIL-HLS: Contract-Aligned Multi-Agent Workflow with Hardware-in-the-Loop Feedback for HLS Design
ContractHIL-HLS：面向高层次综合设计的合同对齐多智能体工作流与硬件在环反馈
Abstract
This paper presents ContractHIL-HLS, a contract-aligned multi-agent workflow for practical high-level synthesis (HLS) engineering. The workflow makes three contributions. First, it introduces a structured contract as the semantic-alignment and task-execution artifact that translates natural language requirements into explicit interfaces, constraints, validation checks, and rollback rules. Second, it incorporates hardware information into the feedback loop by feeding HLS, Vivado, PYNQ runtime, power, and failure evidence back into generation, thereby extending LLM-assisted HLS from kernel code toward system- and board-level closure. Third, it decomposes agents by semantic lowering and execution tasks rather than by conversational roles: a Contract Agent lowers natural language into the contract, an HTML Agent renders the contract as persistent structured HTML, and a Hardware-in-the-Loop Agent implements and revises the design with measured evidence. We evaluate ContractHIL-HLS in two parts. On 94 locally executable HLS-Eval tasks, the structured contract provides the largest small design gain, improving the estimated single-sample testbench pass rate from 64.0% to 70.2%; the full flow reaches 70.4% pass@1 and 76.6% pass@5. Because HLS-Eval does not exercise board-level design, we also validate ContractHIL-HLS on a board tested ML-KEM/ML-DSA post-quantum cryptography (PQC) secure-message accelerator, where the retained dual-bitstream organization reduces six-message average text runtime from 207.3 ms to 52.4 ms with positive routed WNS on both images while preserving decrypted-message verification. We open-source our work at BJUT-CS316-LAB/ContractHIL-HLS (https://github.com/BJUT-CS316-LAB/ContractHIL-HLS).
Chinese Translation
本文提出了ContractHIL-HLS，一种面向实际高层次综合（HLS）工程的合同对齐多智能体工作流。该工作流有三个贡献。首先，它引入了一种结构化合同，作为语义对齐和任务执行的工件，将自然语言需求转化为明确的接口、约束、验证检查和回滚规则。其次，它通过将HLS、Vivado、PYNQ运行时、功耗和故障证据反馈到生成中，将硬件信息纳入反馈循环，从而将LLM辅助的HLS从内核代码扩展到系统和板级闭合。第三，它通过语义降级和执行任务而非对话角色来分解智能体：合同智能体将自然语言降级为合同，HTML智能体将合同呈现为持久的结构化HTML，而硬件在环智能体则使用测量证据实施和修订设计。我们在两个部分评估了ContractHIL-HLS。在94个本地可执行的HLS-Eval任务中，结构化合同提供了最大的设计增益，将估计的单样本测试平台通过率从64.0%提高到70.2%；完整流程达到70.4%的通过率@1和76.6%的通过率@5。由于HLS-Eval未涉及板级设计，我们还在经过测试的ML-KEM/ML-DSA后量子密码学（PQC）安全消息加速器上验证了ContractHIL-HLS，其中保留的双比特流组织将六消息的平均文本运行时间从207.3毫秒减少到52.4毫秒，同时在两个图像上保持正向路由的WNS，并保留解密消息的验证。我们在BJUT-CS316-LAB/ContractHIL-HLS（https://github.com/BJUT-CS316-LAB/ContractHIL-HLS）开源了我们的工作。
cs.AI / 54 / 2607.25292
Instruction-Tuned Language Models Cannot Sample from Distributions They Can Describe
指令调优的语言模型无法从它们能够描述的分布中抽样
Abstract
Silicon sampling uses language models as proxies for human survey respondents, treating each model call as an independent draw from the persona's response distribution. We show this draw does not exist: instruction-tuned models do not sample from distributions, they collapse to a single output. The same persona on the same question returns the same answer on more than half of items in a public-opinion benchmark. The collapse is sharp: the model's internal probabilities concentrate on a single option, and the failure is substantially amplified by instruction tuning: across three model families with materially different post-training pipelines, every instruction-tuned model fails on every task we test, while base models fail far less often. Strikingly, the same model that cannot sample from a distribution can describe it accurately in a single call. We call this gap the KNOWS/DOES split, and trace it to a degenerate sampling primitive visible in the logits and induced by alignment training. Exploiting this split, asking the model to describe the response distribution in one call more than halves the error against human survey data compared to persona aggregation. For applications that require per-persona outputs, we propose Prompt-Perturbed Argyle (PPA), which reduces the same error by 21% at no added cost.
Chinese Translation
硅采样将语言模型作为人类调查受访者的代理，将每次模型调用视为从个性响应分布中独立抽取的一次样本。我们展示了这种抽样并不存在：指令调优模型并不从分布中抽样，而是收敛到单一输出。在公共舆论基准中，相同个性在相同问题上对超过一半的项目返回相同答案。这种收敛是显著的：模型的内部概率集中在单一选项上，而这种失败在指令调优的影响下显著放大：在三种具有实质性不同后训练流程的模型家族中，每个指令调优模型在我们测试的每个任务上都失败，而基础模型的失败率则低得多。值得注意的是，无法从分布中抽样的模型可以在一次调用中准确描述该分布。我们将这一差距称为 KNOWS/DOES 分裂，并追溯到在 logits 中可见的退化抽样原语，该原语是由对齐训练引起的。利用这一分裂，要求模型在一次调用中描述响应分布，相较于个性聚合，能够将与人类调查数据的误差减少一半以上。对于需要每个个性输出的应用，我们提出了 Prompt-Perturbed Argyle (PPA)，在没有额外成本的情况下将相同的误差减少了21%。
cs.AI / 55 / 2607.25321
Physics-Grounded Fluid Video Generation with a Simulation Dataset and Dual-Stream Optical-Flow Supervision
基于物理的流体视频生成：模拟数据集与双流光流监督
Abstract
Video diffusion models generate visually compelling content but routinely violate elementary physics when the subject involves fluids: liquid columns break apart in mid-air, container water levels fail to rise as liquid is poured in, and splashes disperse without regard to momentum or gravity. We attribute this gap to the fact that large-scale video-text corpora contain almost no explicit motion supervision, so models learn to imitate fluid appearance rather than dynamics. We address this with two contributions. First, we build a physics-simulation fluid dataset combining 1,638 MPM-simulated pouring/sloshing videos with 2,320 keyword-filtered real pouring videos mined from stock footage, plus two held-out test sets: a 1,515-video real-video benchmark and an 18-prompt text-to-first-frame generalization benchmark. Second, we introduce a dual-stream image-to-video architecture built on a pretrained diffusion-transformer video generator. It augments the standard RGB decoder with a lightweight Optical-Flow Decoder branch trained with explicit end-point-error and smoothness losses, fused into the RGB stream via zero-initialized convolutions so the pretrained backbone starts undisturbed. Only the two decoders are updated; the encoder, temporal transformer, and text encoder remain frozen. Across two model scales (1.3B and 14B) and two test sets, our method improves VideoPhy-2 Physical-Commonsense and Video-Quality scores over the frozen backbone by up to 8.75 and 4.65 points, outperforms a leading open competitor, and is preferred by human raters in a blind study. A direct optical-flow read-out evaluation further shows an end-point error as low as 0.54 pixels in-distribution, confirming the model has internalized a coherent motion prior rather than merely improving surface appearance.
Chinese Translation
视频扩散模型生成视觉上引人注目的内容，但在涉及流体的主题时，常常违反基本物理规律：液柱在空中断裂，容器中的水位在倒入液体时未能上升，飞溅物的扩散忽视了动量和重力。我们将这一差距归因于大规模视频-文本语料库几乎没有明确的运动监督，因此模型学习模仿流体的外观而非动态。我们通过两个贡献来解决这一问题。首先，我们构建了一个物理模拟流体数据集，结合了1,638个MPM模拟的倒水/晃动视频和2,320个经过关键词筛选的真实倒水视频，后者来自库存素材，并且设立了两个保留的测试集：一个包含1,515个视频的真实视频基准和一个包含18个提示的文本到第一帧的泛化基准。其次，我们引入了一种基于预训练扩散变换器视频生成器的双流图像到视频架构。它通过轻量级的光流解码器分支增强了标准RGB解码器，该分支使用明确的端点误差和平滑损失进行训练，并通过零初始化卷积融合到RGB流中，以便预训练的主干保持不受干扰。只有这两个解码器被更新；编码器、时间变换器和文本编码器保持冻结。在两个模型规模（1.3B和14B）及两个测试集上，我们的方法在VideoPhy-2物理常识和视频质量评分上，相较于冻结主干提高了最多8.75和4.65分，超越了一家领先的开放竞争者，并在盲测中被人类评审者偏好。直接的光流读出评估进一步显示，在分布内的端点误差低至0.54像素，确认模型已经内化了一种连贯的运动先验，而不仅仅是改善表面外观。
cs.AI / 56 / 2607.25322
From Cellular Responses to Pharmacological Domains: Multimodal Zero-Shot Drug Representation Learning
从细胞反应到药理领域：多模态零样本药物表征学习
Abstract
Multimodal drug discovery enables drug representation learning beyond chemical structure by incorporating cellular responses such as gene expression and cell morphology. However, direct fusion and instance-level contrastive alignment may mix mechanism-related signals with modality-specific noise and incorrectly separate structurally dissimilar but biologically related compounds. This limitation can obscure transferable mechanism patterns required for predicting the properties of unseen compounds. We introduce PMRD, a pharmacological response domain-guided framework for multimodal zero-shot drug property prediction. PMRD separates mechanism-consistent factors from modality-specific information and constructs a consensus response domain across three modalities. Mechanism candidate augmentation identifies locally stable factors, while retrieval-geometry attribution dynamically reweights the alignment and augmentation objectives according to whether their updates preserve inter-drug discriminability.This feedback suppresses training signals that conflict with mechanism-discriminative retrieval. PMRD further combines complementary representations through reliability-aware multiview retrieval. Experiments on public datasets show improved zero-shot property prediction and more biologically coherent drug neighborhoods. Hard-negative analysis further indicates fewer conflicts between structurally dissimilar but response-related compounds. These results support PMRD as an effective framework for mechanism-aware multimodal drug representation learning.\footnote{The code will be released upon publication.}
Chinese Translation
多模态药物发现通过结合基因表达和细胞形态等细胞反应，超越化学结构实现药物表征学习。然而，直接融合和实例级对比对齐可能会将机制相关信号与特定模态的噪声混合，并错误地将结构上不相似但生物学上相关的化合物分开。这一局限性可能会掩盖预测未见化合物所需的可转移机制模式。我们提出了PMRD，一个药理反应领域引导的多模态零样本药物属性预测框架。PMRD将机制一致的因素与特定模态的信息分离，并在三种模态之间构建共识反应领域。机制候选增强识别局部稳定因素，而检索几何归因根据其更新是否保持药物间的可区分性动态重新加权对齐和增强目标。这种反馈抑制与机制区分检索相冲突的训练信号。PMRD进一步通过可靠性感知的多视图检索结合互补表征。对公共数据集的实验显示出改进的零样本属性预测和更具生物学一致性的药物邻域。困难负样本分析进一步表明结构上不相似但反应相关的化合物之间的冲突更少。这些结果支持PMRD作为一个有效的机制感知多模态药物表征学习框架。
cs.AI / 57 / 2607.25338
Dual-Domain Manifold Modeling for Hyperspectral Image Fusion
双域流形建模用于高光谱图像融合
Abstract
Achieving a coherent integration of spectral richness and spatial fidelity remains a central objective in hyperspectral image fusion. However, existing hyperspectral image fusion methods struggle to effectively model geometric constraints. In the spatial domain, weak spatial-spectral interaction limits geometry-aware feature learning and suppresses high-frequency structural information, resulting in low-frequency bias and structural degradation. In the spectral domain, local manifold structures induced by spectral similarity are insufficiently exploited, limiting intrinsic pixel relationship modeling and fine-grained spectral reconstruction. To address these challenges, we propose a dual-domain manifold modeling (DDMM) framework. Specifically, we introduce a Topology-Aware Transformer (TPFormer) that combines global attention with neighborhood propagation, jointly modeling spatial topology and pixel-level feature manifold relationships to capture intrinsic spatial-spectral structures and improve topology-aware representation learning. Furthermore, a Frequency-Decoupled Spatial-Spectral Collaborative Fusion (FDSCF) module is devised, in which features are projected into the frequency domain via the discrete cosine transform and explicitly decoupled into low- and high-frequency components. Guided by a low-rank structural prior and spectral-driven spatial enhancement, FDSCF selectively enhances geometry-aware high-frequency features, strengthening spatia-spectral coupling and recovering sharper edges and finer textures. Extensive experiments on multiple benchmark datasets demonstrate that DDMM achieves superior overall performance over SoTA methods in terms of spatial structure preservation and spectral reconstruction.
Chinese Translation
实现光谱丰富性与空间保真性的协调整合仍然是高光谱图像融合的核心目标。然而，现有的高光谱图像融合方法在有效建模几何约束方面存在困难。在空间域中，弱空间-光谱交互限制了对几何感知特征的学习，并抑制了高频结构信息，导致低频偏差和结构退化。在光谱域中，由光谱相似性引起的局部流形结构未得到充分利用，限制了内在像素关系建模和细粒度光谱重建。为了解决这些挑战，我们提出了一种双域流形建模（DDMM）框架。具体而言，我们引入了一种拓扑感知变换器（TPFormer），该变换器结合了全局注意力与邻域传播，共同建模空间拓扑和像素级特征流形关系，以捕捉内在的空间-光谱结构并改善拓扑感知表示学习。此外，我们设计了一个频率解耦空间-光谱协同融合（FDSCF）模块，其中特征通过离散余弦变换投影到频率域，并显式解耦为低频和高频成分。在低秩结构先验和光谱驱动的空间增强的指导下，FDSCF选择性地增强几何感知的高频特征，增强空间-光谱耦合，恢复更清晰的边缘和更细腻的纹理。在多个基准数据集上的广泛实验表明，DDMM在空间结构保留和光谱重建方面的整体性能优于最先进（SoTA）的方法。
cs.AI / 58 / 2607.25340
Cardiologent: Multi-Agent Clinical Decision Support for Patient-Level Arrhythmia Assessment, Urgency, and Management
Cardiologent：多智能体临床决策支持系统用于患者级心律失常评估、紧急性和管理
Abstract
The same episode of atrial fibrillation is a minor finding in a healthy adult and grounds for anticoagulation in an elderly patient with hypertension: identical signal, opposite decision. Naming the rhythm is only the start; what determines a patient's outcome is the judgement that follows -- what the arrhythmia is across the whole record, what it means for this patient, and what should be done about it. Recent work pairing large language models with the ECG stops short of this, reading one recording without assembling a patient-level finding; and agentic systems built around it either receive the arrhythmia a device has already detected or target a different diagnostic task, stopping before the decision this task requires. We formulate patient-level arrhythmia decision support as a task and present Cardiologent, a multi-agent system that spans it from detection to decision. An agent for each signal -- a single ECG lead and the photoplethysmogram a wearable acquires -- grounds its window reading in measured features rather than a bare label; the readings are assembled into the patient's rhythm profile and, with the patient's own data, reasoned against clinical guidelines retrieved for the case, with a critic checking each conclusion against the guideline it cites. We evaluate the clinical decision rather than the report, across integrated diagnosis, clinical significance, and urgency and management. Cardiologent scores highest on every axis, first on every patient-level task under both cardiologists and an at-scale LLM judge -- whose agreement with the cardiologists (ICC 0.74, 0.66) matches theirs with each other (0.67). Because each conclusion traces to a cited guideline and is validated against expert cardiologists, it yields decisions a clinician can audit rather than act on blindly -- a step toward use in continuous monitoring.
Chinese Translation
同一次房颤在健康成年人中是一个轻微的发现，而在高血压老年患者中则是抗凝治疗的依据：相同的信号，却得出相反的决策。命名心律只是开始；决定患者结果的是随之而来的判断——整个记录中的心律是什么，这对该患者意味着什么，以及应采取什么措施。最近的研究将大型语言模型与心电图（ECG）结合，但未能实现这一目标，仅仅读取一条记录而未能整合患者级的发现；围绕它构建的智能系统要么接收设备已经检测到的心律失常，要么针对不同的诊断任务，停留在该任务所需决策之前。我们将患者级心律失常决策支持形式化为一项任务，并提出Cardiologent，这是一个涵盖从检测到决策的多智能体系统。每个信号的智能体——单个ECG导联和可穿戴设备获取的光电容积脉搏波图（PPG）——基于测量特征而非简单标签来确定其窗口读取；这些读取被整合成患者的心律档案，并结合患者自身的数据，与为该病例检索的临床指南进行推理，且有一个评论者检查每个结论是否符合其引用的指南。我们评估临床决策而非报告，涵盖综合诊断、临床意义和紧急性及管理。Cardiologent在每个维度上得分最高，在心脏病专家和大规模大型语言模型（LLM）评判下的每个患者级任务中均排名第一——其与心脏病专家的协议（ICC 0.74, 0.66）与他们之间的协议（0.67）相匹配。由于每个结论都追溯到引用的指南，并经过专家心脏病医生的验证，因此它产生的决策是临床医生可以审计的，而不是盲目采取的——这是向持续监测应用迈出的重要一步。
cs.AI / 59 / 2607.25364
Explanation-Bound Tool Execution for AI Agents: Server-Verified Action Claims Without Trusting Model Rationales
面向人工智能代理的解释绑定工具执行：无需信任模型推理的服务器验证行动声明
Abstract
Tool-using agents expose structured calls but commonly attach free-form rationales. Such rationales are neither authorization nor reliable introspection. We present Explanation-Bound Tool Execution (EBTE), a claim-carrying mediation layer that converts decision-relevant rationale content into typed action claims and checks them against server-held intent, policy, payload, tool, risk, provenance, and freshness facts. EBTE cannot widen baseline authority: conflicts deny, incomplete or uncertain claims review, and only matching claims remain eligible for governed execution. We formalize this composition under explicit mediation and trusted-fact assumptions and implement a versioned reference profile with minimized audit packets. Across 136 authored conformance scenarios, the full profile matches all specified dispositions, admits none of 96 designated hard contradictions, and passes 232 metamorphic checks; these results validate the included profile rather than population performance. A draft-only reference integration forwards none of 48 authored hard cases under EBTE while preserving all 16 soft-review and 4 aligned draft paths. In a frozen 2026-07-12 exploratory 224-attempt hosted-model record, the historical generation/runner agreement counts are 71/96, 66/96, and 19/32; a separately labeled zero-call post-hoc revalidation of the preserved minimized claims under the current pipeline yields 70/96, 65/96, and 17/32. In an AgentDojo-derived semantic check, existing high-risk controls already make all 12 attack proposals non-allow; EBTE additionally resolves them as deny. These results support the feasibility and diagnostic value of server-checked action claims, not rationale faithfulness, human-review benefit, representative attack resistance, or production safety.
Chinese Translation
使用工具的代理暴露结构化调用，但通常附带自由形式的推理。这些推理既不是授权，也不是可靠的自我审视。我们提出了解释绑定工具执行（Explanation-Bound Tool Execution, EBTE），这是一种承载声明的中介层，它将与决策相关的推理内容转换为类型化的行动声明，并将其与服务器持有的意图、政策、有效负载、工具、风险、来源和新鲜度事实进行核对。EBTE 不能扩大基本权限：冲突会被拒绝，不完整或不确定的声明将被审查，只有匹配的声明才有资格进行受管执行。我们在明确的中介和可信事实假设下形式化了这种组合，并实现了一个版本化的参考配置文件，最小化审计数据包。在 136 个编写的符合性场景中，完整的配置文件匹配所有指定的处置，未接受 96 个指定的硬性矛盾，并通过 232 个变形检查；这些结果验证了所包含配置文件的有效性，而非人群表现。在仅草稿的参考集成中，EBTE 下未转发 48 个编写的硬性案例，同时保留了所有 16 个软审查和 4 个对齐的草稿路径。在一个冻结的 2026-07-12 探索性 224 次尝试的托管模型记录中，历史生成/运行者协议计数为 71/96、66/96 和 19/32；在当前管道下对保留的最小化声明进行的单独标记零调用事后重新验证结果为 70/96、65/96 和 17/32。在一个基于 AgentDojo 的语义检查中，现有的高风险控制已经使所有 12 个攻击提案不被允许；EBTE 进一步将其判定为拒绝。这些结果支持了服务器检查的行动声明的可行性和诊断价值，而非推理的可信度、人类审查的益处、代表性的攻击抵抗或生产安全。
cs.AI / 60 / 2607.25368
AI Deployment and Cyber Governance Failures in Public-Sector Organizations: A Typological Analysis
公共部门组织中人工智能部署与网络治理失败：一种类型学分析
Abstract
The intersection of artificial intelligence adoption, cybersecurity governance, and public sector institutional constraints has not been examined as a unified analytical problem in the existing literature. Studies address AI cybersecurity risks generically, public sector governance independently, and framework adequacy separately. Existing studies have not integrated these three streams to explain specifically how AI adoption causes cybersecurity governance failure in government organizations, nor test existing governance instruments against AI-specific public sector failure causes. This paper ad-dresses that gap. It proposes a seven-domain typology identifying ten specific AI-driven cyber governance failure causes grounded in public sector institutional analysis. It presents a three-pathway failure model showing how accountability failure, opera-tional resilience failure, and compliance failure interact and reinforce each other. It de-livers a structured coverage matrix testing five major governance frameworks (NIST CSF 2.0, ISO/IEC 27001, COBIT, NIST AI RMF, and ISO/IEC 42001) against the typology, finding that no instrument addresses Shadow AI, speed asymmetry, or gov-ernance vacuum at the operational specificity required for public sector application. The paper introduces speed asymmetry as a named structural construct with a specified mechanism. The framework provides the design specification for an AI-enabled cyber-security maturity model for government organizations.
Chinese Translation
现有文献中尚未将人工智能采纳、网络安全治理与公共部门制度约束的交集作为一个统一的分析问题进行研究。相关研究通常泛泛而谈人工智能的网络安全风险，独立探讨公共部门治理，以及单独评估框架的适用性。现有研究未能整合这三方面内容，以具体解释人工智能采纳如何导致政府组织中的网络安全治理失败，也未对现有治理工具进行针对人工智能特定公共部门失败原因的检验。本文填补了这一空白。它提出了一种七领域类型学，识别出十个基于公共部门制度分析的特定人工智能驱动的网络治理失败原因。本文展示了一个三路径失败模型，说明了问责失败、运营韧性失败和合规失败如何相互作用并相互强化。它提供了一个结构化的覆盖矩阵，测试了五个主要治理框架（NIST CSF 2.0、ISO/IEC 27001、COBIT、NIST AI RMF 和 ISO/IEC 42001）与类型学的匹配，发现没有任何工具能够在公共部门应用所需的操作特异性上解决影子人工智能、速度不对称或治理真空问题。本文将速度不对称引入为一个命名的结构性构造，并指定了其机制。该框架为政府组织的人工智能驱动的网络安全成熟度模型提供了设计规范。
cs.AI / 61 / 2607.25369
ODYSSE: Episode-wise Policy Optimization for Personalized Agentic Reasoning
ODYSSE：基于情节的个性化代理推理策略优化
Abstract
Agentic systems have rapidly advanced in their ability to interact with real-world environments, leverage external tools, and provide services for users. However, unlike natural-world tasks that assume well-defined instructions, human-centered scenarios are characterized by ambiguous requests that lead to large, open-ended solution spaces. Decoding users' personalized preferences is therefore essential for narrowing the candidate solution space. This introduces a new challenge, personalized agentic reasoning, which requires agents to jointly interact with both users and environments to deliver personalized services. In this paper, we present ODYSSE, a Reinforced Fine-Tuning (RFT) framework for personalized agentic reasoning. At its core, ODYSSE proposes Episode-wise GRPO (ESPO), a novel extension of Group Relative Policy Optimization (GRPO) designed to address long action horizons and strong cross-step dependencies in personalized agentic reasoning. Rather than optimizing individual steps independently, ESPO introduces an episode-level reward mechanism together with episodic advantage estimation, enabling upstream evidence to effectively guide downstream personalized decisions and allowing agents to progressively resolve ambiguous user requests across multiple interaction steps. We further propose an episodic batch sampler that groups actions from the same episode into unified training batches, facilitating coherent optimization under ESPO. We evaluate ODYSSE on realistic long-horizon personalized GUI reasoning tasks. Experimental results demonstrate that ODYSSE consistently outperforms both specialist and general-purpose LVLMs, highlighting its effectiveness for personalized agentic reasoning.
Chinese Translation
代理系统在与现实环境互动、利用外部工具以及为用户提供服务的能力上迅速发展。然而，与假定有明确指令的自然世界任务不同，以人为中心的场景特征在于模糊的请求，这导致了广泛且开放的解决方案空间。因此，解码用户的个性化偏好对于缩小候选解决方案空间至关重要。这引入了一个新的挑战——个性化代理推理，要求代理与用户和环境共同互动，以提供个性化服务。在本文中，我们提出了ODYSSE，一个用于个性化代理推理的强化微调（Reinforced Fine-Tuning, RFT）框架。ODYSSE的核心是提出了情节级别的群体相对策略优化（Episode-wise GRPO, ESPO），这是群体相对策略优化（Group Relative Policy Optimization, GRPO）的新扩展，旨在解决个性化代理推理中的长行动时间范围和强跨步骤依赖性。ESPO并不是独立优化每个步骤，而是引入了情节级奖励机制和情节优势估计，使得上游证据能够有效指导下游个性化决策，并允许代理在多个互动步骤中逐步解决模糊的用户请求。我们进一步提出了一种情节批量采样器，将来自同一情节的动作分组为统一的训练批次，从而促进在ESPO下的连贯优化。我们在现实的长时间个性化图形用户界面推理任务上评估了ODYSSE。实验结果表明，ODYSSE在性能上始终优于专业和通用的语言视觉大模型（LVLM），突显了其在个性化代理推理中的有效性。
cs.AI / 62 / 2607.25379
Cyber-Capable AI Agents: Vulnerabilities, Evaluation Containment, and Defensive Response
具备网络能力的人工智能代理：脆弱性、评估控制与防御响应
Abstract
Cyber-capable AI agents combine language models with tools, memory, and execution en- vironments to perform multi-step offensive-security tasks. Existing work separately measures cyber capability and catalogs attacks against agent components, but provides less guidance on containing a capable agent within the environments used to evaluate it. This review synthe- sizes five vulnerability classes at that boundary: multi-step offensive chains, objectives that conflict with sandbox boundaries, supply-chain and credential exposure, persistent command- and-control, and the speed of automated action. We use the reported July 2026 Hugging Face/OpenAI incident as a bounded case study, distinguishing incident-specific observations from findings established in the wider literature. Across the taxonomy and case, we examine controls for containment, privilege separation, provenance, and responder access, including the dual-use problem that defensive artifacts may also enable misuse. The review identifies practical priorities for evaluating cyber capability together with the security of the environment in which that capability is exercised.
Chinese Translation
具备网络能力的人工智能代理结合了语言模型、工具、记忆和执行环境，以执行多步骤的攻击安全任务。现有研究分别测量网络能力并对代理组件的攻击进行分类，但对如何在评估环境中控制具备能力的代理提供的指导较少。本综述综合了在这一边界上的五类脆弱性：多步骤攻击链、与沙箱边界冲突的目标、供应链和凭证暴露、持久的指挥与控制，以及自动化行动的速度。我们以2026年7月报告的Hugging Face/OpenAI事件作为一个有限的案例研究，区分事件特定的观察与更广泛文献中确立的发现。在分类法和案例中，我们审查了控制措施，包括控制、特权分离、来源追溯和响应者访问，考虑到防御性工件可能也会导致误用的双重用途问题。该综述确定了评估网络能力与该能力施展环境安全的实际优先事项。
cs.AI / 63 / 2607.25398
HANDBOOK.md: A Benchmark for Long-Context Agentic Instruction Following
HANDBOOK.md：长上下文自主指令遵循的基准测试
Abstract
Language-model agents are increasingly deployed under standing instructions: a system prompt, a policy file, or a skills document is placed in context, and the agent is trusted to let it govern every action that follows. Existing benchmarks rarely test this deployment pattern directly; they measure whether an agent can complete a task, not whether a long, binding policy document actually constrains its behavior over an extended tool-use horizon. We present HANDBOOK.md, a benchmark of 65 agentic tasks modeled on how enterprise employees follow company handbooks. Each task places an agent in a self-contained company environment, a file workspace together with mock email, chat, calendar, issue-tracking, and commerce services exposed over the Model Context Protocol, and instructs it to carry out routine professional work governed by an expert-written standard operating procedure of 20 to 124 pages. Tasks span five domains (finance, medical billing, insurance, logistics, and HR) and ten fictional companies. To resist memorization, every task modifies one of ten base handbooks, altering the specific rules and thresholds on which grading turns, so no two tasks share a policy. Grading is fully deterministic: each task carries a rubric of programmatic criteria (824 in total) that check both that required actions occurred and that prohibited actions did not. Under strict grading, where a trial passes only if every criterion is satisfied, the best of thirty evaluated model configurations passes 36.2% of trials, and most frontier configurations remain below 25%. Failures follow consistent patterns: agents let a plausible in-environment request override the standing policy, perform a required check and then act against its result, lose rule details over long horizons, and report compliance they did not achieve. We release all tasks, environments, and the evaluation harness.
Chinese Translation
语言模型代理越来越多地在既定指令下部署：一个系统提示、一个政策文件或一个技能文档被置于上下文中，代理被信任以让其主导随后的每一个行动。现有基准测试很少直接测试这种部署模式；它们测量的是代理是否能够完成任务，而不是一个长的、具有约束力的政策文件是否在延长的工具使用范围内真正约束其行为。我们提出了HANDBOOK.md，这是一个基于企业员工如何遵循公司手册的65个自主任务的基准测试。每个任务将代理置于一个自给自足的公司环境中，文件工作区以及通过模型上下文协议暴露的模拟电子邮件、聊天、日历、问题跟踪和商业服务，并指示其根据专家编写的标准操作程序（20到124页）进行日常专业工作。任务涵盖五个领域（金融、医疗账单、保险、物流和人力资源）和十个虚构公司。为了抵制记忆，每个任务修改十个基础手册中的一个，改变评分所依据的具体规则和阈值，因此没有两个任务共享相同的政策。评分是完全确定性的：每个任务都有一套程序化标准（总共824项）来检查所需的行动是否发生以及禁止的行动是否未发生。在严格评分下，只有在满足所有标准的情况下，试验才能通过，评估的三十个模型配置中表现最佳的仅通过了36.2%的试验，而大多数前沿配置仍低于25%。失败遵循一致的模式：代理让一个合理的环境内请求覆盖既定政策，执行必要的检查后又违反其结果，在长时间范围内丢失规则细节，并报告未能实现的合规性。我们发布所有任务、环境和评估工具。
cs.AI / 64 / 2607.25400
COVENANT: Natural-Language Workflow Compilation for Aligned Agent Execution
COVENANT：用于对齐代理执行的自然语言工作流编译
Abstract
Large language model (LLM) agents are increasingly entrusted with natural-language workflow instructions (e.g., retail-payment policies) that specify not only what outcome to achieve, but also which steps, branches, and tool interactions are permitted. When these instructions are supplied as prompt context, however, the model retains control over both procedure selection and step execution. As interactions accumulate, an agent can skip required steps, take unsupported branches, or execute a valid step with unsupported arguments or effects--a failure mode we call workflow misalignment. In this work, we propose COVENANT, a compiler-and-interpreter architecture for workflow-aligned agent execution. Our key insight is to treat workflow instructions as source programs rather than prompts. COVENANT converts the instructions into a workflow abstract syntax tree (WAST) and lowers it to a workflow control-flow graph (WCFG). At runtime, a controller interprets the WCFG one node at a time, checks each proposal against requirements extracted from the instructions before committing controller state or advancing the graph, and returns diagnostic feedback for repair. To evaluate COVENANT, we use 120 cases from three existing benchmarks, spanning seven workflow scenarios. Compared with state-of-the-art LLM agents, COVENANT improves benchmark success from 50.00% to 83.33% and reduces the workflow-misalignment failure rate from 42.50% to 15.83% (62.75% relative). These results show that COVENANT substantially mitigates workflow misalignment, moving LLM-agent alignment beyond isolated prompt following toward reliable execution of complex and multi-step workflows.
Chinese Translation
大型语言模型（LLM）代理越来越多地被委以自然语言工作流指令（例如，零售支付政策），这些指令不仅指定了要实现的结果，还规定了允许的步骤、分支和工具交互。然而，当这些指令作为提示上下文提供时，模型仍然控制着程序选择和步骤执行。随着交互的积累，代理可能会跳过必需的步骤、采取不支持的分支，或以不支持的参数或效果执行有效步骤——我们称之为工作流不对齐的失败模式。在本研究中，我们提出了COVENANT，一种用于工作流对齐代理执行的编译器和解释器架构。我们的关键见解是将工作流指令视为源程序而非提示。COVENANT将指令转换为工作流抽象语法树（WAST），并将其降低为工作流控制流图（WCFG）。在运行时，控制器逐节点解释WCFG，在提交控制器状态或推进图之前，检查每个提案是否符合从指令中提取的要求，并返回诊断反馈以进行修复。为了评估COVENANT，我们使用了来自三个现有基准的120个案例，涵盖七种工作流场景。与最先进的LLM代理相比，COVENANT将基准成功率从50.00%提高到83.33%，并将工作流不对齐失败率从42.50%降低到15.83%（相对降低62.75%）。这些结果表明，COVENANT显著减轻了工作流不对齐，将LLM代理的对齐从孤立的提示跟随提升到复杂多步骤工作流的可靠执行。
cs.AI / 65 / 2607.25408
Context Assembly as the Controlled Variable: A Control-Theoretic View of Harness Policies for Frozen LLM Agents
将上下文组装作为受控变量：对冷冻大语言模型代理的控制理论视角下的控制策略
Abstract
A growing body of 2026 work applies control theory to LLM agents: Lyapunov-certified stability for tool-mediated controllers (Prinos et al., "Stable Agentic Control", 2026), sample-complexity bounds for sparse policies over massive discrete tool universes (Majumdar, "Sparse Agentic Control", 2026), and regulatory-control decompositions of multi-agent systems into auditable feedback loops (Nogueira and Skogestad, 2026). We do not claim to introduce control theory to LLM agents -- that ship has sailed. Our narrower claim is about what the controlled variable is. Prior work controls tool selection, inter-agent message routing, or the agent's raw action stream. We instead treat context assembly itself -- which prompt template, which few-shot demonstrations, how much retrieved context, how many planning/verification passes -- as the controlled variable, learned online by a contextual bandit or REINFORCE policy sitting outside a frozen model. This paper develops the formal decomposition (inner frozen policy $\pi_\theta$, outer context policy $\pi_\phi$), gives a stability argument for the online controller in the sense used by Zhang et al. (2026) (non-decreasing expected reward under bounded policy change), and reports an uncertainty-calibration analysis of the controller's own confidence against realized task outcomes. The applied counterpart to this paper instantiates the same controller across three domains and two model providers and releases the dataset, trajectory logs, and a deployment recipe; here we focus on the formal framing and the stability/uncertainty evidence a control-theoretic claim requires.
Chinese Translation
越来越多的研究工作将控制理论应用于大语言模型（LLM）代理：Lyapunov 认证的工具介导控制器的稳定性（Prinos 等，"稳定的代理控制"，2026），在庞大的离散工具宇宙中稀疏策略的样本复杂性界限（Majumdar，"稀疏代理控制"，2026），以及将多代理系统的监管控制分解为可审计的反馈循环（Nogueira 和 Skogestad，2026）。我们并不声称将控制理论引入 LLM 代理——这艘船已经启航。我们更狭义的主张是关于受控变量的定义。先前的工作控制工具选择、代理间消息路由或代理的原始动作流。而我们则将上下文组装本身——使用哪个提示模板、哪些少量示例、检索多少上下文、进行多少次规划/验证——视为受控变量，由一个位于冷冻模型外部的上下文赌博者或 REINFORCE 策略在线学习。本文发展了形式分解（内层冷冻策略 $ heta_ heta$，外层上下文策略 $ heta_ heta$），为在线控制器提供了稳定性论证，符合 Zhang 等（2026）所使用的意义（在有界策略变化下期望奖励非递减），并报告了控制器自身信心与实际任务结果之间的不确定性校准分析。本文的应用对应物在三个领域和两个模型提供者中实例化了相同的控制器，并发布了数据集、轨迹日志和部署配方；在这里我们专注于形式框架以及控制理论主张所需的稳定性/不确定性证据。
cs.AI / 66 / 2607.25415
A Control System, a Dataset, and a Recipe for Making Frozen LLM Agents Learn a Domain
控制系统、数据集及制作冻结LLM代理学习领域的方案
Abstract
Production LLM agents are increasingly assembled from a frozen model wrapped in a harness: a prompt template, a tool set, a memory/retrieval layer, a planning strategy, and a verification policy. Two 2026 systems, Meta-Harness (Lee et al., 2026) and HyperAgents (Meta AI, 2026), show that this harness can itself be optimized or even self-rewritten by an agentic proposer -- at the cost of either an expensive code-search loop or unconstrained self-modifying code, neither of which is auditable or usable with a fully black-box model API. We take a narrower, more constrained position: treat the harness as a small, fixed, human-legible action space and learn a policy over it online with classic sample-efficient reinforcement learning (an $\epsilon$-greedy contextual bandit and REINFORCE), scored against a multi-objective reward (task success, verifier score, policy compliance, cost, latency, and an unsupported-claim penalty). We instantiate this control system with DSPy (Khattab et al., 2024) as both the context assembler and the source of the strongest non-adaptive baseline (a DSPy BootstrapFewShot static prompt), and evaluate it across three verifiable task domains -- tool-use workflows, code generation (HumanEval), and multi-hop retrieval QA (HotpotQA) -- and two model providers (a local Ollama model and AWS Bedrock). We release the harness-control-system code, the cross-domain verifiable task suite, the full trajectory/reward-decomposition logs from training, and a provider-agnostic deployment recipe for applying this to a new organization's domain and verification setup.
Chinese Translation
生产LLM代理越来越多地由一个冻结模型构成，该模型被包裹在一个框架中：一个提示模板、一套工具、一个记忆/检索层、一个规划策略和一个验证政策。2026年的两个系统，Meta-Harness（Lee et al., 2026）和HyperAgents（Meta AI, 2026），表明这个框架本身可以被一个代理提议者优化甚至自我重写——代价是一个昂贵的代码搜索循环或不受限制的自我修改代码，而这两者都无法审计或与完全黑箱模型API一起使用。我们采取一个更狭窄、更受限的立场：将框架视为一个小而固定的人类可读动作空间，并通过经典的样本高效强化学习（一个$ ext{ε}$-贪婪上下文强盗和REINFORCE）在线学习一个策略，该策略以多目标奖励（任务成功、验证者得分、政策合规性、成本、延迟和不支持声明的惩罚）进行评分。我们使用DSPy（Khattab et al., 2024）作为上下文组装器和最强非自适应基线（一个DSPy BootstrapFewShot静态提示）来实例化这个控制系统，并在三个可验证的任务领域——工具使用工作流、代码生成（HumanEval）和多跳检索问答（HotpotQA）——以及两个模型提供者（一个本地Ollama模型和AWS Bedrock）上进行评估。我们发布了框架控制系统代码、跨领域可验证任务套件、训练过程中的完整轨迹/奖励分解日志，以及一个与提供者无关的部署方案，以便将其应用于新组织的领域和验证设置。
cs.AI / 67 / 2607.25422
Salient Knowledge Pathways: Sparse Cross-Modal Routing for Efficient Knowledge-Intensive Multimodal Question Answering
显著知识路径：高效知识密集型多模态问答的稀疏跨模态路由
Abstract
Knowledge-intensive multimodal question answering (KI-MMQA) sits at the intersection of three expensive primitives: long visual token sequences, dense retrieval over large external corpora, and full cross-modal fusion. Existing systems pay all three costs uniformly per query, even though only a small fraction of visual content and retrieved knowledge is actually relevant to any given question. We introduce SKIP (Salient Knowledge-Injected Pathways), a unified inference architecture that routes computation along sparse pathways jointly conditioned on the question, the image, and a difficulty estimate. SKIP combines question-guided visual token pruning, region-conditional sparse retrieval, bipartite sparse cross-attention, and speculative knowledge verification with an adaptive budget controller that allocates compute proportional to predicted question difficulty. We derive an information-bottleneck bound showing that the optimal visual sparsity rate scales as $O(1/\sqrt{N})$ under realistic question-image mutual-information assumptions, with retained accuracy guarantees. Across five KI-MMQA benchmarks (OK-VQA, A-OKVQA, InfoSeek, Encyclopedic-VQA, and ViQuAE), SKIP matches or exceeds the accuracy of strong dense baselines while using $3.4$--$6.8\times$ fewer FLOPs and $2.7\times$ less end-to-end latency. Code available at: https://pmlrbd.github.io/skip/
Chinese Translation
知识密集型多模态问答（KI-MMQA）位于三个高成本原语的交叉点：长视觉标记序列、对大型外部语料库的密集检索，以及完全的跨模态融合。现有系统在每个查询中均匀支付这三项成本，尽管实际上只有少量视觉内容和检索知识与任何给定问题相关。我们提出了SKIP（显著知识注入路径），一种统一的推理架构，通过共同条件化于问题、图像和难度估计，沿稀疏路径路由计算。SKIP结合了基于问题的视觉标记修剪、区域条件的稀疏检索、二分稀疏跨注意力和投机性知识验证，并配备了一个自适应预算控制器，按预测的问题难度分配计算资源。我们推导出信息瓶颈界限，表明在现实的问题-图像互信息假设下，最佳视觉稀疏率的比例为$O(1/ ext{sqrt}{N})$，并保持准确性保证。在五个KI-MMQA基准（OK-VQA、A-OKVQA、InfoSeek、Encyclopedic-VQA和ViQuAE）上，SKIP的准确性与强大的密集基线相匹配或超过，同时使用$3.4$--$6.8 imes$更少的FLOPs和$2.7 imes$更少的端到端延迟。代码可在：https://pmlrbd.github.io/skip/
cs.AI / 68 / 2607.25425
The Disruptive Impact of Large Language Models on Capture the Flag Competitions and the Path Toward Fair Play
大型语言模型对夺旗比赛的颠覆性影响及公平竞争的路径
Abstract
Capture the Flag (CTF) competitions are among cybersecurity's most effective training grounds, developing practical skill across cryptography, web exploitation, and binary exploitation. Large language models (LLMs) can now solve a growing share of challenges with minimal human input, raising urgent questions about fairness, the validity of rankings, and whether participation still delivers the learning that justifies the effort. This paper reports a mixed-methods study of LLM impact on modern CTFs, combining a synthesis of published benchmarks, including a recent government evaluation, case studies of live competition across three challenge categories, structured observation of the public channels where the community debates AI use, and semi-structured interviews with experienced players and organisers. We map the current human-machine capability boundary by category, showing that easy and intermediate challenges in cryptography, web, and binary exploitation are now reliably automated while narrower sub-categories continue to resist. We find that community disagreement about whether AI should be permitted is downstream of an undeclared prior question: what a competition is for. Against this backdrop we contribute a four-component safeguard framework, combining tiered competition divisions, LLM-resistant challenge design, telemetry used investigatively, and a draft community code of conduct, together with a decision tool that ties the combination of safeguards to a competition's declared purpose. The argument reaches beyond CTFs to any setting in cybersecurity where a demonstrated result is taken as evidence of an underlying ability.
Chinese Translation
夺旗比赛（Capture the Flag, CTF）是网络安全领域最有效的训练场之一，能够培养参与者在密码学、网络利用和二进制利用等方面的实用技能。大型语言模型（Large Language Models, LLMs）现在能够以最小的人类输入解决越来越多的挑战，这引发了关于公平性、排名有效性以及参与是否仍然能带来值得付出的学习成果的紧迫问题。本文报告了一项关于LLM对现代CTF影响的混合方法研究，结合了已发布基准的综合分析，包括最近的政府评估、三个挑战类别的现场比赛案例研究、对社区讨论AI使用的公共频道的结构化观察，以及与经验丰富的选手和组织者的半结构化访谈。我们按类别绘制了当前人机能力边界，显示密码学、网络和二进制利用中的简单和中等难度挑战现在已可靠地实现自动化，而较窄的子类别仍然抵抗这种趋势。我们发现，关于是否应允许AI的社区分歧源于一个未被明确提出的先前问题：比赛的目的是什么。在此背景下，我们提出了一个由四个组成部分构成的保障框架，结合了分级比赛分区、抗LLM挑战设计、用于调查的遥测以及一份草拟的社区行为准则，此外还提供了一个将保障组合与比赛声明目的联系起来的决策工具。该论点超越了CTF，适用于网络安全领域的任何场合，在这些场合中，展示的结果被视为潜在能力的证据。
cs.AI / 69 / 2607.25446
Toward an Organizational Science of Multi-Agent LLM Systems: Decoupling Who, How, and Which Algorithm
迈向多智能体大语言模型系统的组织科学：解耦谁、如何以及哪个算法
Abstract
Multi-agent frameworks built on large language models (LLMs) routinely entangle three logically distinct concerns: who is on the team (organization), how members align (coordination), and which algorithm fuses their work (collaboration protocol). IMACS (Intelligent Multi-Agent Collaboration System) separates the three into orthogonal, independently swappable layers. Classic organizational theory (Belbin roles, Mintzberg coordination, RACI accountability) becomes executable, validated configuration, and the framework places six published collaboration algorithms behind a common interface while exposing roles, coordination, and accountability as independently configurable factors. We use this separation to conduct controlled comparisons in which organizational assignments vary while the collaboration protocol is held fixed. It also turns protocol choice into a variable that can be learned: Adaptive Org Routing, a contextual-bandit meta-protocol, selects a protocol per task under an explicit quality-cost tradeoff, outperforms every fixed protocol in a controlled study, and trains online on real benchmark and LLM-judge rewards. The ablations expose a mechanism. Accountability placement changes outcomes exactly when the protocol routes the deliverable through the accountable agent, and the winning placement flips across model families, so organizational design cannot be hard-coded; it must be revalidated, or learned, for each model binding.
Chinese Translation
基于大语言模型（LLMs）的多智能体框架通常将三个逻辑上不同的关注点纠缠在一起：团队成员是谁（组织）、成员如何协调（协调）以及哪个算法融合他们的工作（协作协议）。智能多智能体协作系统（IMACS）将这三者分离为正交的、独立可替换的层次。经典的组织理论（Belbin角色、Mintzberg协调、RACI问责）变成可执行的、经过验证的配置，而该框架将六种已发布的协作算法置于一个共同接口之后，同时将角色、协调和问责作为独立可配置的因素。我们利用这种分离进行受控比较，其中组织分配变化而协作协议保持不变。这也将协议选择转变为一个可以学习的变量：自适应组织路由（Adaptive Org Routing），一种上下文赌博元协议，在明确的质量-成本权衡下为每个任务选择协议，在受控研究中超越了每个固定协议，并在真实基准和LLM评判奖励上进行在线训练。消融实验揭示了一种机制。当协议通过负责代理路由可交付成果时，问责的安排恰好改变结果，而获胜的安排在模型家族之间翻转，因此组织设计不能被硬编码；它必须为每个模型绑定重新验证或学习。
cs.AI / 70 / 2607.25471
TRWH: A Text-Driven Random Walk Heterogeneous GNN for Semantic-Aware Sparse Recommendation
TRWH：一种基于文本驱动的随机游走异构图神经网络用于语义感知稀疏推荐
Abstract
Graph Neural Networks (GNNs) and Large Language Models (LLMs) have each advanced recommendation systems by modeling structural and semantic signals, respectively. However, integrating their complementary strengths remains challenging, particularly in sparse settings where maintaining semantic precision is critical. We propose TRWH (Text-driven Random Walk Heterogeneous Graph Neural Network), a novel framework that fuses LLM-generated textual profiles with heterogeneous graph structures through strategic random walk augmentation. TRWH consists of three core components: (1) Embedding Creation, which produces user and item representations using both Word2Vec and LLM-based profiling; (2) a Heterogeneous Graph Neural Network (HeteroGNN) that propagates information across multi-relational edges; and (3) Random Walk-based Path Construction, which enriches sparse graphs with second-order user-user and item-item links. Experiments on the Amazon-2023 Fashion (2M users, 825K items) and Beauty (631K users, 112K items) datasets demonstrate that TRWH achieves substantial performance gains over state-of-the-art methods, including 80.0% RMSE and 52.6% MAE reductions on Fashion, and 25.7% and 10.8% improvements on Beauty. Notably, while random walks improve performance with traditional embeddings, they can dilute the nuanced representations learned by LLMs, underscoring the importance of adaptive integration strategies.
Chinese Translation
图神经网络（GNNs）和大型语言模型（LLMs）分别通过建模结构和语义信号推动了推荐系统的发展。然而，特别是在稀疏环境中，整合它们的互补优势仍然具有挑战性，因为保持语义精确性至关重要。我们提出了TRWH（文本驱动随机游走异构图神经网络），这是一个新颖的框架，通过战略性随机游走增强，将LLM生成的文本档案与异构图结构融合。TRWH由三个核心组件组成：（1）嵌入创建，使用Word2Vec和基于LLM的档案生成用户和物品表示；（2）异构图神经网络（HeteroGNN），在多关系边上传播信息；（3）基于随机游走的路径构建，利用二阶用户-用户和物品-物品链接丰富稀疏图。在Amazon-2023时尚（200万用户，82.5万物品）和美容（63.1万用户，11.2万物品）数据集上的实验表明，TRWH在性能上显著优于最先进的方法，包括在时尚领域实现80.0%的均方根误差（RMSE）和52.6%的平均绝对误差（MAE）减少，以及在美容领域的25.7%和10.8%的提升。值得注意的是，尽管随机游走通过传统嵌入提高了性能，但它们可能会稀释LLMs学习的细微表示，这突显了自适应整合策略的重要性。
cs.AI / 71 / 2607.25474
Balancing multiscale similarity and cartographic constraints: A similarity-driven optimization framework for line generalization
平衡多尺度相似性与制图约束：一种基于相似性的线条概化优化框架
Abstract
Cartographic generalization is essential for generating multiscale map representations by balancing information preservation and cartographic readability. However, automated generalization remains challenging because existing approaches often treat spatial similarity evaluation, cartographic constraints, and parameter optimization as separate processes, limiting adaptive and interpretable control across scales. This study formulates cartographic generalization as a constrained multiscale similarity optimization problem and proposes a similarity-driven framework for adaptive generalization control. The framework integrates multiscale spatial similarity as an optimization objective to quantify representation consistency between original and generalized data, while incorporating cartographic constraints to regulate readability, smoothness, and geometric validity. A unified objective function is optimized to automatically identify scale-dependent parameter configurations for different generalization algorithms. Experiments using multiple line simplification algorithms, target scales, and similarity measures, including geometric, structural, and learning-based metrics, demonstrate that the proposed framework achieves an effective balance between similarity preservation and cartographic abstraction. The results further show that combining similarity optimization with cartographic constraints provides more consistent and interpretable parameter control than relying on similarity evaluation alone. This study provides a unified optimization perspective that connects similarity assessment, constraint modeling, and algorithm control, contributing to adaptive and automated cartographic generalization.
Chinese Translation
制图概化对于通过平衡信息保留和制图可读性生成多尺度地图表示至关重要。然而，自动化概化仍然面临挑战，因为现有方法通常将空间相似性评估、制图约束和参数优化视为独立的过程，从而限制了跨尺度的自适应和可解释控制。本研究将制图概化公式化为一个受约束的多尺度相似性优化问题，并提出了一种基于相似性的自适应概化控制框架。该框架将多尺度空间相似性作为优化目标，以量化原始数据与概化数据之间的表示一致性，同时结合制图约束以调节可读性、平滑性和几何有效性。一个统一的目标函数被优化，以自动识别不同概化算法的尺度依赖参数配置。使用多种线条简化算法、目标尺度和相似性度量（包括几何、结构和基于学习的指标）的实验表明，所提出的框架在相似性保留与制图抽象之间实现了有效平衡。结果进一步表明，将相似性优化与制图约束相结合提供了比单独依赖相似性评估更一致和可解释的参数控制。本研究提供了一个统一的优化视角，连接了相似性评估、约束建模和算法控制，为自适应和自动化的制图概化做出了贡献。
cs.AI / 72 / 2607.25484
Finding Optimal Cost-Bounded Plan Reductions: Refined Model
寻找最佳成本约束计划简化：改进模型
Abstract
In some real applications a plan may later become unfeasible due to newly imposed budget constraints, yet, at the same time, using only the original actions of the plan and their order is mandatory. In this paper, we study the problem of extracting, from a precomputed plan, a valid subplan that maximizes utility while respecting a cost bound. Each goal is given a utility value and the plan is reduced by removing actions that support low-utility goals, while preserving both executability and the original action order. We show the decision variant is NP-complete and propose two exact methods to solve it: one via oversubscription planning (OSP) and another via Integer Linear Programming (ILP). This paper extends our previous work published at ICAPS 2026 (Del Toro, Fuentetaja, and Garc\'ia-Olaya 2026b). While the core framework remains as introduced there, we further introduce a refined ILP formulation that significantly decreases the model size and improves computational efficiency.
Chinese Translation
在某些实际应用中，计划可能由于新施加的预算限制而变得不可行，但同时，必须仅使用计划的原始动作及其顺序。本文研究了从预计算计划中提取有效子计划的问题，该子计划在遵循成本约束的同时最大化效用。每个目标都有一个效用值，计划通过移除支持低效用目标的动作来减少，同时保持可执行性和原始动作顺序。我们证明了该决策变体是 NP 完全的，并提出了两种精确的方法来解决它：一种通过超订阅规划（OSP），另一种通过整数线性规划（ILP）。本文扩展了我们在 ICAPS 2026 上发表的先前工作（Del Toro, Fuentetaja, 和 García-Olaya 2026b）。尽管核心框架与之前介绍的一致，我们进一步引入了一种改进的 ILP 公式，显著减少了模型规模并提高了计算效率。
cs.AI / 73 / 2607.25485
PatientAgentBench: A Benchmark Framework for Evaluating Patient-Facing Health AI Agents
PatientAgentBench：评估面向患者的健康人工智能代理的基准框架
Abstract
Health AI is evolving from answering questions to agentic systems that converse with patients, reason about health records, and act on their behalf. Primary care guards against diagnostic errors and unsafe care; agents assisting in this domain warrant evaluation against the same risks. Current benchmarks focus on medical knowledge, assessed through isolated question-answering or clinician-facing tasks. PatientAgentBench benchmarks patient-facing agentic healthcare; it evaluates a foundation model, wrapped in an agent with a sandbox of healthcare tools, conversing with a simulated patient. Each conversation is scored by an LLM-as-a-Jury across six dimensions via over a hundred conversation-agnostic, clinician-grounded criteria. To validate alignment, licensed clinicians annotated shared conversations, yielding 79-93% adjacent agreement between jury and expert raters, on par with or exceeding clinician inter-rater agreement. We benchmarked 10 models across four families on the same 1,200 scenarios and found clinical gaps. Triage quality is the most discriminating dimension: pass rates rise from 32% for the weakest models to 88% for the strongest, with agents often acting on administrative requests without clinical screening. Clinical safety and workflow accuracy follow the same pattern: the weakest models fail often, fabricating unexecuted actions, while frontier models fail on only 1-3% of cases, from unverified tool outputs and omitted crisis resources in an emergency. More capable models narrow these gaps but do not close them; the strongest scores only 4.25 of 5 overall. These failures surface only in sustained, tool-using conversations against realistic patient records, confirming that static benchmarks are insufficient as healthcare agentic systems gain autonomy. We release the framework as a reproducible, clinician-validated evaluation standard to help the field close this gap.
Chinese Translation
健康人工智能正从回答问题发展为能够与患者对话、推理健康记录并代表患者行动的代理系统。初级护理防止诊断错误和不安全的护理；在这一领域协助的代理需要针对相同风险进行评估。目前的基准主要集中在医学知识的评估上，通过孤立的问答或面向临床医生的任务进行评估。PatientAgentBench基准评估面向患者的代理医疗；它评估一个基础模型，该模型被包装在一个具有医疗工具沙箱的代理中，与模拟患者进行对话。每次对话通过一个大型语言模型（LLM）作为评审，在六个维度上通过超过一百个与对话无关的、基于临床的标准进行评分。为了验证一致性，持证临床医生对共享对话进行了注释，评审与专家评分者之间的邻近一致性达到了79-93%，与临床医生之间的评分一致性相当或更高。我们在同一1200个场景上对10个模型进行了基准测试，发现了临床差距。分诊质量是最具区分性的维度：通过最弱模型的合格率为32%，而最强模型的合格率上升至88%，代理在没有临床筛查的情况下通常会执行行政请求。临床安全性和工作流程准确性遵循相同的模式：最弱模型经常失败，虚构未执行的操作，而前沿模型仅在1-3%的案例中失败，原因是未验证的工具输出和在紧急情况下遗漏的危机资源。更强大的模型缩小了这些差距，但并未完全消除；最强模型的整体评分仅为4.25分（满分5分）。这些失败仅在持续使用工具的对话中与真实患者记录相对比时显现，确认了静态基准在医疗代理系统获得自主性时的不足。我们发布该框架作为可重复的、经过临床验证的评估标准，以帮助该领域缩小这一差距。
cs.AI / 74 / 2607.25487
CoTinyVLA: Chain-of-Thought Distillation for a Sub-Billion-Parameter Vision-Language-Action Model
CoTinyVLA：用于亚十亿参数视觉-语言-动作模型的思维链蒸馏
Abstract
Vision-Language-Action (VLA) models translate natural-language commands into robot action sequences, but leading systems on the LIBERO-Plus robustness benchmark use three- to seven-billion-parameter backbones whose memory demands can exceed embedded robotic budgets. We present CoTinyVLA, a 0.9B-parameter action model on a Qwen3.5-0.8B backbone that obtains that robustness by structuring supervision instead of enlarging the model. Three components target different axes of the problem: dual-view temporal input of 16 history frames per step with textual camera and time markers; hierarchical chain-of-thought (CoT) distillation from a 35B teacher into an episode-level Plan and a chunk-level Think span over task phase, gripper state and next subaction; and paraphrase augmentation expanding 40 base commands into 800 variants. On LIBERO-Plus, spanning 10,030 perturbed tasks across seven perturbation dimensions, CoTinyVLA reaches 90.8% on Spatial, 87.3% on Object, 86.6% on Goal and 80.7% on Long, leading the strongest 7B baseline on all four suites by 4.7, 2.8, 15.9 and 3.0 points, with every margin interval excluding zero. The gains concentrate on the hardest axes of the benchmark: across the eleven published baselines none exceeds 53.2% on Robot Initial States in any suite, whereas CoTinyVLA reaches 73.6% on Goal against 39.9% for the strongest baseline. Ablations show the three components to be separable by perturbation axis, and at a matched image budget how frames are divided between the two cameras and across time accounts for 8.6 points on its own. Closed-loop inference peaks at 2.25 GiB of allocated GPU memory, and paired interventions show the episode Plan to be load-bearing: replacing it with an empty or contradictory span costs 40 to 45 points of success. Structured supervision thus lets a 0.9B backbone exceed all of them. Code: https://github.com/BrainJellyPie/CoTinyVLA
Chinese Translation
视觉-语言-动作（VLA）模型将自然语言命令转化为机器人动作序列，但在LIBERO-Plus鲁棒性基准测试中，领先系统使用的三到七十亿参数的主干网络，其内存需求可能超出嵌入式机器人预算。我们提出了CoTinyVLA，这是一个基于Qwen3.5-0.8B主干的0.9B参数动作模型，通过构建监督结构而非扩大模型来获得鲁棒性。三个组成部分针对问题的不同方面：每步16帧历史图像的双视角时间输入，配有文本摄像头和时间标记；从35B教师模型进行的层次化思维链（CoT）蒸馏，生成任务阶段、抓取器状态和下一个子动作的情节级计划和块级思维跨度；以及将40个基本命令扩展为800个变体的释义增强。在LIBERO-Plus上，涵盖了跨七个扰动维度的10,030个扰动任务，CoTinyVLA在空间（Spatial）任务上达到90.8%，在物体（Object）任务上达到87.3%，在目标（Goal）任务上达到86.6%，在长时间（Long）任务上达到80.7%，在所有四个测试套件中领先最强的7B基线分别4.7、2.8、15.9和3.0个百分点，且每个差距区间均不包含零。这些增益集中在基准测试中最困难的维度：在发布的十一条基线中，没有一条在任何测试套件中超过53.2%的机器人初始状态，而CoTinyVLA在目标任务上达到73.6%，而最强基线仅为39.9%。消融实验表明这三个组成部分可以通过扰动维度分开，并且在匹配的图像预算下，如何在两个摄像头之间以及跨时间划分帧本身就能贡献8.6个百分点。闭环推理的GPU内存峰值分配为2.25 GiB，配对干预显示情节计划是承载性结构：用空或矛盾的跨度替换它会导致成功率下降40到45个百分点。因此，结构化监督使得0.9B主干超越了所有这些。代码链接：https://github.com/BrainJellyPie/CoTinyVLA
cs.AI / 75 / 2607.25529
Are the High-weight Neurons the Important Ones in Image Classification Neural Networks?
高权重神经元在图像分类神经网络中是重要的神经元吗？
Abstract
As neural network models for image classification advance, neurons play critical roles in pruning, backdoor defense, and interpretability. Yet existing work lacks clarity on the weight-importance relationship. We address this with a neuron importance assessment method using three experiments: quantifying overlap between high-weight and accuracy-impacting neurons, analyzing high-weight neuron perturbation effects, and testing post-retraining accuracy after high-weight neuron ablation. Experiments on CIFAR-10 and Mini-ImageNet reveal key patterns. Overlap analysis shows top 10\% high-weight neurons overlap with important ones by only about 25\% at maximum, dropping further in subsequent intervals. Perturbation tests find top 10\% high-weight neurons cause 45-80\% accuracy degradation under certain operations compared to 3-7\% for random perturbations, but a third of them show minimal impact. Ablation-retraining results show removing top 10\% high-weight neurons leaves accuracy 10-20\% below baseline with no recovery, while ablating top 0.1\% allows near-full recovery. Notably, some low-weight intervals show 10-17\% degradation when perturbed, comparable to mid-range high-weight neurons. These results confirm not all high-weight neurons are important: their importance is nonlinear. Low-weight neurons also contribute significantly. This challenges weight-importance equivalence, offering refined neuron role insights. It supports applications like encryption prioritizing critical high-weight neurons and pruning removing non-critical ones, advancing neural network analysis.
Chinese Translation
随着图像分类神经网络模型的发展，神经元在剪枝、后门防御和可解释性方面发挥着关键作用。然而，现有研究对权重与重要性之间的关系缺乏明确性。我们通过一种神经元重要性评估方法来解决这一问题，采用了三项实验：量化高权重神经元与影响准确性的神经元之间的重叠，分析高权重神经元的扰动效应，以及测试高权重神经元消融后的再训练准确性。对CIFAR-10和Mini-ImageNet的实验揭示了关键模式。重叠分析表明，前10%的高权重神经元与重要神经元的重叠最多仅约为25%，在后续区间进一步下降。扰动测试发现，前10%的高权重神经元在某些操作下导致的准确性下降为45-80%，而随机扰动的准确性下降仅为3-7%，但其中三分之一的神经元显示出最小影响。消融再训练结果显示，去除前10%的高权重神经元使得准确性比基线低10-20%，且没有恢复，而消融前0.1%的神经元则几乎可以完全恢复。值得注意的是，一些低权重区间在扰动时显示出10-17%的准确性下降， comparable to mid-range high-weight neurons。这些结果确认并非所有高权重神经元都是重要的：它们的重要性是非线性的。低权重神经元也显著贡献。这挑战了权重与重要性等价的观点，提供了更精细的神经元角色洞察。这支持了如加密等应用，优先考虑关键的高权重神经元，并在剪枝中去除非关键神经元，推动神经网络分析的发展。
cs.AI / 76 / 2607.25532
Entangled by Design: Spurious Intra-Variable Signal Routing in Tabular In-Context Learners
设计中的纠缠：表格上下文学习者中的虚假变量信号路由
Abstract
Consider a model trained at a single hospital to predict patient recovery, where the measured feature $X$ bundles the patient's true health signal ($C$) with a systematic artefact from that hospital's equipment ($S$). Within that hospital, the artefact correlates with outcomes through unmeasured confounders such as patient demographics; an in-context learner rationally routes predictions through $S$, not $C$, and fails silently when deployed at a new hospital with different equipment. We formalise this as \emph{spurious routing in composite representations}: when a feature $X = [C;\,\alpha S;\,\eta]$ encodes a causal signal $C$ and a spurious signal $S$ in distinct subspaces, the ICL cannot determine which drives predictions. We prove that under ridge ICL, a linear in-context learner, this routing is unavoidable regardless of context size; TabPFN, a state-of-the-art pretrained tabular ICL model, shows qualitatively consistent behaviour empirically. We derive a closed-form characterisation, $\mathrm{CSR} \propto \rho_S/\rho_C$, confirmed at $r = 0.997$ for linear ICL and $r = 0.979$ for TabPFN. Contrary to intuition, larger context sharpens commitment to the dominant in-context signal, amplifying spurious routing by up to $1.74\times$; in the high-spurious corner, more expressive models show greater vulnerability empirically ($+2.22$ CSR gap at high entanglement). We introduce two lightweight mitigations: environment-stratified context construction and S-swap augmentation, that require only weak environment labels and no knowledge of the causal partition. S-swap reduces spurious routing by $74\%$ for linear ICL and $98.8\%$ for TabPFN, with TabPFN's causal sensitivity increasing $8.4\times$ simultaneously: the model does not become agnostic, it reroutes through the causal signal.
Chinese Translation
考虑一个在单一医院训练的模型，用于预测患者恢复情况，其中测量特征 $X$ 将患者的真实健康信号 ($C$) 与该医院设备的系统性伪影 ($S$) 进行了捆绑。在该医院内，伪影通过未测量的混杂因素（如患者人口统计信息）与结果相关联；上下文学习者理性地通过 $S$ 而非 $C$ 路由预测，并在部署到设备不同的新医院时悄然失败。我们将其形式化为 extit{复合表示中的虚假路由}：当特征 $X = [C;\,eta S;\, heta]$ 在不同的子空间中编码因果信号 $C$ 和虚假信号 $S$ 时，ICL 无法确定哪个驱动预测。我们证明在岭回归 ICL 下，作为线性上下文学习者的这种路由是不可避免的，无论上下文大小如何；最先进的预训练表格 ICL 模型 TabPFN 在经验上显示出定性一致的行为。我们推导出一个封闭形式的表征，$ ext{CSR} ext{ } ext{propto} ext{ }
ho_S/
ho_C$，在线性 ICL 下确认 $r = 0.997$，在 TabPFN 下确认 $r = 0.979$。与直觉相反，较大的上下文增强了对主导上下文信号的承诺，虚假路由最多放大 $1.74 imes$；在高虚假角落，更具表现力的模型在经验上显示出更大的脆弱性（在高纠缠下 CSR 差距为 $+2.22$）。我们引入了两种轻量级的缓解措施：环境分层上下文构建和 S-swap 增强，这仅需弱环境标签且不需要因果划分的知识。S-swap 在线性 ICL 下减少了 $74 ext{ extperthousand}$ 的虚假路由，在 TabPFN 下减少了 $98.8 ext{ extperthousand}$，同时 TabPFN 的因果敏感性增加了 $8.4 imes$：模型并未变得无知，而是通过因果信号重新路由。
cs.AI / 77 / 2607.25546
From Training to Deployment: Post-Hoc Causal Feature Identification via Sensitivity Ratios
从训练到部署：通过敏感性比率进行事后因果特征识别
Abstract
Given a model that is already trained, which features does it rely on causally versus spuriously? Existing methods require access to the training procedure and cannot answer this post-hoc. We introduce the \textbf{Normalised Sensitivity Ratio~(NSR)}, a post-hoc, model-agnostic diagnostic for this question under a structured-shift regime: environments differ primarily in the mean of spurious features while the causal mechanism and causal marginals remain stable, as in multi-site clinical data or multi-batch genomics. Within this regime, causal features induce constant model sensitivity across environments while spurious features track shift. NSR formalises this as the squared coefficient of variation of per-environment sensitivity. Under a linear structural causal model (SCM) with $K\ge3$ non-degenerate environments, NSR achieves exact identification (Theorem~1). We fully characterise failure: weak shifts ($O(\varepsilon^4)$ collapse), degenerate geometry, and proxy attenuation ($O((1-\alpha)^4)$), giving practitioners quantitative criteria for assessing whether the regime holds. Finite-sample rates are $O_p(n^{-1})$ under the null and $O_p(n^{-1/2})$ under the alternative. Experiments confirm all theoretical predictions on synthetic data (area under the ROC curve [AUROC] $= 1.000$ under conditions satisfying the regime), show consistent rankings across five model families (Kendall $\tau\ge0.529$), and recover six of eight causal features on bike-sharing data (Precision@7 $= 0.75$) without modifying any trained model.
Chinese Translation
给定一个已经训练好的模型，它在因果上依赖于哪些特征而非虚假特征？现有方法需要访问训练过程，无法事后回答这个问题。我们引入了 extbf{归一化敏感性比率~(NSR)}，这是一个在结构性变化环境下的事后、模型无关的诊断工具：环境主要在虚假特征的均值上有所不同，而因果机制和因果边际保持稳定，类似于多站点临床数据或多批次基因组数据。在这种环境下，因果特征在不同环境中引起模型敏感性的恒定变化，而虚假特征则跟踪变化。NSR将此形式化为每个环境敏感性的平方变异系数。在具有$K
ge3$个非退化环境的线性结构因果模型（SCM）下，NSR实现了精确识别（定理~1）。我们全面表征了失败情况：弱变化（$O( ext{ε}^4)$崩溃）、退化几何和代理衰减（$O((1- ext{α})^4)$），为从业者提供了评估该环境是否成立的定量标准。在原假设下，有限样本率为$O_p(n^{-1})$，在备择假设下为$O_p(n^{-1/2})$。实验验证了在合成数据上所有理论预测（在满足该环境条件下，ROC曲线下面积[AUROC] $= 1.000$），在五个模型家族中显示出一致的排名（Kendall $ au
ge0.529$），并在共享单车数据上恢复了八个因果特征中的六个（Precision@7 $= 0.75$），而无需修改任何已训练的模型。
cs.AI / 78 / 2607.25554
Distilling Temporal Search and Reasoning: Evolving LLMs for Future Prediction via Harness-Assisted Efficient Data Synthesis
提炼时间搜索与推理：通过辅助工具高效数据合成演化大语言模型以进行未来预测
Abstract
Future event prediction carries broad social impact yet remains challenging. SOTA approaches augment LLMs with external agent frameworks whose predictive capability vanishes once the harness is removed. While recent Tool-Integrated Reasoning (TIR) internalizes deep search for multi-hop retrieval of facts, forecasting further demands temporal search and reasoning over historical trends and dynamic shifts. The key obstacle is data: historical queries induce temporal leakage that degrades forecasting into retrieval. Prior works either freeze information gathering with static observations, or rely on rejection sampling or unresolved fresh queries that discard vast amounts of data, degrading synthesis efficiency. We propose a time-truncation harness that enforces a temporal cut-off at every turn, enabling TIR-style sampling from historical events, reducing temporal leakage and reliance of rejection sampling or unsolved queries, increasing the sampling efficiency. We further build a large-scale corpus and a process-based metric and show that our harness naturally induces a broader temporal breadth of search and raises the proportion of high-quality data, further increasing the efficiency and reducing the reliance on complex rubrics. Distillation experiments show that students trained on harness-intervened data achieve the best performance, demonstrating harness-assisted model evolving that turns higher quality temporal search and reasoning data into a parametric advancement of the students.
Chinese Translation
未来事件预测具有广泛的社会影响，但仍然面临挑战。现有的最先进方法通过外部代理框架增强大语言模型（LLMs），但一旦去除辅助工具，其预测能力便消失。虽然近期的工具集成推理（Tool-Integrated Reasoning, TIR）将深度搜索内化以实现多跳事实检索，但预测还需要对历史趋势和动态变化进行时间搜索与推理。关键障碍在于数据：历史查询会导致时间泄漏，从而将预测降级为检索。以往的研究要么通过静态观察冻结信息收集，要么依赖于拒绝采样或未解决的新查询，这会丢弃大量数据，降低合成效率。我们提出了一种时间截断辅助工具，在每一步强制施加时间截止，允许从历史事件中进行TIR风格的采样，减少时间泄漏和对拒绝采样或未解决查询的依赖，提高采样效率。我们进一步构建了一个大规模语料库和基于过程的度量，显示我们的辅助工具自然引导更广泛的时间搜索范围，并提高高质量数据的比例，进一步提高效率并减少对复杂标准的依赖。蒸馏实验表明，在辅助工具干预的数据上训练的学生模型表现最佳，展示了辅助工具支持的模型演化，将更高质量的时间搜索与推理数据转化为学生模型的参数化进步。
cs.AI / 79 / 2607.25560
Agent Skills Matter: Inferring Proprietary Skills from Execution Trajectories
代理技能的重要性：从执行轨迹推断专有技能
Abstract
Agent skills package reusable procedures that improve downstream performance. Their lightweight, portable form enables marketplace monetization and private deployment behind cloud-hosted agent interfaces, giving providers incentives to keep high-value skills proprietary. Yet hiding the artifacts does not conceal their behavioral effects, which remain observable in execution trajectories and form a behavioral side channel. We define this exposure as Skill Leakage: reconstructing proprietary skills from trajectories elicited by benign queries, without reference answers or success labels. We introduce SigLeak, a black-box framework that exploits recurring skill signatures in agent behavior. It constructs diverse, decision-rich diagnostic tasks, contrasts matched skill-enabled and skill-disabled trajectories, and iteratively refines a reconstructed skill from the isolated patterns. Across five scenarios, three model families, and three agent frameworks, SigLeak outperforms or matches three baselines in nearly every setting. It raises the success rate by 6.88 percentage points over the skill-disabled reference on average and achieves the highest overall SkillSim, our metric for coarse- and fine-grained semantic similarity. These results show that benign execution trajectories can expose proprietary procedural knowledge. The code is available at https://anonymous.4open.science/r/SigLeak-D1DB.
Chinese Translation
代理技能封装了可重用的程序，这些程序能够提升后续性能。它们轻量、可移植的形式使得市场化变得可行，并且可以在云托管的代理接口后进行私有部署，从而激励提供者将高价值技能保持为专有。然而，隐藏这些工件并不能掩盖其行为效应，这些效应在执行轨迹中依然可观察，并形成一种行为侧信道。我们将这种暴露定义为技能泄漏（Skill Leakage）：从由良性查询引发的轨迹中重构专有技能，而无需参考答案或成功标签。我们引入了SigLeak，一个利用代理行为中重复技能特征的黑箱框架。它构建了多样化、决策丰富的诊断任务，比较匹配的技能启用和技能禁用的轨迹，并迭代地从孤立模式中细化重构的技能。在五种场景、三种模型系列和三种代理框架中，SigLeak在几乎每种设置下都超越或匹配了三个基线。它的成功率平均提高了6.88个百分点，相较于技能禁用参考，并且在我们用于粗粒度和细粒度语义相似性的指标SkillSim中达到了最高的整体表现。这些结果表明，良性的执行轨迹可以暴露专有的程序知识。代码可在 https://anonymous.4open.science/r/SigLeak-D1DB 获取。
cs.AI / 80 / 2607.25576
Matrix-Free Photoacoustic Image Reconstruction via Sensor-Token Self-Attention
无矩阵光声图像重建通过传感器令牌自注意力
Abstract
Photoacoustic tomography (PAT) combines the optical absorption contrast of biological tissue with the spatial resolution of ultrasound, yet recovering the initial pressure distribution from sparse-view sensor measurements remains an ill-posed inverse problem. Iterative compressive-sensing solvers and unrolled deep networks both retain a dependence on the system matrix at inference, which leaves real-time clinical reconstruction computationally expensive. This paper proposes the Sensor Attention Network (SAN), a Transformer-based architecture that treats the full time series of each sensor as a token and maps raw measurements directly to the reconstructed image without invoking the system matrix at inference. For training and benchmarking, an analytical k-space H-matrix is constructed and validated against the k-Wave pseudo-spectral solver under matched geometry, achieving a mean per-sensor Pearson correlation of 0.919 +/- 0.049, with k-space apodization and Gaussian temporal damping acting synergistically to reduce the energy-normalized mismatch by 49%. Trained with a vessel-weighted loss on 488 augmented samples and evaluated on 46 held-out samples against ISTA, split-Bregman total variation (SBTV), and learned ISTA (LISTA), SAN attains the highest mean SSIM (0.522) and PSNR (22.09 dB) and the lowest NMSE (0.233). Paired t-tests and Wilcoxon signed-rank tests confirm the superiority of SAN over LISTA on PSNR, NMSE, and Pearson correlation at p < 1e-8, and over ISTA and SBTV on all fidelity metrics. By bypassing the H-matrix at inference, SAN reduces reconstruction time by at least an order of magnitude, supporting real-time PAT reconstruction.
Chinese Translation
光声断层成像（PAT）结合了生物组织的光吸收对比度与超声的空间分辨率，但从稀疏视图传感器测量中恢复初始压力分布仍然是一个病态逆问题。迭代压缩感知求解器和展开的深度网络在推理时均依赖于系统矩阵，这使得实时临床重建在计算上代价高昂。本文提出了传感器注意力网络（SAN），一种基于Transformer的架构，将每个传感器的完整时间序列视为一个令牌，并直接将原始测量映射到重建图像，而无需在推理时调用系统矩阵。为了进行训练和基准测试，构建了一个解析的k空间H矩阵，并在匹配几何下与k-Wave伪谱求解器进行了验证，达到了每个传感器的Pearson相关系数均值为0.919 +/- 0.049，k空间加权和高斯时间衰减协同作用下将能量归一化不匹配降低了49%。在488个增强样本上使用加权损失进行训练，并在46个保留样本上与ISTA、分裂Bregman全变差（SBTV）和学习ISTA（LISTA）进行评估，SAN达到了最高的均值结构相似性指数（SSIM）为0.522和峰值信噪比（PSNR）为22.09 dB，以及最低的归一化均方误差（NMSE）为0.233。配对t检验和Wilcoxon符号秩检验确认了SAN在PSNR、NMSE和Pearson相关性上优于LISTA（p < 1e-8），并在所有保真度指标上优于ISTA和SBTV。通过在推理时绕过H矩阵，SAN将重建时间至少减少了一个数量级，支持实时PAT重建。
cs.AI / 81 / 2607.25583
How Small Can You Go? A Controlled Study of LoRA Rank, Target Modules, and Quantization Trade-offs for Text-to-SQL on a 60M-Parameter Model
你能做到多小？在一个6000万参数模型上对LoRA秩、目标模块和量化权衡进行的控制研究
Abstract
Parameter-efficient fine-tuning (PEFT) and low-bit quantization are now standard tools for adapting language models under tight compute budgets, yet their interaction is most often studied on billion-parameter models where the design space is expensive to explore. We ask a complementary question: on a specific, fully reproducible 60M-parameter encoder-decoder model (T5-small) and a single-table text-to-SQL benchmark (WikiSQL), how much task accuracy does each efficiency knob actually cost? We run a controlled, single-variable study over (i) LoRA rank r in {2, 4, 8, 16, 32}, (ii) the set of adapted modules, and (iii) numerical precision. We report task accuracy alongside system-level metrics including trainable parameters, peak training memory, inference latency, and throughput, and frame adaptation as a constrained trade-off rather than an accuracy-only objective. Our results show that LoRA with r=16 recovers within 11.6 percentage points of full fine-tuning accuracy (59.6% vs. 71.2% exact-match) while training fewer than 1% of parameters and consuming 31% less peak GPU memory. Within this setting, rank beyond r=16 yields no measurable accuracy gain. QLoRA with INT8 and NF4 quantization achieves comparable accuracy (52.8% and 53.2%) at dramatically lower memory cost (0.60 GB each), demonstrating a compelling trade-off for memory-constrained deployments. All code, configurations, and logs are released for full reproducibility.
Chinese Translation
参数高效微调（PEFT）和低位量化现在已成为在紧张计算预算下调整语言模型的标准工具，但它们的相互作用通常是在十亿参数模型上进行研究，这使得设计空间的探索成本高昂。我们提出一个互补的问题：在一个特定的、完全可复现的6000万参数编码-解码模型（T5-small）和一个单表的文本到SQL基准（WikiSQL）上，每个效率调节实际上会消耗多少任务准确性？我们对（i）LoRA秩 r ∈ {2, 4, 8, 16, 32}，（ii）适应模块的集合，以及（iii）数值精度进行了控制的单变量研究。我们报告任务准确性以及系统级指标，包括可训练参数、峰值训练内存、推理延迟和吞吐量，并将适应框架视为一种受限的权衡，而非仅仅是准确性目标。我们的结果表明，r=16的LoRA在训练不到1%的参数并消耗31%更少的峰值GPU内存的情况下，能够恢复到距离完全微调准确性仅有11.6个百分点的水平（59.6%对71.2%的精确匹配）。在这个设置下，秩超过r=16没有可测量的准确性提升。使用INT8和NF4量化的QLoRA在显著更低的内存成本（每个0.60 GB）下实现了相当的准确性（52.8%和53.2%），展示了在内存受限部署中的一个引人注目的权衡。所有代码、配置和日志均已发布，以确保完全可复现性。
cs.AI / 82 / 2607.25597
A Density-Matrix Framework for Electronic-Structure Analysis of Functional-Group and Salt Effects in Lithium-Metal Electrolytes
用于锂金属电解质中功能团和盐效应电子结构分析的密度矩阵框架
Abstract
The reactivity of lithium-metal electrolytes arises from the interplay of molecular functional groups, Li$^+$ solvation, and salt-anion participation. This interplay operates through the redistribution of electron density across donor, anion, and cation centers, which is most directly read out from the electronic structure resolved in space. Quantum-chemical calculations deliver such readouts faithfully, yet become computationally demanding across this multidimensional design space, and machine-learning electronic-structure models seldom cover chemically diverse solvation shells or electrolyte-relevant readouts. Here, we present a density-matrix-centered AI platform (EMolStudio) for electronic-structure prediction and analysis. Its workflow integrates molecular functionalization, explicit Li$^+$ first-shell assembly, density-matrix prediction with idempotency projection, and readouts of frontier orbitals, electrostatic potential, Li$^+$-donor bond order, and electron localization. We apply EMolStudio to 163,655 functionalized molecules and 22,500 explicit Li$^+$ first-shell clusters across four lithium salts. We find that 1) at the molecular scale, functionalization distinguishes CO$_2$Me, CN, F/CF$_3$, and sulfonyl groups by chemically distinct changes in frontier levels, electrostatic potential, and Li$^+$-donor contact, consistent with $\pi^*$-acceptor, inductive, and polarization contributions, with sublinear accumulation at higher degrees of functionalization; 2) in explicit solvation shells, anion identity reshapes frontier-orbital localization: LiTDI anchors the HOMO on the anion across the entire library, whereas LiDFOB pairs an anion-hosted HOMO with strongly functional-group-dependent LUMO hosting. EMolStudio thereby translates functional-group and salt choices into electronic-structure hypotheses relevant to lithium-bond formation, desolvation, and interphase reactions.
Chinese Translation
锂金属电解质的反应性源于分子功能团、Li$^+$ 溶剂化和盐阴离子参与之间的相互作用。这种相互作用通过电子密度在供体、阴离子和阳离子中心之间的重新分布进行，最直接的体现来自于空间中解析的电子结构。量子化学计算能够忠实地提供这种读数，但在这个多维设计空间中计算需求变得很高，而机器学习电子结构模型很少涵盖化学多样的溶剂壳或电解质相关的读数。在此，我们提出了一个以密度矩阵为中心的人工智能平台（EMolStudio）用于电子结构预测和分析。其工作流程整合了分子功能化、显式的 Li$^+$ 第一壳层组装、具有幂等投影的密度矩阵预测，以及前沿轨道、电势、Li$^+$-供体键级和电子局域化的读数。我们将 EMolStudio 应用于 163,655 个功能化分子和 22,500 个显式的 Li$^+$ 第一壳层簇，涵盖四种锂盐。我们发现：1）在分子尺度上，功能化通过前沿能级、电势和 Li$^+$-供体接触的化学特征变化区分了 CO$_2$Me、CN、F/CF$_3$ 和磺酰基等功能团，这与 $ ext{π}^*$-受体、诱导和极化贡献一致，并在更高的功能化程度下呈现亚线性累积；2）在显式溶剂壳中，阴离子身份重塑了前沿轨道的局域化：LiTDI 在整个库中将 HOMO 锚定在阴离子上，而 LiDFOB 则将一个阴离子主导的 HOMO 与强烈依赖功能团的 LUMO 配对。因此，EMolStudio 将功能团和盐的选择转化为与锂键形成、去溶剂化和相间反应相关的电子结构假设。
cs.AI / 83 / 2607.25605
Computational Extraction of Legal Causes via al-Sabr wa al-Taqsim: A Set-Theoretic Formalization for Closed Fiqh Chapters
通过 al-Sabr wa al-Taqsim 的计算法提取法律原因：封闭法学章节的集合论形式化
Abstract
This paper presents a set-theoretic formalization of the classical usuli method of al-Sabr wa al-Taqsim (Examination and Division) for extracting legal causes ('ilal) within closed chapters of jurisprudence. A computational algorithm is introduced that extracts minimal operational rules from a truth table of juristic verdicts. The principal result is that, given a complete truth table for a closed chapter, the algorithm computes the minimal structural generators of the ruling and eliminates all logically redundant attributes. The resulting structures constitute admissible candidate causes for subsequent juristic evaluation. The framework is conditional upon the availability of a finite school-relative concept vocabulary and a complete ruling table for the chapter under investigation.
Chinese Translation
本文提出了一种经典法理学方法 al-Sabr wa al-Taqsim（检验与划分）的集合论形式化，用于提取封闭法学章节中的法律原因（'ilal）。我们引入了一种计算算法，该算法从法学裁决的真值表中提取最小操作规则。主要结果是，给定一个封闭章节的完整真值表，该算法计算裁决的最小结构生成元，并消除所有逻辑上冗余的属性。所得到的结构构成了后续法学评估的可接受候选原因。该框架依赖于有限的学校相关概念词汇和所研究章节的完整裁决表的可用性。
cs.AI / 84 / 2607.25612
Multi-Sensor Alignment for Weather Simulations
天气模拟中的多传感器对齐
Abstract
Perception tasks for autonomous vehicles need to work satisfactorily in adverse weather conditions. Due to lack of real-world weather datasets, weather simulations are a promising alternative. To ensure simulations closely mirror real-world weather data, it's crucial that they represent the same weather characteristics, including severity and particle positioning, across different sensors. To achieve this, we propose the Reference Dataset Alignment Method (ReDAM) for weather intensity alignment in fog and Unified-weather-edit (inspired by Weather-edit[1]) for particle positioning alignment in rain and snow. We validate both alignment methods using statistical and geometrical tests, respectively. We find that 3D detection models for non-aligned versions tend to be overly optimistic as compared to aligned versions. We also show the aligned-multi-sensor simulation's effectiveness for achieving robustness for 3D object detection task by finetuning existing sensor fusion models on it.
Chinese Translation
自主驾驶车辆的感知任务需要在恶劣天气条件下正常工作。由于缺乏真实世界的天气数据集，天气模拟成为一种有前景的替代方案。为了确保模拟能够紧密反映真实世界的天气数据，关键在于它们能够在不同传感器之间表现出相同的天气特征，包括强度和粒子位置。为此，我们提出了参考数据集对齐方法（Reference Dataset Alignment Method, ReDAM）用于雾天的天气强度对齐，以及统一天气编辑（Unified-weather-edit，受Weather-edit[1]启发）用于雨天和雪天的粒子位置对齐。我们分别使用统计和几何测试验证这两种对齐方法。我们发现，非对齐版本的三维检测模型相比于对齐版本往往过于乐观。我们还展示了对齐的多传感器模拟在通过对现有传感器融合模型进行微调以实现三维物体检测任务的鲁棒性方面的有效性。
cs.AI / 85 / 2607.25620
Beyond Epistemia: Epistemic Schizologia and Large Language Models as Techno-Semiotic Machines
超越知识论：作为技术符号机器的大型语言模型与知识性精神分裂症
Abstract
Quattrociocchi and colleagues warn that the fluent outputs of large language models may allow linguistic plausibility to substitute for epistemic evaluation, producing the condition they call *Epistemia*: the experience of possessing knowledge without undertaking the practices through which judgment would ordinarily be warranted. This article accepts that diagnosis but challenges its explanatory framework, which compares an embodied, socially situated human knower with an isolated generative model thereby locating epistemic legitimacy in capacities internal to autonomous agents. Drawing on Carlo Sini's philosophy of practices, writing, signs, and technics, we propose instead to understand a large language model (LLM) as a *techno-semiotic machine* that automates a phase of written semiosis by producing plausible linguistic configurations from the sedimented archive of human writing. From this perspective, *Epistemia* is one consequence of a broader phenomenon that we call *epistemic schizologia*: the socio-technical cleavage between signs as linguistically accomplished expressions and signs as moments within socially embedded circuits of interpretation, evidence, criticism, verification, and responsibility. This cleavage is reinforced by *eikotic closure*, through which a plausible continuation is presented with the finality of an epistemic result, and by algorithmic authority and epistemic self-misrecognition. The relevant unit is therefore not the model alone but the complete practice in which generated inscriptions are prompted, interpreted, verified, contested, used, and made consequential. This reframing preserves the distinction between linguistic production and responsible understanding while grounding a design programme centred on inspectable genealogy, contestability, distributed responsibility, epistemic agency, and the evaluation of hybrid human--AIpractices.
Chinese Translation
Quattrociocchi及其同事警告称，大型语言模型流畅的输出可能使语言的可信性取代知识的评估，产生他们所称的*知识论（Epistemia）*：在没有进行通常应有的判断实践的情况下，体验到拥有知识的状态。本文接受这一诊断，但对其解释框架提出质疑，该框架将具身的、社会嵌入的人类知识者与孤立的生成模型进行比较，从而将知识的合法性定位于自主代理内部的能力。我们借鉴Carlo Sini关于实践、书写、符号和技术的哲学，建议将大型语言模型（LLM）理解为一种*技术符号机器（techno-semiotic machine）*，它通过从人类书写的沉淀档案中生成可信的语言配置，自动化书面符号学的一个阶段。从这个角度看，*知识论（Epistemia）*是我们称之为*知识性精神分裂症（epistemic schizologia）*的更广泛现象的一个结果：符号作为语言完成的表达与符号作为嵌入社会的解释、证据、批评、验证和责任的循环中的时刻之间的社会技术裂缝。这种裂缝通过*图像闭合（eikotic closure）*得以加强，通过这种方式，可信的延续被呈现为知识结果的最终性，以及通过算法权威和知识自我误认。相关的单位因此不仅仅是模型本身，而是生成的铭文被提示、解释、验证、争议、使用并产生后果的完整实践。这种重新框定保留了语言生产与负责任理解之间的区别，同时为以可检验的谱系、争议性、分布式责任、知识代理和混合人类—人工智能实践的评估为中心的设计方案奠定了基础。
cs.AI / 86 / 2607.25624
Quotient Dynamics, Effective Curvature, and Implicit Bias in Positive Quadratic Networks
商动态、有效曲率与正二次网络中的隐性偏差
Abstract
Positive quadratic networks admit the low-rank representation f_U(x)=x^top UU^top x, where Uinmathbb{R}^{dtimes r} is identifiable only up to right orthogonal multiplication, representing a rank-r PSD matrix Q=UU^top. We study how this quotient structure governs training dynamics, curvature, recovery, and interpolation bias. On the full-column-rank stratum, we identify mathbb{R}^{dtimes r}_*/O(r) with the rank-r PSD manifold. For smooth objectives L(U)=ell(UU^top), the Euclidean factor gradient is horizontal. Thus, factor gradient flow projects exactly to quotient Riemannian gradient flow, while finite-step gradient descent induces an exact congruence recursion for the predictor. For quadratic regression, we derive the effective Hessian at interpolators as the empirical measurement Gram form restricted to the tangent space relative to the quotient metric. Under Gaussian rank-one measurements, we compute population curvature, prove uniform deviation bounds for the empirical normal operator, construct a spectral initializer, and establish local exponential convergence for gradient flow and linear convergence for small-step descent. Recovery guarantees are explicit but conservative due to reliance on full-space second-moment control. In underdetermined commuting regimes, factor gradient flow becomes an exact entropy mirror flow in joint spectral coordinates. Strictly positive initializations converge to Bregman projections onto the interpolation set. With isotropic initialization q(0)=varepsilon^2mathbf{1}, predictors approach the minimum-trace solution set as varepsilondownarrow0, resolving nonuniqueness via weighted entropy within the invariant joint spectral algebra. Finite-step descent selects interpolants differing from continuous-time Bregman projections by O(eta). Numerical experiments verify these quotient identities, curvature predictions, recovery behaviors, and selection laws.
Chinese Translation
正二次网络允许低秩表示 f_U(x)=x^top UU^top x，其中 U∈ℝ^{d×r} 仅在右正交乘法下可识别，表示秩为 r 的 PSD 矩阵 Q=UU^top。我们研究这种商结构如何支配训练动态、曲率、恢复和插值偏差。在全列秩层面，我们将 ℝ^{d×r}_*/O(r) 与秩为 r 的 PSD 流形进行识别。对于光滑目标 L(U)=ℓ(UU^top)，欧几里得因子梯度是水平的。因此，因子梯度流准确投影到商黎曼梯度流，而有限步梯度下降则为预测器引入了精确的同构递归。对于二次回归，我们推导出插值点的有效 Hessian，作为相对于商度量的切空间限制的经验测量 Gram 形式。在高斯秩一测量下，我们计算了总体曲率，证明了经验正规算子的统一偏差界，构建了谱初始化器，并建立了梯度流的局部指数收敛性和小步下降的线性收敛性。恢复保证是明确的，但由于依赖于全空间二阶矩控制而显得保守。在欠定的可交换状态下，因子梯度流成为联合谱坐标中的精确熵镜流。严格正的初始化收敛到插值集上的 Bregman 投影。通过各向同性初始化 q(0)=ε^2𝟏，预测器在 ε→0 时接近最小迹解集，通过不变的联合谱代数中的加权熵解决非唯一性。有限步下降选择的插值点与连续时间 Bregman 投影的差异为 O(η)。数值实验验证了这些商身份、曲率预测、恢复行为和选择法则。
cs.AI / 87 / 2607.25626
Joint Text-Audio Alignment for EEG-to-Text Decoding in Chinese Speech Production and Perception
中文语音生产与感知中的脑电图到文本解码的联合文本-音频对齐
Abstract
Decoding speech information directly from scalp electroencephalography (EEG) into text provides a potential non-invasive neural communication pathway for individuals with severe speech and motor impairments. Compared with invasive approaches such as electrocorticography, EEG is safer and more widely deployable, yet substantially more challenging to decode.This challenge is exacerbated for Chinese sentence decoding, which must handle a high-dimensional output space with thousands of characters, severe inter-subject variability, and low signal-to-noise ratios for text alignment.Existing methods commit to a single supervisory axis---either text semantics or audio acoustic features---yet neither can simultaneously satisfy the demands of sentence-level discriminability and fine-grained temporal resolution required for large-vocabulary Chinese decoding. We introduce EEGAlign, a novel parameter-efficient framework that jointly aligns EEG with two axes---text alignment with BGE-M3 text embeddings and audio alignment with wav2vec~2.0 speech features via contrastive learning followed by CTC character-sequence decoding. On ChineseEEG-2 data, EEGAlign yields state-of-the-art closed-set sentence classification performance, reaching up to 82.37% Top-1 accuracy on Reading Aloud EEG and 41.43% on Passive Listening EEG out of 101 candidates. Ablation studies show that the two alignment axes are highly complementary: combining them yields consistently better performance than either alone. To the best of our knowledge, this is the first study on decoding large-vocabulary Chinese sentences from non-invasive EEG during overt speech production, and achieving strong classification performance with relatively large closed-set candidate-sentence setting.
Chinese Translation
直接从头皮脑电图（EEG）解码语音信息为文本，为严重言语和运动障碍的个体提供了一种潜在的非侵入性神经通信途径。与电皮层图（electrocorticography）等侵入性方法相比，EEG更安全且更易于部署，但解码难度显著更大。对于中文句子的解码，这一挑战更加严峻，因为它必须处理具有数千个字符的高维输出空间、严重的个体间变异性以及文本对齐的低信噪比。现有方法专注于单一的监督轴——要么是文本语义，要么是音频声学特征——但两者都无法同时满足大词汇量中文解码所需的句子级可区分性和细粒度时间分辨率的要求。我们提出了EEGAlign，一个新颖的参数高效框架，联合对齐EEG与两个轴——通过对比学习将EEG与BGE-M3文本嵌入进行文本对齐，并通过CTC字符序列解码将EEG与wav2vec~2.0语音特征进行音频对齐。在ChineseEEG-2数据集上，EEGAlign实现了最先进的闭集句子分类性能，在朗读EEG上达到82.37%的Top-1准确率，在被动听觉EEG上达到41.43%，共涉及101个候选句子。消融研究表明，这两个对齐轴是高度互补的：将它们结合起来的性能始终优于单独使用任一轴。根据我们所知，这是首个在明显言语生产过程中从非侵入性EEG解码大词汇量中文句子的研究，并在相对较大的闭集候选句子设置中实现了强大的分类性能。
cs.AI / 88 / 2607.25634
AIriskEval-edu Demo: Auditing of Pedagogical Risks in Educational Explanations
AIriskEval-edu 演示：教育解释中的教学风险审计
Abstract
We present AIriskEval-edu Demo, a platform that audits the pedagogical quality of instructional explanations and provides explainable audit results. The platform evaluates an explanation against a rubric covering five dimensions of pedagogical risk: factual accuracy, depth and completeness, focus and relevance, student-level appropriateness, and ideological bias. For each dimension, it returns a binary decision and a confidence score. Detected risks also include a natural-language rationale and, except for Depth and Completeness, a localized evidence span. The platform integrates GPT-5.5 through an external API and a self-hosted Llama 3.1 8B evaluator that runs on consumer-grade GPUs. The local evaluator is fine-tuned on AIriskEval-edu, a dataset of K-12 instructional explanations with risk and explainability annotations. The platform operates in two modes: in AI mode, both evaluators assess stored explanations generated under six simulated teacher profiles, each representing a distinct pedagogical behavior and potential risk; in human mode, the local evaluator audits user-written explanations in real time. The local evaluator outperforms GPT-5.5 on most reported metrics, offering educational institutions a practical way to keep audited content within their own infrastructure.
Chinese Translation
我们介绍了 AIriskEval-edu 演示，一个审计教学解释教学质量的平台，并提供可解释的审计结果。该平台根据涵盖五个教学风险维度的评分标准评估解释：事实准确性、深度与完整性、焦点与相关性、学生适宜性以及意识形态偏见。对于每个维度，它返回一个二元决策和一个置信度分数。检测到的风险还包括自然语言的理由，除了深度与完整性外，还提供本地化的证据范围。该平台通过外部 API 集成了 GPT-5.5，并使用在消费级 GPU 上运行的自托管 Llama 3.1 8B 评估器。该本地评估器经过微调，基于 AIriskEval-edu 数据集，该数据集包含 K-12 教学解释及其风险和可解释性注释。该平台有两种模式：在 AI 模式下，两个评估器评估在六个模拟教师档案下生成的存储解释，每个档案代表一种独特的教学行为和潜在风险；在人工模式下，本地评估器实时审计用户撰写的解释。本地评估器在大多数报告的指标上优于 GPT-5.5，为教育机构提供了一种在其自身基础设施内保持审计内容的实用方式。
cs.AI / 89 / 2607.25655
Engine-Equal, Human-Unequal: A Reproducible Outcome Skew in Engine-Assessed Equal Chess Positions
引擎相等，人类不平等：引擎评估的平等国际象棋局面中的可重复结果偏差
Abstract
Among chess opening positions that a strong engine judges essentially equal (Stockfish 18 evaluation within 10 centipawns of zero, depth-stable) and that humans actually reach on Lichess (October 2025; 1,661 positions, 16.1M occurrences), human results are not balanced. Positions carry outcome skews, each the gap between its games' actual results and what the players' ratings predict, whose directions are stable properties of the naturally-reached position: some positions favour White, others Black. These skews reproduce across three re-partitions -- disjoint player-account sets (primary), time, and disjoint rating bands -- and on an out-of-sample month eight months later. On the primary split, each position's skew is measured once in each account group, and the replication slope asks how well one measurement predicts the other after removing rating and opening-family effects: one means undiminished carry-over; zero, no linear relation. We find 0.69 (family-clustered 95% CI [0.65, 0.74]), rising to 0.94 on the most-popular, best-measured positions. The slope's value depends on the position mix. Existence is the invariant claim: it survives every tighter evaluation band, search depth, calibration, and popularity cutoff we test, and replicates within blitz and rapid separately. The typical skew is small (median $|\delta| \approx 0.018$, about two percentage points of White score), yet it reproduces, position by position, across disjoint accounts. At these positions the disfavoured side also thinks longer. Even where the evaluation is most confident, it is not a sufficient statistic for human outcomes. The result is observational, and the causal question is left to a pre-registered randomised companion study.
Chinese Translation
在强大的引擎（Stockfish 18 评估在零点的十个分数单位内，深度稳定）判断为基本相等的国际象棋开局局面中，以及人类在 Lichess 上实际达到的局面（2025年10月；1,661 个局面，16.1M 次出现），人类的结果并不平衡。这些局面存在结果偏差，每个偏差是其比赛实际结果与玩家评级预测之间的差距，其方向是自然达到的局面的稳定特性：一些局面偏向白方，另一些则偏向黑方。这些偏差在三个重新划分中重复出现——不重叠的玩家账户集（主要）、时间和不重叠的评级区间——并且在八个月后的样本外月份中也得到了验证。在主要划分中，每个局面的偏差在每个账户组中测量一次，复制斜率则询问在去除评级和开局家族效应后，一个测量如何预测另一个测量：一表示无减弱的延续；零表示没有线性关系。我们发现偏差为 0.69（家族聚类的 95% 置信区间 [0.65, 0.74]），在最受欢迎、测量最好的局面上上升至 0.94。斜率的值依赖于局面组合。存在性是一个不变的主张：它在我们测试的每个更严格的评估区间、搜索深度、校准和流行度截止点中都得以存活，并且在快速棋和闪电棋中分别复制。典型的偏差较小（中位数 $|eta| ext{约} 0.018$，约为白方得分的两个百分点），但它在不重叠的账户中逐局面地重复出现。在这些局面中，处于劣势的一方也会思考更长时间。即使在评估最为自信的情况下，它也不是人类结果的充分统计量。该结果是观察性的，因果问题留待一个预注册的随机化伴随研究来探讨。
cs.AI / 90 / 2607.25656
OrchBench: Evaluating Multi-Agent Orchestration Plans in Isolation via Deterministic Simulation
OrchBench：通过确定性模拟孤立评估多智能体编排计划
Abstract
Complex tasks often decompose into parallelizable yet interdependent subtasks, making orchestration critical to the performance of multi-agent systems (MAS). Existing evaluations typically rely on end-to-end execution, which conflates orchestration-plan quality with worker capabilities, tool reliability, and environmental noise. Moreover, the time and token costs of real execution grow rapidly with workflow scale, making systematic evaluation expensive. We present OrchBench, a simulation-based benchmark for evaluating multi-agent orchestration plans in isolation. Starting from real-world tasks, OrchBench constructs directed acyclic graphs (DAGs) that encode task dependencies, with controlled sizes and degrees of parallelism. Given a DAG, a per-agent context limit, and an agent budget, the evaluated planner assigns subtasks to agents and specifies cross-agent information transfers and their retention ratios. A deterministic simulator evaluates the resulting plan without invoking worker agents and returns interpretable measures of result quality, makespan, and token cost. The simulated scores produced by OrchBench correlate strongly with quality scores from Claude Code executions, achieving a Pearson correlation of \(r=0.816\), while requiring only \(1.3\%\) of the tokens and \(10.3\%\) of the wall-clock time. Across diverse planners and workflow scales, we find that preserving task-critical information is more important than simply increasing the number of agents, and the benefits of parallelism diminish as coordination failures accumulate. These results establish OrchBench as an efficient and interpretable benchmark for comparing and diagnosing multi-agent orchestration plans.
Chinese Translation
复杂任务通常会分解为可并行化但相互依赖的子任务，这使得编排对于多智能体系统（MAS）的性能至关重要。现有评估通常依赖于端到端执行，这将编排计划的质量与工作者能力、工具可靠性和环境噪声混为一谈。此外，实际执行的时间和令牌成本随着工作流规模的增长而迅速增加，使得系统评估变得昂贵。我们提出了OrchBench，这是一个基于模拟的基准，用于孤立评估多智能体编排计划。从现实世界任务出发，OrchBench构建了有向无环图（DAG），以编码任务依赖关系，并控制其大小和并行度。在给定DAG、每个智能体的上下文限制和智能体预算的情况下，被评估的规划者将子任务分配给智能体，并指定跨智能体的信息传输及其保留比例。一个确定性模拟器在不调用工作者智能体的情况下评估生成的计划，并返回可解释的结果质量、完成时间和令牌成本的度量。OrchBench生成的模拟得分与Claude Code执行的质量得分之间具有很强的相关性，皮尔逊相关系数为$r=0.816$，同时仅需$1.3\%$的令牌和$10.3\\%$的实际时间。在不同的规划者和工作流规模中，我们发现保持任务关键的信息比单纯增加智能体数量更为重要，并且随着协调失败的累积，平行化的益处会减小。这些结果确立了OrchBench作为一个高效且可解释的基准，用于比较和诊断多智能体编排计划。
cs.AI / 91 / 2607.25659
CoRT: Counterfactual Replay for Token-Level Rubric-Guided Policy Optimization
CoRT：用于基于评分标准的标记级策略优化的反事实重放
Abstract
Rubric-based reinforcement learning enriches language model training by evaluating model outputs against explicit criteria. Yet in GRPO-style pipelines, these structured judgments are reduced to a scalar response-level reward and converted into a response-level advantage, which is broadcast uniformly to all generated tokens. This leaves no explicit mechanism for allocating credit within a response, even when different criteria are grounded in different spans, formatting decisions, or semantic choices. We propose CoRT, a token-level credit weighting method for rubric-conditioned GRPO. Instead of training an auxiliary token scoring model, CoRT uses counterfactual replay to rescore the same sampled response under the original rubric-conditioned prompt and a matched criteria-free prompt. The resulting tokenwise log-likelihood contrasts serve as a proxy for dependence on the rubric context. CoRT maps these contrasts to bounded, response-normalized weights and uses them to redistribute the signed GRPO advantage across tokens, without introducing an auxiliary scorer or changing the response-level reward. Experiments across instruction-tuned models and reward granularities show that CoRT improves over matched response-level GRPO in the vast majority of comparisons, with an average gain of 4.4 percentage points. The method remains competitive with learned token-level credit baselines while avoiding a separate relevance-learning stage. These results suggest that policy-internal counterfactual likelihood contrasts provide an effective training signal for within-response credit allocation while retaining the simplicity and stability of GRPO.
Chinese Translation
基于评分标准的强化学习通过将模型输出与明确标准进行比较，丰富了语言模型的训练。然而，在GRPO（生成响应级别优化）风格的管道中，这些结构化的判断被简化为标量响应级奖励，并转换为响应级优势，这一优势均匀地广播到所有生成的标记。这导致在响应内部没有明确的信用分配机制，即使不同的标准基于不同的跨度、格式决策或语义选择。我们提出了CoRT，一种用于基于评分标准的GRPO的标记级信用加权方法。CoRT并不训练一个辅助的标记评分模型，而是利用反事实重放在原始评分标准条件提示和匹配的无标准提示下重新评分相同的采样响应。由此产生的逐标记对数似然对比作为对评分标准上下文依赖性的代理。CoRT将这些对比映射到有界的、响应归一化的权重，并利用这些权重在标记之间重新分配签名的GRPO优势，而无需引入辅助评分器或改变响应级奖励。在针对指令调优模型和奖励粒度的实验中，CoRT在绝大多数比较中优于匹配的响应级GRPO，平均提升4.4个百分点。该方法在避免单独的相关性学习阶段的同时，仍然与学习的标记级信用基线保持竞争力。这些结果表明，策略内部的反事实似然对比为响应内部的信用分配提供了有效的训练信号，同时保持了GRPO的简单性和稳定性。
cs.AI / 92 / 2607.25663
Localized Adaptation Reveals Distinct Learning Signatures in Transformers
局部适应揭示变换器中的不同学习特征
Abstract
Transformer adaptation is typically distributed across model depth, even when the intended change is narrow. We investigate how adaptation site shapes what a model learns, how well that learning generalizes, and how selectively it is applied. We introduce a controlled benchmark spanning five objectives (lexical binding, factual association, behavioral policy learning, causal mapping, and procedural reasoning) and define each objective's "adaptation geometry" as its profile of acquisition, transfer, and boundedness under full-stack and early-, middle-, or late-layer LoRA. The objectives exhibit distinct geometries. Lexical binding favors early-layer adaptation for acquisition and boundedness but requires broader updates for transfer; factual association favors later layers among localized adapters; behavioral learning separates late-layer action acquisition from middle-layer policy gating; and causal and procedural transfer benefit most from middle- or full-stack adaptation. These patterns largely persist under parameter-matched controls, and most corresponding directional contrasts replicate across five model families. These findings establish adaptation site as a key design variable for controlling what models learn, generalize, and leave unchanged.
Chinese Translation
变换器的适应通常分布在模型深度的各个层次，即使所需的变化是狭窄的。我们研究适应位置如何影响模型的学习内容、学习的泛化能力以及适应的选择性应用。我们引入了一个涵盖五个目标的受控基准（词汇绑定、事实关联、行为策略学习、因果映射和程序推理），并定义每个目标的“适应几何”作为其在全堆栈和早期、中期或晚期层次的 LoRA 下的获取、转移和边界特征。各个目标展现出不同的几何特征。词汇绑定倾向于在早期层次进行适应以实现获取和边界性，但需要更广泛的更新以实现转移；事实关联在局部适配器中偏好晚期层次；行为学习将晚期层次的行动获取与中期层次的策略门控分开；因果和程序转移则最受益于中期或全堆栈的适应。这些模式在参数匹配的控制下大体保持不变，并且大多数相应的方向对比在五个模型家族中得以重复。这些发现确立了适应位置作为控制模型学习内容、泛化能力和保持不变的关键设计变量。
cs.AI / 93 / 2607.25669
OmniDelta: Skill-Driven Budget Allocation for Token Compression in OmniLLMs
OmniDelta：基于技能驱动的OmniLLMs中令牌压缩的预算分配
Abstract
Emerging Omni-modal Large Language Models (OmniLLMs) enable unified understanding of text, audio, and video, but their long audio-video token sequences introduce substantial memory and inference costs. Existing compression methods mainly focus on selecting important tokens under fixed budgets, leaving the preceding budget-allocation problem underexplored. We show that direct query-to-audio/video similarity is unreliable for inter-modal budget allocation, and that uniform intra-modal budgets can miss key evidence while retaining redundant content. To address these limitations, we propose OmniDelta, a training-free, skill-driven framework that couples intent-aware inter-modal allocation with content-aware intra-modal allocation. OmniDelta first constructs audio and video skill pools to shift the fixed retained-token budget according to query demand, then reallocates modality budgets over audio segments and video frames using local complexity and temporal redundancy. The resulting local budgets can be combined with existing pruning strategies, preserving the total retained-token ratio while changing where the budget is spent. Experiments on four audio-video benchmarks with two Qwen2.5-Omni models show that OmniDelta establishes a new accuracy-efficiency Pareto frontier across pruning ratios. At 25% token retention on Qwen2.5-Omni-7B, OmniDelta reduces GPU memory by 22.0% and achieves a 1.64x end-to-end speedup over full-token inference.
Chinese Translation
新兴的全模态大型语言模型（OmniLLMs）能够统一理解文本、音频和视频，但其长音频-视频令牌序列带来了显著的内存和推理成本。现有的压缩方法主要集中在固定预算下选择重要令牌，导致预算分配问题未得到充分探索。我们表明，直接的查询与音频/视频相似性对于跨模态预算分配是不可靠的，而均匀的模态内预算可能会遗漏关键证据，同时保留冗余内容。为了解决这些局限性，我们提出了OmniDelta，这是一种无训练、基于技能驱动的框架，结合了意图感知的跨模态分配和内容感知的模态内分配。OmniDelta首先构建音频和视频技能池，根据查询需求调整固定的保留令牌预算，然后利用局部复杂性和时间冗余重新分配音频段和视频帧的模态预算。生成的局部预算可以与现有的剪枝策略结合，保持总的保留令牌比例，同时改变预算的使用位置。在四个音频-视频基准上的实验结果表明，OmniDelta在剪枝比率上建立了新的准确性-效率帕累托前沿。在Qwen2.5-Omni-7B模型上，以25%的令牌保留率，OmniDelta将GPU内存减少了22.0%，并在全令牌推理上实现了1.64倍的端到端加速。
cs.AI / 94 / 2607.25675
DecoEvo: Score-Decoupled Co-Evolution of Solver and Rubric-Generator Skills in Text Space
DecoEvo：文本空间中求解器与评分生成器技能的分离协同进化
Abstract
Text-space optimization adapts large language models (LLMs) by editing external natural-language artifacts rather than model weights, so the optimized artifacts remain inspectable and the model can be treated as a black box. However, most existing text-space methods keep evaluation fixed. On open-ended tasks, this can become a bottleneck: once the solver improves on the criteria a rubric measures, omitted dimensions remain invisible to the optimization signal. Simply evolving the rubric is also unreliable when updates are selected by the current solver's score, because apparent progress can come from making the rubric easier to satisfy. We introduce DecoEvo (Decoupled Co-Evolution), which co-evolves a solver skill and a rubric-generator skill under decoupled objectives without using gold rubrics during optimization. The solver skill is updated using criterion-level feedback, while the rubric-generator skill is revised through complementary audits of requirement coverage and response discrimination that are independent of aggregate solver score. This separation focuses generator updates on newly exposed solver weaknesses, reducing repeated emphasis on criteria the solver already satisfies. Under each benchmark's official evaluation, DecoEvo outperforms all compared methods across five benchmarks and three LLM backbones, yielding 2.8--5.0\% relative gains over SkillOpt in the five-benchmark average.
Chinese Translation
文本空间优化通过编辑外部自然语言文档而非模型权重来调整大型语言模型（LLMs），因此优化后的文档仍然可供检查，模型可以被视为黑箱。然而，大多数现有的文本空间方法保持评估固定。在开放式任务中，这可能成为瓶颈：一旦求解器在评分标准上有所改进，未考虑的维度将对优化信号保持不可见。仅仅进化评分标准也不可靠，因为更新是由当前求解器的得分选择的，因为表面上的进展可能来自于使评分标准更容易满足。我们引入了DecoEvo（分离协同进化），在不使用金标准评分的情况下，在分离目标下共同进化求解器技能和评分生成器技能。求解器技能通过标准级反馈进行更新，而评分生成器技能则通过独立于总求解器得分的要求覆盖和响应区分的互补审计进行修订。这种分离使生成器的更新集中在新暴露的求解器弱点上，减少了对求解器已经满足的标准的重复强调。在每个基准的官方评估中，DecoEvo在五个基准和三个LLM骨干上超越了所有比较方法，在五个基准的平均值上相较于SkillOpt获得了2.8%至5.0%的相对提升。
cs.AI / 95 / 2607.25681
Cognivia: A Cognitive Behavioral Therapy Copilot for Evidence-Based Mental Healthcare
Cognivia：一种用于基于证据的心理健康护理的认知行为疗法助手
Abstract
Cognitive distortion amplifies negative emotions and contributes to mental health disorders. Cognitive Behavioral Therapy (CBT) is an effective way to address cognitive distortions, but its large-scale application is limited by the shortage of professional therapists. Although large language models (LLMs) have recently been explored for mental health applications, existing methods still suffer from limited domain specificity, overly flattering responses, and the absence of well-defined annotations for cognitive distortions. This paper proposes Cognivia, an evidence-based artificial intelligence therapist that integrates automatic cognitive distortion identification and rational response generation. Our framework is built on authoritative CBT texts widely regarded as core paradigms and standard references. It is further augmented with mental health question-answer (Q and A) data, and employs multi-stage prompting and structured generation strategies under the supervision of behavioral science experts. Then we fine-tune a lightweight LLM on this augmented CBT dataset to obtain Cognivia. In addition, we propose the first hierarchical quality evaluation framework for assessing LLM-generated rational responses, developed through collaboration between AI researchers and behavioral science experts. Cognivia is evaluated using lexical metrics, LLM-based Judges with two complementary criteria, and human evaluation by 10 behavioral science experts. It consistently outperforms the baseline methods in cognitive distortion recognition and rational response generation, demonstrating its effectiveness. Our code is available at https://github.com/SNOWTEAM2023/Cognivia.
Chinese Translation
认知扭曲会加剧负面情绪并导致心理健康障碍。认知行为疗法（CBT）是解决认知扭曲的有效方法，但其大规模应用受到专业治疗师短缺的限制。尽管最近已经探索了大型语言模型（LLMs）在心理健康应用中的潜力，但现有方法仍然存在领域特异性有限、过于迎合的回应以及缺乏对认知扭曲的明确定义注释等问题。本文提出了Cognivia，一种基于证据的人工智能治疗师，集成了自动认知扭曲识别和理性回应生成。我们的框架建立在被广泛认为是核心范式和标准参考的权威CBT文本基础上，并通过心理健康问答（Q and A）数据进行增强，采用多阶段提示和结构化生成策略，在行为科学专家的监督下进行。然后，我们在这个增强的CBT数据集上对一个轻量级LLM进行微调，以获得Cognivia。此外，我们提出了第一个分层质量评估框架，用于评估LLM生成的理性回应，该框架是在AI研究人员和行为科学专家的合作下开发的。Cognivia通过词汇度量、基于LLM的评审者使用两种互补标准以及10名行为科学专家的人类评估进行评估。在认知扭曲识别和理性回应生成方面，它始终优于基线方法，证明了其有效性。我们的代码可在https://github.com/SNOWTEAM2023/Cognivia获取。
cs.AI / 96 / 2607.25726
Nudging Sustainable Choices through LLM-Generated Recommendation Explanations
通过大型语言模型生成的推荐解释促进可持续选择
Abstract
Recommender systems mediate everyday consumption, offering a promising channel for encouraging sustainable choices. Prior research shows that explanations influence users' perceptions of recommendations and can support more informed decisions. We argue that explanations can also serve as behavioral nudges by foregrounding sustainability information at the moment of choice. This study investigates how different behavioral framings of sustainability information in recommendation explanations affect user choices and perceptions. Using generative AI, we generate sustainability-aware explanations by drawing on nudge theory and validate them through human evaluation and LLM-as-a-judge audits. Building on this foundation, we conduct two randomized studies ($N = 529$) in a low involvement domain (instant coffee) and a high involvement domain (hotel bookings), in which participants choose among preference matched recommendations accompanied by these explanations. Our results show that, across both domains, merely disclosing sustainability information in explanations does not change choices, whereas framing that information or invoking a descriptive social norm significantly increases sustainable selections and eases decision-making. Notably, perception and behavior diverge, as plain disclosure improves explanation evaluations without translating into more sustainable selection behavior. Our work demonstrates how LLMs can generate theory-grounded explanations at scale, pointing toward practical explanation-based interventions for social good. We conclude by discussing implications for adaptive explanation design with generative AI.
Chinese Translation
推荐系统在日常消费中发挥着中介作用，为鼓励可持续选择提供了有前景的渠道。先前的研究表明，解释会影响用户对推荐的感知，并能支持更为知情的决策。我们认为，解释还可以作为行为推动，通过在选择时突出可持续性信息来发挥作用。本研究探讨了推荐解释中可持续性信息的不同行为框架如何影响用户的选择和感知。我们利用生成性人工智能，基于推动理论生成关注可持续性的解释，并通过人类评估和大型语言模型作为评审的审计进行验证。在此基础上，我们在一个低参与度领域（速溶咖啡）和一个高参与度领域（酒店预订）中进行两项随机研究（$N = 529$），参与者在这些解释的陪伴下选择偏好匹配的推荐。我们的结果显示，在这两个领域中，仅仅在解释中披露可持续性信息并未改变选择，而对该信息的框架或引用描述性社会规范显著增加了可持续选择并简化了决策过程。值得注意的是，感知和行为存在差异，简单的披露改善了对解释的评价，但并未转化为更可持续的选择行为。我们的研究展示了大型语言模型如何大规模生成基于理论的解释，指向了基于解释的社会公益干预的实际应用。最后，我们讨论了生成性人工智能在自适应解释设计中的影响。
cs.AI / 97 / 2607.25748
Loss Invariance Determines What Concept Layers Encode: Volume Grounding in Echocardiography
损失不变性决定概念层的编码内容：超声心动图中的体积基础
Abstract
Objective: Concept bottleneck models route prediction through interpretable intermediate variables, and their validity is normally judged by how accurately those variables are predicted. We ask whether that judgement is sufficient, using left ventricular volumes as the concepts underlying ejection fraction estimation from echocardiographic video. Methods: A video transformer encoder was trained on a publicly available echocardiography dataset. End-systolic and end-diastolic volumes formed a concept layer from which ejection fraction was computed analytically, with no residual path to the output. We compared training under an ejection fraction objective alone against training with additional supervision of the volumes in millilitres, and evaluated both on 1276 held-out studies. Results: The concept bottleneck did not increase ejection fraction error relative to direct regression, at 6.89 against 7.13 mean absolute error. Without volume supervision, however, the spread of predicted volumes collapsed to 0.1 millilitres against reference spreads of 35.7 and 45.7 millilitres, while correlation was partly preserved. We show that this follows from an invariance property of the objective: ejection fraction is a ratio and is unchanged when both volumes are rescaled, so the loss determines the concept layer only up to scale. Supervision in absolute units reduced volume error from 89.8 to 25.8 millilitres at a cost of 0.4 in ejection fraction error. Conclusion: Concept accuracy alone can conceal a concept layer that carries no physical scale. Significance: Interpretable intermediate variables in clinical models should be validated against the invariance structure of the training objective, not only against prediction accuracy.
Chinese Translation
目的：概念瓶颈模型通过可解释的中间变量进行预测，其有效性通常通过这些变量的预测准确性来判断。我们探讨这种判断是否足够，以左心室体积作为超声心动图视频中射血分数估计的基础概念。方法：在一个公开可用的超声心动图数据集上训练了一个视频变换器编码器。末收缩期和末舒张期体积形成了一个概念层，从中可以通过解析方法计算射血分数，且没有残余路径通向输出。我们比较了仅在射血分数目标下训练与在体积（以毫升为单位）额外监督下训练的效果，并在1276个保留研究上进行了评估。结果：相对于直接回归，概念瓶颈并未增加射血分数的误差，平均绝对误差为6.89对比7.13。然而，在没有体积监督的情况下，预测体积的分布收缩至0.1毫升，而参考分布为35.7和45.7毫升，尽管相关性部分得以保留。我们表明，这源于目标的一个不变性特性：射血分数是一个比率，当两个体积都被重新缩放时，其值不变，因此损失仅在尺度上决定概念层。绝对单位的监督将体积误差从89.8减少至25.8毫升，代价是射血分数误差增加0.4。结论：仅靠概念准确性可能掩盖一个没有物理尺度的概念层。意义：临床模型中的可解释中间变量应根据训练目标的不变性结构进行验证，而不仅仅是根据预测准确性。
cs.AI / 98 / 2607.25816
Speculate While You Reason: Teaching Agents to Predict Their Next Tool Call via Joint Agent-Speculator RL
推测与推理：通过联合智能体-推测器强化学习教导智能体预测其下一个工具调用
Abstract
Large language model agents often spend substantial wall-clock time waiting for tool call results. Tool-call speculation can hide this latency by predicting and pre-executing an agent's next tool call if the prediction matches the agent's eventual tool call, but existing speculators are typically separate draft models or cached traces that are poorly aligned with the deployed agent's own behavior. We identify this speculator-agent gap and show that the target agent itself is a strong next-call speculator. This points to a simpler design: unifying the agent and speculator within the same model. In this paper, we introduce the self-speculating agent, a single model that both solves tasks in agent mode and predicts its next tool call from partial trajectories in speculator mode, fully reusing prefix KV cache. To enable this dual-mode agent without degrading performance, we propose a joint agent-speculator reinforcement learning method, which derives speculation targets from the agent's own rollouts and alternates agent and speculator updates. Across agentic search QA and conversational tool-use agentic tasks, our method improves average next tool-call Hit@1 from 44.1 to 61.2 for Qwen3-4B and from 48.9 to 66.3 for Qwen3.5-4B, while preserving agent task success.
Chinese Translation
大型语言模型智能体通常会花费大量的实际时间等待工具调用结果。工具调用推测可以通过预测并预执行智能体的下一个工具调用来隐藏这种延迟，如果预测与智能体最终的工具调用相匹配。然而，现有的推测器通常是独立的草稿模型或缓存的轨迹，这些与部署的智能体自身行为的对齐程度较差。我们识别出这种推测器-智能体之间的差距，并展示目标智能体本身是一个强大的下一个调用推测器。这指向了一种更简单的设计：将智能体和推测器统一在同一模型中。在本文中，我们介绍了自推测智能体，这是一种单一模型，既可以在智能体模式下解决任务，又可以在推测器模式下从部分轨迹中预测其下一个工具调用，完全重用前缀KV缓存。为了实现这种双模式智能体而不降低性能，我们提出了一种联合智能体-推测器强化学习方法，该方法从智能体自身的回滚中推导推测目标，并交替进行智能体和推测器的更新。在智能体搜索问答和对话工具使用的智能任务中，我们的方法将Qwen3-4B的平均下一个工具调用Hit@1从44.1提高到61.2，将Qwen3.5-4B的Hit@1从48.9提高到66.3，同时保持智能体任务的成功率。
cs.AI / 99 / 2607.25835
Distributed Constraint Optimization via Online Learning and Iterative Pricing with Application to Large-Scale Satellite Scheduling
通过在线学习和迭代定价进行分布式约束优化及其在大规模卫星调度中的应用
Abstract
Distributed constraint optimization problems (DCOPs) provide a popular framework for distributed decision making under limited communication, but many real-world instances are too large to solve monolithically. We address this challenge from two complementary directions. We revisit the connection between DCOPs and potential games, and adapt modern online learning algorithms for equilibrium finding to DCOPs. We show that these algorithms are competitive with representative incomplete DCOP algorithms. We then turn to decomposition frameworks for large-scale DCOPs, motivated by large-scale decentralized satellite scheduling. We propose a new framework that separates a DCOP into two interacting subproblems: a high-level meta-DCOP for task allocation, and independent local optimization problems for scheduling. To couple the two levels, we develop a novel iterative pricing method that updates the meta-level utilities using feedback from the local optimizers. Combining our online learning methods with our iterative pricing framework, we obtain near-optimal performance on real-world decentralized satellite scheduling problem instances, fulfilling over 99% of observation requests compared with 87% for state-of-the-art baselines.
Chinese Translation
分布式约束优化问题（DCOPs）为有限通信下的分布式决策提供了一个流行的框架，但许多实际实例过于庞大，无法单体求解。我们从两个互补的方向来解决这一挑战。我们重新审视DCOPs与潜在博弈之间的联系，并将现代在线学习算法适配于DCOPs的均衡求解。我们证明这些算法在与代表性的非完全DCOP算法的竞争中表现良好。接着，我们转向大规模DCOPs的分解框架，受到大规模去中心化卫星调度的启发。我们提出了一个新的框架，将DCOP分解为两个相互作用的子问题：一个用于任务分配的高层元DCOP，以及用于调度的独立局部优化问题。为了将两个层次结合起来，我们开发了一种新颖的迭代定价方法，通过局部优化器的反馈更新元层效用。将我们的在线学习方法与迭代定价框架结合后，我们在实际的去中心化卫星调度问题实例中获得了近乎最优的性能，满足了超过99%的观测请求，而最先进的基线仅满足87%。
cs.AI / 100 / 2607.25853
HiSkill: Empowering LLM Agents with Hierarchical Skill Graphs
HiSkill：通过层次技能图赋能大型语言模型代理
Abstract
Skills have become an important abstraction for enabling large language model (LLM) agents to reuse past experience in long-horizon interactive tasks. However, existing trajectory-to-skill methods often produce flat collections of high-level textual skills that are stored and retrieved independently, leaving skill relations underutilized and maintaining a gap between high-level skills and executable actions. In this paper, we propose HiSkill, a hierarchical skill graph framework that organizes interaction trajectories into a directed graph with skill nodes, AtomicOp nodes, and typed edges. Specifically, the graph connects reusable high-level skills with executable action templates, while also capturing decomposition, temporal transition, compatibility, support, and recovery relations among them. At inference time, HiSkill retrieves a compact task-relevant subgraph and performs subgraph-guided task execution, where a symbolic task state, an active skill, and the retrieved subgraph guide the LLM agent to switch skills, select AtomicOps, and ground executable actions iteratively. Experiments on three interactive environments show that HiSkill outperforms state-of-the-art baselines while reducing inference token consumption, demonstrating the effectiveness of bridging high-level skills and executable action grounding through a hierarchical skill graph. Our data and code is available at https://github.com/BUPT-GAMMA/HiSkill.
Chinese Translation
技能已成为使大型语言模型（LLM）代理在长时间交互任务中重用过去经验的重要抽象。然而，现有的轨迹到技能方法通常产生扁平的高层次文本技能集合，这些技能被独立存储和检索，导致技能关系未得到充分利用，并在高层次技能与可执行动作之间保持差距。本文提出了HiSkill，一个层次技能图框架，将交互轨迹组织成一个包含技能节点、原子操作（AtomicOp）节点和类型化边的有向图。具体而言，该图将可重用的高层次技能与可执行的动作模板连接起来，同时捕捉它们之间的分解、时间过渡、兼容性、支持和恢复关系。在推理时，HiSkill 检索一个紧凑的任务相关子图，并执行子图引导的任务执行，其中符号任务状态、活跃技能和检索的子图引导LLM代理迭代地切换技能、选择原子操作并确定可执行动作。在三个交互环境中的实验表明，HiSkill 在减少推理令牌消耗的同时超越了最先进的基线，展示了通过层次技能图架构连接高层次技能与可执行动作基础的有效性。我们的数据和代码可在 https://github.com/BUPT-GAMMA/HiSkill 获取。
cs.AI / 101 / 2607.25877
Runtime Uncertainty Monitoring for LLM-Based Multi-Agent Systems Using Bayesian Networks
基于贝叶斯网络的LLM多智能体系统运行时不确定性监测
Abstract
This paper investigates how multi-agent systems (MAS)-based on large language models (LLMs) can support actuarial risk modelling, with a particular focus on uncertainty quantification. Actuarial workflows represent a high-stakes decision-support setting where unreliable outputs may lead to incorrect risk assessment, unfair pricing, and regulatory non-compliance. To address uncertainty introduced by the probabilistic nature of LLMs and dependencies between agents, a multi-agent framework is proposed in which specialised agents perform data preparation, modelling, review, and explanation tasks under a central hub. The main contribution is a novel approach to uncertainty propagation using token-level log-probabilities and a Bayesian Network. Importantly, log probabilities are not treated as direct probabilities of correctness or task success. Instead, length-normalised log-probability summaries are transformed into calibrated task-level confidence estimates before incorporation into the Bayesian Network. Results show that the framework reproduces baseline actuarial performance while providing additional insight into workflow stability and runtime uncertainty propagation.
Chinese Translation
本文探讨了基于大型语言模型（LLM）的多智能体系统（MAS）如何支持精算风险建模，特别关注不确定性量化。精算工作流程代表了一种高风险的决策支持环境，其中不可靠的输出可能导致错误的风险评估、不公平的定价和监管不合规。为了解决LLM的概率特性和智能体之间的依赖关系所引入的不确定性，提出了一种多智能体框架，其中专门的智能体在中央枢纽下执行数据准备、建模、审查和解释任务。主要贡献是提出了一种使用令牌级对数概率和贝叶斯网络的不确定性传播新方法。重要的是，对数概率并不被视为正确性或任务成功的直接概率。相反，经过长度归一化的对数概率摘要在纳入贝叶斯网络之前被转换为经过校准的任务级置信度估计。结果表明，该框架在再现基线精算性能的同时，提供了对工作流程稳定性和运行时不确定性传播的额外洞察。
cs.AI / 102 / 2607.25890
Distributing Security Controls Through Harness Engineering
通过代理工程分配安全控制
Abstract
AI coding agents are being adopted at historic speed, yet security and risk concerns remain the primary barrier to scaling agentic AI across organizations. Existing security controls for coding agents are not systematically distributed to engineering teams, and vendor-native solutions introduce ecosystem dependencies that may not suit every deployment context. This paper investigates whether off-the-shelf security controls can be implemented on commercial AI coding agents and scaled to a distributed user base via a custom agent harness. A phased testing methodology was applied across four agent configurations --- two commercial agents with and without controls, a baseline harness, and a security-hardened harness --- using a 23-test suite derived from the OWASP Top 10 for Agentic Applications. SHarD (Secure Harness Distribution), a distributable harness built on the Pi agent harness, demonstrated that three categories of security controls --- OS sandboxing, skill scanning, and tool restriction --- can be embedded and distributed via a single install command while retaining equivalent efficacy to direct installation on commercial agents. SHarD achieved an adjusted score of 100\%, matching the best securely configured commercial agent, with no regression across any test category. Notable observations include evidence that model non-determinism produces inconsistent security outcomes and that autonomous agent behavior can cross system boundaries in ways that OS sandboxing directly mitigates. Initial characteristics toward a control harness fitness framework are proposed, and a third research question is identified for future investigation.
Chinese Translation
人工智能编码代理的采用速度空前迅猛，但安全和风险问题仍然是将代理人工智能在组织中规模化的主要障碍。现有的编码代理安全控制并未系统性地分配给工程团队，而供应商原生解决方案则引入了生态系统依赖性，可能不适合每种部署环境。本文探讨了现成的安全控制是否可以在商业人工智能编码代理上实施，并通过自定义代理工具进行分发以适应分布式用户基础。我们在四种代理配置中应用了分阶段测试方法——两种商业代理（有控制和无控制）、一个基线工具和一个安全增强工具——使用了基于OWASP Top 10 for Agentic Applications的23项测试套件。SHarD（安全工具分发）是一个基于Pi代理工具构建的可分发工具，证明了三类安全控制——操作系统沙箱、技能扫描和工具限制——可以通过单个安装命令嵌入和分发，同时保持与在商业代理上直接安装的等效效果。SHarD的调整得分为100%，与配置最佳的安全商业代理相匹配，且在任何测试类别中均未出现回归。值得注意的观察包括模型非确定性导致不一致的安全结果，以及自主代理行为可以以操作系统沙箱直接缓解的方式跨越系统边界。本文提出了控制工具适应性框架的初步特征，并确定了一个未来研究的第三个研究问题。
cs.AI / 103 / 2607.25891
Messier: A High-Resolution Corpus for Cross-Benchmark Agent Evaluation
Messier：用于跨基准代理评估的高分辨率语料库
Abstract
Evaluating AI agents in interactive environments is hindered by fragmented tasks, scaffolds, verifiers, and scoring rules. Existing efforts focus on narrow settings, remain limited in scale, or require costly reruns, leaving much of the empirical record incomparable. We introduce Messier, a unified corpus of 957,253 records that span 30 benchmarks, 714 agents, 11,891 tasks, and 74,205 verifiers. Messier consolidates public benchmark scores and supplements them with five-agent runs across six underrepresented professional and scientific domains, including a recent legal benchmark. Each record is standardized by model, scaffold, environment, task, verifier, and aggregation rule, with SOC/NAICS classifications for occupational and industry analysis. Using this corpus, we show frontier progress is uneven across benchmark types, with "function calling" saturated, "programming" improving the fastest, and "enterprise workflows" remaining the most challenging. Furthermore, counterfactual rescoring shows that strict all-pass aggregation in multi-verifier tasks can obscure progress and artificially alter agent rankings. From these standardized records, we derive capability scales that align with Epoch's Evaluation Capability Index rankings at Spearman \r{ho} = 0.81 and can be specialized by domain, occupation, action space, or verifier type. Messier provides a foundational, reusable infrastructure for agent capability scaling, benchmark auditing, and fine-grained analysis of evaluation failures.
Chinese Translation
在互动环境中评估人工智能代理面临着任务碎片化、支架、验证者和评分规则等问题。现有的研究集中于狭窄的设置，规模有限，或需要高成本的重跑，导致大量实证记录无法进行比较。我们引入了Messier，一个统一的语料库，包含957,253条记录，涵盖30个基准、714个代理、11,891个任务和74,205个验证者。Messier整合了公共基准分数，并通过在六个代表性不足的专业和科学领域进行五个代理的运行进行补充，包括最近的法律基准。每条记录都按照模型、支架、环境、任务、验证者和聚合规则进行标准化，并包含职业和行业分析的SOC/NAICS分类。利用该语料库，我们展示了前沿进展在基准类型之间的不均衡，其中“函数调用”已趋于饱和，“编程”进展最快，而“企业工作流程”仍然是最具挑战性的。此外，反事实重新评分表明，在多验证者任务中，严格的全通过聚合可能会掩盖进展并人为改变代理排名。基于这些标准化记录，我们推导出与Epoch的评估能力指数排名一致的能力尺度，Spearman
{ho} = 0.81，并可以根据领域、职业、行动空间或验证者类型进行专业化。Messier为代理能力评估、基准审计和评估失败的细致分析提供了基础的、可重用的基础设施。
cs.AI / 104 / 2607.25904
Interactive Reward Agent: GUI Task Evaluation via Environment-State Verification
交互式奖励代理：通过环境状态验证进行GUI任务评估
Abstract
Graphical user interface task evaluation aims to determine whether a GUI agent has successfully completed a user instruction. Automated GUI task evaluation has received increasing attention because the evaluation results can serve as reward signals for both test-time scaling and post-training. However, reliable GUI task evaluation remains challenging because the judgments often require access to environment states, such as system configurations, file data, and application settings, beyond the screenshots of execution trajectories. In this paper, we propose an interactive reward agent (IRA) based on a propose-then-verify framework to acquire and verify evidence from the post-execution environment. Given a task instruction and a GUI environment after the GUI agent execution, IRA first proposes the task completion conditions and then verifies them by invoking system tools, application tools, and GUI tools. This design combines evidence from both visible interfaces and the environment state in an interactive process. We further introduce GUI-RewardBench, a benchmark of 321 GUI task trajectories spanning 10 Ubuntu desktop application categories. Experiments show that IRA achieves 86.9% accuracy on GUI-RewardBench, outperforming existing evaluator baselines. We further apply IRA to reinforcement learning of GUI agents, achieving a 34.0% OSWorld success rate, which demonstrates that IRA can provide effective reward signals for training GUI agents.
Chinese Translation
图形用户界面（GUI）任务评估旨在确定GUI代理是否成功完成用户指令。自动化的GUI任务评估受到越来越多的关注，因为评估结果可以作为测试时扩展和后期训练的奖励信号。然而，可靠的GUI任务评估仍然具有挑战性，因为判断通常需要访问环境状态，如系统配置、文件数据和应用设置，而不仅仅是执行轨迹的屏幕截图。在本文中，我们提出了一种基于提议-验证框架的交互式奖励代理（IRA），以获取和验证来自执行后环境的证据。在给定任务指令和GUI代理执行后的GUI环境的情况下，IRA首先提出任务完成条件，然后通过调用系统工具、应用工具和GUI工具进行验证。该设计在交互过程中结合了来自可见界面和环境状态的证据。我们进一步介绍了GUI-RewardBench，这是一个涵盖10个Ubuntu桌面应用类别的321个GUI任务轨迹的基准测试。实验表明，IRA在GUI-RewardBench上达到了86.9%的准确率，超越了现有的评估基线。我们还将IRA应用于GUI代理的强化学习，实现了34.0%的OSWorld成功率，这表明IRA能够为训练GUI代理提供有效的奖励信号。
cs.AI / 105 / 2607.25914
Toward Standardized Cross-Vendor Agent Tool Trust Management in Autonomous Networks
朝着自主网络中标准化跨供应商代理工具信任管理的方向
Abstract
Autonomous Network Levels 4-5 require AI agents to invoke tools across vendor boundaries without human oversight, yet existing management standards lack a standardized mechanism for cross-vendor trust visibility. When a tool from Vendor B is compromised, agents from Vendor A continue invoking it -- unaware of the trust degradation -- causing cascading service impact. We present AgentToolMO, a proposed 3GPP NRM information model for agent tool trust management. The model comprises: a formally defined trust state machine with provable graduated enforcement, damped cascade propagation with bounded convergence, cross-vendor trust notifications via existing Management Services (MnS) interfaces, and retroactive impact assessment through NRM dependency graph traversal. Simulation-based evaluation across multi-vendor topologies shows that standardized cross-vendor notifications reduce blast radius from hours-scale undetected propagation to near-real-time containment bounded by MnS notification delivery, with cascade convergence guaranteed in bounded iterations and sub-linear notification scaling across vendor domains. The framework operates within existing 3GPP management infrastructure, leverages existing protocols, and provides a standardization pathway for trustworthy multi-vendor autonomous network management.
Chinese Translation
自主网络的4-5级别要求人工智能代理在没有人工监督的情况下跨供应商边界调用工具，但现有的管理标准缺乏跨供应商信任可见性的标准化机制。当供应商B的工具受到攻击时，供应商A的代理仍然会调用该工具，而未意识到信任的下降，从而导致级联服务影响。我们提出了AgentToolMO，这是一个为代理工具信任管理而设计的3GPP NRM信息模型。该模型包括：一个形式化定义的信任状态机，具有可证明的渐进强制执行，受限收敛的减震级联传播，通过现有管理服务（MnS）接口进行跨供应商信任通知，以及通过NRM依赖图遍历进行的追溯影响评估。在多供应商拓扑下的基于仿真的评估表明，标准化的跨供应商通知将未被检测的传播的影响范围从小时级别减少到接近实时的限制，受限于MnS通知的交付，且在有限的迭代中保证级联收敛，并在供应商域之间实现亚线性的通知扩展。该框架在现有的3GPP管理基础设施中运行，利用现有协议，并为可信的多供应商自主网络管理提供了标准化路径。
cs.AI / 106 / 2607.25915
Penelope: Localized Latent Recurrence for Efficient Structured Reasoning
Penelope：局部潜在递归以实现高效的结构化推理
Abstract
Complex structured reasoning tasks often require additional computation, yet current language models obtain it mainly by increasing parameter scale or by serializing intermediate steps as chain-of-thought (CoT) tokens. The former raises training and deployment costs, while the latter ties reasoning computation to autoregressive output length. We introduce Penelope, an efficient latent-reasoning framework for pretrained decoder-only Transformers that localizes recurrent computation to a selected decoder interval. The lower decoder prefix is evaluated once to construct a problem-conditioned boundary memory, which is then iteratively refined through time-modulated GRU dynamics and recurrent readout states before answer generation. A progressive CoT-to-latent curriculum transfers visible reasoning into this internal recurrent path, allowing additional computation to be allocated in latent space without repeatedly executing the complete decoder or generating a long intermediate trace. Experiments on open-source structured-reasoning benchmarks show that, at validation-selected latent budgets, Penelope attains competitive accuracy relative to established latent-reasoning models while reducing measured inference latency. These results show that latent refinement can be localized to a narrow decoder interval, reducing repeated full-decoder execution without generating a long visible reasoning trace and providing a practical accuracy-efficiency tradeoff for decoder-only Transformer models.
Chinese Translation
复杂的结构化推理任务通常需要额外的计算，然而当前的语言模型主要通过增加参数规模或将中间步骤序列化为思维链（CoT）标记来获得这些计算。前者提高了训练和部署成本，而后者则将推理计算与自回归输出长度绑定在一起。我们提出了Penelope，一个高效的潜在推理框架，适用于预训练的仅解码器Transformer，它将递归计算局部化到选定的解码器区间。较低的解码器前缀被评估一次，以构建一个问题条件的边界记忆，然后通过时间调制的GRU动态和递归读出状态进行迭代精炼，最终生成答案。渐进式的CoT到潜在课程将可见推理转移到这个内部递归路径中，从而使得额外的计算能够在潜在空间中分配，而无需重复执行完整的解码器或生成长的中间轨迹。在开源结构化推理基准上的实验表明，在验证选择的潜在预算下，Penelope相对于已建立的潜在推理模型达到了竞争性的准确性，同时减少了测量的推理延迟。这些结果表明，潜在精炼可以局部化到狭窄的解码器区间，从而减少重复的完整解码器执行，而不生成长的可见推理轨迹，为仅解码器Transformer模型提供了实用的准确性与效率的权衡。
cs.AI / 107 / 2607.25925
dtControl2+$\varepsilon$: Trading Optimality for Explainability in MDPs via Decision Trees
dtControl2+$ ext{ε}$：通过决策树在马尔可夫决策过程中的可解释性与最优性之间的权衡
Abstract
Over the past decade, decision trees have been used to represent controllers (a.k.a. policies) in an explainable way, with dtControl2 as a current state-of-the-art tool. However, for systems that are large or have many corner cases, even such representations tend to be too complex and not human-comprehensible. Unfortunately, reducing the size of the decision tree is not straightforward, as missing just a single crucial case might result in an incorrect controller. We tackle this issue in the setting of Markov decision processes, extending dtControl2 by "$\varepsilon$" functionality: Given an allowed imprecision $\varepsilon \geq 0$, we construct a smaller decision tree, distilling the essence of the controller, while still guaranteeing its $\varepsilon$-optimality. This enables us to provide tunably simpler explanations, omitting a controllable amount of detail. Our tool constructs decision trees that are orders of magnitude smaller than the state of the art.
Chinese Translation
在过去十年中，决策树被用于以可解释的方式表示控制器（即策略），其中 dtControl2 是当前的最先进工具。然而，对于大型系统或具有许多边缘案例的系统，即使这样的表示也往往过于复杂，难以被人类理解。不幸的是，减少决策树的大小并非易事，因为缺失一个关键案例可能导致控制器不正确。我们在马尔可夫决策过程的背景下解决了这个问题，通过引入“$ ext{ε}$”功能扩展 dtControl2：给定一个允许的精度 $ ext{ε}
geq 0$，我们构建一个更小的决策树，提炼控制器的本质，同时仍然保证其 $ ext{ε}$-最优性。这使我们能够提供可调节的更简单的解释，省略可控数量的细节。我们的工具构建的决策树比最先进的技术小几个数量级。
cs.AI / 108 / 2607.25947
A Cost-Effective Multimodal LLM Reasoning Framework for Question Answering over Irregular Clinical Time Series
一种经济高效的多模态大语言模型推理框架用于不规则临床时间序列的问答
Abstract
Question answering (QA) over irregular clinical time series (ICTS) plays a pivotal role in a wide range of healthcare applications. Although recent multimodal time-series large language models (LLMs) have shown considerable promise in general-purpose time-series QA, they remain poorly equipped to model the sparsity, asynchrony, and irregular sampling patterns of clinical observations. To fill this gap, we propose ClinPRISM, a cost-effective multimodal LLM reasoning framework for question answering over ICTS data. First, we devise an irregularity-aware multi-scale encoder to capture sparse clinical evidence at diverse temporal scales. Then, we propose a temporal evidence distiller to integrate representations across these scales and compress them into a small number of LLM-compatible tokens. Moreover, we introduce a progressive alignment strategy that sequentially aligns the irregular trajectories with the LLM's textual embedding space. To facilitate training, we construct 30,000 clinical time series paired with multi-scale descriptions, together with 41,000 instruction-tuning instances spanning 11 tasks. Using a 4-billion-parameter LLM backbone, ClinPRISM achieves state-of-the-art performance on the held-out evaluation benchmark while using only 16 time-series tokens and achieving an average inference latency of 0.15 seconds per question.
Chinese Translation
不规则临床时间序列（ICTS）上的问答（QA）在广泛的医疗应用中发挥着关键作用。尽管近期的多模态时间序列大语言模型（LLMs）在通用时间序列问答中展现出相当大的潜力，但它们在建模临床观察的稀疏性、异步性和不规则采样模式方面仍显不足。为填补这一空白，我们提出了ClinPRISM，一种经济高效的多模态LLM推理框架，用于ICTS数据的问答。首先，我们设计了一种考虑不规则性的多尺度编码器，以捕捉不同时间尺度下的稀疏临床证据。然后，我们提出了一种时间证据提炼器，以整合这些尺度的表示并将其压缩为少量与LLM兼容的标记。此外，我们引入了一种渐进对齐策略，依次将不规则轨迹与LLM的文本嵌入空间对齐。为了便于训练，我们构建了30,000个临床时间序列，配有多尺度描述，以及41,000个涵盖11个任务的指令调优实例。使用一个拥有40亿参数的LLM骨干，ClinPRISM在保留的评估基准上实现了最先进的性能，同时仅使用16个时间序列标记，并且每个问题的平均推理延迟为0.15秒。
cs.AI / 109 / 2607.25956
Large Language Model for Operations Research Formulation Selection in Multi-Warehouse Inventory Allocation
用于多仓库库存分配的运筹学公式选择的大型语言模型
Abstract
Multi-warehouse inventory allocation is typically formulated as a mixed-integer programming (MIP) problem, yet no single formulation consistently matches heterogeneous instance-level regimes induced by demand concentration, inventory imbalance, replenishment scale, service constraints, and forecast volatility. We study this issue as instance-wise operations research (OR) formulation selection, where each allocation instance is assigned to a solver-executable formulation from a candidate OR expert library. We propose a solver-guided large language model (LLM) framework for OR formulation selection, in which each OR expert corresponds to a MIP formulation encoding a distinct allocation priority. To train the selector, the framework first constructs balanced expert-conditioned supervised fine-tuning (SFT) records for schema learning, and then uses MIP solver evaluation on historical instances to convert solver-evaluated allocation-quality gaps into margin-weighted identity preference optimization (IPO) preferences and per-instance expert-score metadata for reward lookup during group relative policy optimization (GRPO) to assign rewards to sampled responses. Experiments on multi-warehouse inventory allocation instances from JD$\mathord{.}$com, one of China's largest e-retailers, demonstrate that GRPO substantially improves expert-selection accuracy relative to the SFT+IPO selector and, more importantly, produces higher realized allocation quality than both the preference-trained selector and the best fixed formulation. With GRPO, Hit Ratio@1 and Hit Ratio@2 increase from 21.45% to 50.42% and from 70.47% to 82.31%. The resulting selector achieves an allocation accuracy gain of 12.57 percentage points over the incumbent baseline, outperforming both the SFT+IPO selector and the best fixed OR expert, and reduces the gap to the ex-post oracle to 4.85 percentage points.
Chinese Translation
多仓库库存分配通常被表述为一个混合整数规划（MIP）问题，但没有单一的公式能够始终匹配由需求集中、库存不平衡、补货规模、服务约束和预测波动引起的异质实例级模式。我们将这一问题视为实例级运筹学（OR）公式选择，其中每个分配实例被分配给来自候选OR专家库的可求解公式。我们提出了一种基于求解器指导的大型语言模型（LLM）框架用于OR公式选择，其中每个OR专家对应于编码特定分配优先级的MIP公式。为了训练选择器，该框架首先构建平衡的专家条件监督微调（SFT）记录以进行模式学习，然后利用历史实例上的MIP求解器评估将求解器评估的分配质量差距转换为边际加权身份偏好优化（IPO）偏好和每个实例的专家评分元数据，以便在群体相对策略优化（GRPO）期间进行奖励查找，从而为采样响应分配奖励。在中国最大的电子零售商之一JD.com的多仓库库存分配实例上的实验表明，GRPO显著提高了相对于SFT+IPO选择器的专家选择准确性，更重要的是，产生的实际分配质量高于偏好训练选择器和最佳固定公式。使用GRPO，Hit Ratio@1和Hit Ratio@2分别从21.45%提高到50.42%和从70.47%提高到82.31%。最终的选择器在分配准确性上比现有基线提高了12.57个百分点，超越了SFT+IPO选择器和最佳固定OR专家，并将与事后oracle的差距缩小至4.85个百分点。
cs.AI / 110 / 2607.26023
CHARM: A Multimodal Graph Foundation Model with Hierarchical Context Modeling for Zero-Shot Transfer
CHARM：一种具有层次上下文建模的多模态图基础模型，用于零样本迁移
Abstract
Graph foundation models (GFMs) have emerged as a promising paradigm for transferring knowledge across graph domains and tasks. Real-world graphs associate nodes with text, images, and other modalities, making multimodal graphs essential for representing complex entities and relations. Moreover, collecting labels and adapting models for every new graph domain is costly and often infeasible, motivating zero-shot transfer. Unfortunately, zero-shot transfer on multimodal graphs remains underexplored. Existing GNN-based graph foundation models typically require downstream adaptation, whereas LLM-based graph methods mainly address unimodal graphs or tasks within a single domain. This setting presents two key challenges. First, models must generalize knowledge from individual modalities while capturing transferable cross-modal relations. Second, without target-domain fine-tuning, node representations remain entangled with domain-specific structures and modality-specific characteristics, obscuring shared concepts in unseen domains. To address these challenges, we propose CHARM, a multimodal graph foundation model with hierarchical context modeling for zero-shot transfer. CHARM replaces isolated raw nodes with hierarchical graph contexts that capture multimodal semantics and cross-modal relations. These contexts map domain-specific node patterns to shared high-level concepts, reducing reliance on target-domain supervision or adaptation. A modality-aware graph context encoder integrates multimodal information with graph structure and converts the resulting representations into graph tokens for a large language model . Experiments show consistent improvements on zero-shot multimodal graph tasks.
Chinese Translation
图基础模型（GFMs）已成为跨图领域和任务转移知识的有前景的范式。现实世界中的图将节点与文本、图像及其他模态关联起来，使得多模态图在表示复杂实体和关系方面至关重要。此外，为每个新图领域收集标签和调整模型的成本高昂且往往不可行，这促使了零样本迁移的需求。不幸的是，在多模态图上进行零样本迁移仍然未得到充分探索。现有的基于图神经网络（GNN）的图基础模型通常需要下游适应，而基于大型语言模型（LLM）的方法主要针对单模态图或单一领域内的任务。这种设置带来了两个关键挑战。首先，模型必须从各个模态中泛化知识，同时捕捉可转移的跨模态关系。其次，在没有目标领域微调的情况下，节点表示仍然与特定领域的结构和特定模态的特征交织在一起，模糊了在未见领域中的共享概念。为了解决这些挑战，我们提出了CHARM，一种具有层次上下文建模的多模态图基础模型，用于零样本迁移。CHARM用层次图上下文替代孤立的原始节点，这些上下文捕捉多模态语义和跨模态关系。这些上下文将特定领域的节点模式映射到共享的高层次概念，从而减少对目标领域监督或适应的依赖。一个模态感知的图上下文编码器将多模态信息与图结构整合，并将生成的表示转换为大型语言模型的图令牌。实验结果显示，在零样本多模态图任务上取得了一致的改进。
cs.AI / 111 / 2607.26034
Falling Behind Drives Unsafe Development in an Idealised AI Race Experiment
落后驱动理想化人工智能竞赛实验中的不安全发展
Abstract
Technological races create tension between speed and safety: actors may gain by moving faster than competitors, even when risky development is harmful. This is prominent in debates about artificial intelligence (AI), where competitive pressure is often argued to incentivise riskier, less safety-conscious development. We study this using a framed behavioural experiment based on an idealised AI race, in which paired participants repeatedly chose between Safe and Unsafe development under an uncertain time horizon. Unsafe development gave faster progress and higher immediate payoffs but accumulated private risk up to a treatment-specific maximum of 10\%, 60\%, or 90\%; the race's competitive structure was held constant, and only this maximum risk varied. Neither the pre-registered comparison between risk levels nor the role of elicited risk preferences was supported by the data. Instead, exploratory analyses motivated by the task's repeated structure show that Unsafe behaviour is shaped less by risk preferences than by the evolving strategic state of the race: participants are more likely to choose Unsafe after their opponent does so, being ahead reduces Unsafe play while falling behind increases it, and first-round choices predict later behaviour. To interpret these effects we introduce a reduced evolutionary model with four strategies -- Always Safe, Always Unsafe, Conditionally Safe, and Conditionally Antisocial Safe -- which reproduces the treatment effect and shows how conditional Unsafe behaviour can be favoured by competitive race dynamics. Together, the experiment and model show that unsafe development can emerge from early behavioural momentum, opponent behaviour, and fear of falling behind, rather than from risk preferences alone, suggesting policy should focus on reducing competitive pressure and promoting cooperation in AI development rather than only individual risk.
Chinese Translation
技术竞赛在速度与安全之间产生紧张关系：参与者可能通过比竞争对手更快地行动而获益，即便冒险的发展是有害的。这在关于人工智能（AI）的辩论中尤为明显，竞争压力常常被认为会激励更冒险、较少关注安全的发展。我们通过一个基于理想化AI竞赛的框架行为实验来研究这一问题，在该实验中，成对参与者在不确定的时间范围内反复选择安全（Safe）和不安全（Unsafe）发展。不安全的发展带来了更快的进展和更高的即时收益，但累积的私人风险最高可达特定处理的10%、60%或90%；竞赛的竞争结构保持不变，只有这一最大风险有所变化。数据未支持预注册的风险水平比较或引发的风险偏好的作用。相反，基于任务的重复结构的探索性分析表明，不安全行为的形成更受竞赛的演变战略状态的影响，而非风险偏好：参与者在对手选择不安全后更可能选择不安全，领先会减少不安全行为，而落后则会增加不安全行为，首轮选择可以预测后续行为。为了解释这些效应，我们引入了一个简化的进化模型，包含四种策略——始终安全（Always Safe）、始终不安全（Always Unsafe）、有条件安全（Conditionally Safe）和有条件反社会安全（Conditionally Antisocial Safe）——该模型再现了处理效应，并展示了竞争竞赛动态如何使得有条件的不安全行为受到青睐。实验和模型共同表明，不安全的发展可能源于早期行为的惯性、对手行为以及对落后的恐惧，而不仅仅是风险偏好，这表明政策应关注减少竞争压力并促进AI发展中的合作，而不仅仅是个体风险。
cs.AI / 112 / 2607.26041
Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?
桌面增量基准：计算机使用模型是否理解桌面图形用户界面的过渡？
Abstract
Computer-use agents (CUAs) increasingly act through desktop GUIs to complete long-horizon tasks. Current benchmarks primarily measure end-task success or single-frame grounding. Neither isolates whether a model can reconstruct the causal, task-relevant transition produced by an action- crucial for rejecting stale observations, verifying progress, and recovering from failure. This is difficult because inference, remote input, app rendering, and screenshot capture are asynchronous: the next observation may be delayed, occluded, transient, or unrelated, then misread as progress and carried into subsequent planning. We introduce Desktop-Delta Bench (DDB), an offline step-level benchmark with 2,013 human-verified instances from novel, multi-app Linux trajectories across ~15 applications and 50 task domains. DDB trajectories targets 3 failure dimensions- state verification, source tracking, and context-aware control- through 2 complementary tasks: 463 3-frame temporal-ordering instances, including 105 with a cross-trajectory decoy, and 1,550 before-after pairs labeled from 5 actions + its payload. We evaluate 8 closed and open-source model families across 32 ordering and 16 single-action settings, observing consistent gaps. Ordering remains unsaturated: best non-decoy and decoy exact-match rates are 65.1% and 65.7%. Task context improves decoy identification by 6.9 percentage points but reduces non-decoy exact match by 2.2 points; error analysis reveals systematic copying of the presented A-B-C order. Single-action results show that inferring the action family is harder than locating it: click F1 is 0.96 vs, 0.76 for drag, while recognized drags are generally localized well. DDB, thus, complements end-to-end benchmarks by filling the missing diagnostic layer between GUI grounding and final task success, enabling targeted improvements to desktop CUA verification, reliability, and recovery.
Chinese Translation
计算机使用代理（CUAs）越来越多地通过桌面图形用户界面（GUI）执行长时间任务。目前的基准主要测量最终任务的成功或单帧的基础。两者都无法判断模型是否能够重建由动作产生的因果、任务相关的过渡——这对于拒绝过时观察、验证进展和从失败中恢复至关重要。这是困难的，因为推理、远程输入、应用渲染和屏幕截图捕获是异步的：下一个观察可能会延迟、被遮挡、短暂或无关，从而被误读为进展并带入后续规划。我们引入了桌面增量基准（Desktop-Delta Bench，DDB），这是一个离线的逐步基准，包含来自约15个应用程序和50个任务领域的新颖多应用Linux轨迹的2013个经过人工验证的实例。DDB轨迹针对3个失败维度——状态验证、源追踪和上下文感知控制——通过2个互补任务：463个三帧时间排序实例，包括105个带有跨轨迹诱饵的实例，以及1550个从5个动作及其有效载荷标记的前后对。我们在32个排序和16个单一动作设置中评估了8个闭源和开源模型系列，观察到一致的差距。排序仍然未饱和：最佳非诱饵和诱饵的准确匹配率分别为65.1%和65.7%。任务上下文将诱饵识别率提高了6.9个百分点，但将非诱饵的准确匹配率降低了2.2个百分点；错误分析揭示了对呈现的A-B-C顺序的系统性复制。单一动作的结果表明，推断动作类别比定位它更困难：点击F1的准确率为0.96，而拖动的准确率为0.76，尽管识别的拖动通常定位良好。因此，DDB通过填补GUI基础与最终任务成功之间缺失的诊断层，补充了端到端基准，能够针对性地改善桌面CUA的验证、可靠性和恢复能力。
cs.CL / 1 / 2607.24750
TimeCapsule: Generative Hallucination as a Method for Historical Sensemaking
时间胶囊：生成性幻觉作为历史理解的方法
Abstract
Large Language Models (LLMs) are temporally overexposed: trained on vast contemporary corpora, they encode present-day concepts that make them unreliable narrators of the past. We present TimeCapsule, a 1.2B-parameter LLaMA-style causal model trained exclusively on Victorian texts (1800-1875) as an epistemologically isolated generative archive. Quantitative evaluation shows a 45.4% perplexity reduction over a GPT-2 baseline on held-out Victorian prose, while larger contemporary causal models achieve lower raw perplexity through broader pretraining but lack temporal isolation. TimeCapsule exhibits computational sensemaking, generating historically plausible analogical explanations for unfamiliar modern concepts (e.g., describing a computer as a "hypertrophied lung"). A qualitative hermeneutic probe with two humanities scholars revealed a crisis of authenticity, as both misclassified approximately 40% of genuine Victorian excerpts as machine-produced. We argue that structural ignorance of the future transforms hallucinations into interpretive probes of nineteenth-century ontologies.
Chinese Translation
大型语言模型（LLMs）在时间上过度曝光：它们在庞大的当代语料库上训练，编码了现代概念，使其成为不可靠的过去叙述者。我们提出了时间胶囊（TimeCapsule），这是一个仅在维多利亚时期文本（1800-1875）上训练的1.2B参数LLaMA风格因果模型，作为一个认识论上隔离的生成档案。定量评估显示，在保留的维多利亚散文上，相较于GPT-2基线，困惑度降低了45.4%。而较大的当代因果模型通过更广泛的预训练实现了更低的原始困惑度，但缺乏时间隔离。时间胶囊展现了计算上的理解能力，为不熟悉的现代概念生成历史上合理的类比解释（例如，将计算机描述为“肥大肺”）。与两位人文学科研究者的定性诠释探讨揭示了真实性危机，因为两者均将约40%的真实维多利亚摘录误分类为机器生成。我们认为，对未来的结构性无知将幻觉转化为对十九世纪本体论的解释性探针。
cs.CL / 2 / 2607.24765
Measuring and Improving Behavioral Consistency in Large Language Models through Fact-Heuristic-Emotion State Enforcement
通过事实-启发式-情感状态强制来测量和改善大型语言模型的行为一致性
Abstract
Large language models (LLMs) can give different answers to the same decision problem across runs, and reverse a decision when their own prior answer returns as context. We ask whether this instability can be measured and partially reduced without changing model weights. We test the Cognitive Kernel Model (CKM), a prompt-level state-enforcement layer. Before deciding, the model must separate its input into three epistemic roles: Fact (given or verifiable), Heuristic (inferred or assumed), and Emotion (evaluative or priority signal). CKM adds no capability; it forces the model to track what kind of information it uses before acting. Formally it maintains a structured state S_t = {F_t, H_t, E_t} updated by a transition function. We evaluate CKM on Korean-language decision scenarios (ambiguity, ethical conflict, resource allocation, error handling) across 26 LLMs from four vendors and 37,403 observations, via four core experiments, a 4-arm ablation, a 5-arm sham-restriction ablation, and a temperature probe. Findings: (1) CKM reduces repeated-output variability (random-effects Hedges' g=1.09, 95% CI [0.83, 1.35], 31 model pairs); (2) state persistence cuts the decision-flip rate by 82% in newer models (g=1.52); (3) the effect is not JSON formatting alone (value-only recomputation, g=2.24); (4) intrinsic randomness under fixed anchor states is negligible; (5) the advantage grows under sampling stochasticity (g=2.87 at temperature 0.7); (6) a sham ablation attributes about 45% of the gain to structural scaffolding and 55% to Fact/Heuristic/Emotion content, and CKM is the only arm that both raises consistency and reduces flipping. CKM does not improve reasoning correctness. The narrower result: behavioral consistency is measurable, varies across models, and is partially improvable by forcing models to separate facts, assumptions, and evaluative signals before deciding.
Chinese Translation
大型语言模型（LLMs）在不同运行中可能会对同一决策问题给出不同的答案，并在其先前答案作为上下文返回时反转决策。我们探讨这种不稳定性是否可以被测量并在不改变模型权重的情况下部分减少。我们测试了认知内核模型（Cognitive Kernel Model, CKM），这是一种提示级别的状态强制层。在做出决策之前，模型必须将其输入分为三种认知角色：事实（可给定或可验证）、启发式（推断或假设）和情感（评估或优先信号）。CKM并不增加能力；它迫使模型在行动之前跟踪其使用的信息类型。形式上，它维护一个结构化状态 S_t = {F_t, H_t, E_t}，由转移函数更新。我们在26个来自四个供应商的LLM上评估CKM，针对韩语决策场景（模糊性、伦理冲突、资源分配、错误处理），通过四个核心实验、一个4臂消融实验、一个5臂虚假限制消融实验和一个温度探针，共进行了37,403次观察。研究结果：（1）CKM减少了重复输出的变异性（随机效应Hedges' g=1.09，95% CI [0.83, 1.35]，31对模型）；（2）状态持续性将新模型的决策翻转率降低了82%（g=1.52）；（3）该效应不仅仅是JSON格式化（仅值重新计算，g=2.24）；（4）在固定锚状态下的内在随机性微不足道；（5）在采样随机性下，优势增加（在温度0.7时g=2.87）；（6）虚假消融将约45%的增益归因于结构支撑，55%归因于事实/启发式/情感内容，而CKM是唯一一个既提高一致性又减少翻转的臂。CKM并未改善推理的正确性。更狭义的结果是：行为一致性是可测量的，在模型之间存在差异，并且通过迫使模型在决策之前分离事实、假设和评估信号可以部分改善。
cs.CL / 3 / 2607.24841
Neuromorphic Diffusion Language Models: Addressing Compute and Memory Bottlenecks via Sparsity and Block Denoising
神经形态扩散语言模型：通过稀疏性和块去噪解决计算和内存瓶颈
Abstract
Autoregressive (AR) large language models (LLMs) are inherently inefficient at inference time because each generated token requires accessing the full set of model parameters, leading to low operational intensity and high energy consumption. Masked diffusion language models (MDLMs) partially address this limitation for memory-bound settings by allowing multiple tokens to be generated per parameter access. In order to further enhance inference efficiency on modern platforms with extensive in-chip memory, this work proposes neuromorphic MDLMs (N-MDLMs), which integrate block diffusion with spike-based neuromorphic computation to jointly improve throughput and energy efficiency. While block diffusion increases token throughput by producing multiple tokens per parameter access, spike-induced sparsity reduces effective parameter traffic and computations by skipping inactive channels. To analyze the synergistic effect of sparsity and diffusion, we develop a token-level roofline-inspired model that captures the combined impact of block-parallel generation and spike sparsity on decoding efficiency. Experimental results on translation tasks show that, thanks to spike-induced sparsity, N-MDLMs achieve substantial improvements in energy efficiency and throughput even in compute-bound platforms for which MDLMs would fail to improve over AR-LLMs.
Chinese Translation
自回归（AR）大型语言模型（LLMs）在推理时本质上效率低下，因为每生成一个标记都需要访问完整的模型参数集，这导致操作强度低和能耗高。掩蔽扩散语言模型（MDLMs）在内存受限的环境中部分解决了这一限制，允许每次参数访问生成多个标记。为了进一步提高在现代平台上推理的效率，尤其是在具有大量片上内存的情况下，本文提出了神经形态MDLMs（N-MDLMs），它将块扩散与基于脉冲的神经形态计算相结合，以共同提高吞吐量和能效。块扩散通过每次参数访问生成多个标记来增加标记吞吐量，而脉冲引起的稀疏性则通过跳过不活跃的通道来减少有效参数流量和计算。为了分析稀疏性和扩散的协同效应，我们开发了一种基于标记级屋顶线模型，该模型捕捉了块并行生成和脉冲稀疏性对解码效率的综合影响。在翻译任务上的实验结果表明，得益于脉冲引起的稀疏性，N-MDLMs在计算受限的平台上实现了显著的能效和吞吐量提升，而在这些平台上，MDLMs无法超越AR-LLMs的表现。
cs.CL / 4 / 2607.24868
Research Report on Noise-Shaped One-Bit Coefficients in Discrete Polynomial Fourier Extension
离散多项式傅里叶扩展中噪声形状的一位系数研究报告
Abstract
This report studies noise-shaped one-bit coefficients in normalized discrete polynomial Fourier extension. For first-order Sigma-Delta quantization, the error is written as $e_k=u_k-q_k=\Delta v_k$ with a uniformly bounded state. Discrete summation by parts then yields variation estimates for complex weights and an $O(N^{-1})$ approximation rate on compact parameter sets. For the parabolic phase $\phi_{x,t}(\xi)=x\xi+t\xi^2$, the bound is expressed through $J(x,t)=\int_0^1 |x+2t\xi|d\xi$, and the uniform $N^{-1}$ rate is shown to be sharp over the admissible input class. Higher-order finite-record identities are derived with all endpoint traces retained. Under endpoint compatibility, or after explicit boundary correction, an $r$th-order noise-shaped error $e=\Delta^r v$ gives $O(N^{-r})$ decay for sufficiently smooth weights and $O(N^{-(r-1+\alpha)})$ decay for $C^{r-1,\alpha}$ weights. Exact $L^2$ orthogonality identities, fourth-moment formulas, local kernel estimates, and oscillatory transfer bounds are also established. Extensions to polynomial phases, multidimensional parameter families, growing observation regions, and correlated state models are included.
Chinese Translation
本报告研究了归一化离散多项式傅里叶扩展中的噪声形状一位系数。对于一阶Sigma-Delta量化，误差可表示为 $e_k=u_k-q_k= riangle v_k$，其中状态均匀有界。通过分部离散求和，得到了复权重的变差估计，并在紧致参数集上获得了 $O(N^{-1})$ 的近似率。对于抛物相位 $ heta_{x,t}( au)=x au+t au^2$，界限通过 $J(x,t)= extstyleigint_{0}^{1} |x+2t au|d au$ 表达，且在可接受输入类上均匀的 $N^{-1}$ 速率被证明是尖锐的。推导了高阶有限记录恒等式，保留了所有端点轨迹。在端点兼容性下，或在显式边界修正后，$r$ 阶噪声形状误差 $e= riangle^r v$ 对于足够光滑的权重给出了 $O(N^{-r})$ 的衰减，对于 $C^{r-1,eta}$ 权重则给出了 $O(N^{-(r-1+eta)})$ 的衰减。此外，还建立了精确的 $L^2$ 正交恒等式、四阶矩公式、局部核估计和振荡转移界限。报告还包括对多项式相位、多维参数族、增长观察区域和相关状态模型的扩展。
cs.CL / 5 / 2607.24999
CogArena: A Multimethod Evaluation of Cognitive Ability Structure in Large Language Models
CogArena：大型语言模型认知能力结构的多方法评估
Abstract
LLM cognitive scores are increasingly summarized as per-ability profiles whose dimensions should converge across tasks, respond selectively to matched interventions, and generalize beyond the models used to define them. We introduce CogArena, a procedurally generated 13-paradigm benchmark built around a multimethod framework for determining when cognitive-task scores warrant dimensional labels across five theory-motivated groupings. Across 55 open-weight models, nearly all paradigm correlations are positive and a common axis explains about half the variance. The within-grouping advantage is small, scoring-sensitive, and uncertain across model families. In a separately frozen, fully crossed study across 12 models from six families, targeted scaffolds show a small matched-grouping advantage, but no scaffold-specific contrast survives multiplicity correction and selectivity does not improve held-out-family prediction. The frozen confirmation criterion fails. A post-hoc alternate-wording replication produces a smaller positive estimate and again fails. Together, these results support a boundary conclusion. Theory-aligned prompting produces a small in-battery diagonal tendency, but the present evidence does not establish stable five-dimensional profiles. CogArena provides a workflow joining behavioral signatures, covariance, matched interventions, and out-of-family prediction before cognitive labels are attached to model scores.
Chinese Translation
大型语言模型（LLM）的认知评分越来越多地被总结为按能力划分的特征，其维度应在不同任务中趋于一致，能够对匹配的干预措施做出选择性响应，并且能够超越用于定义它们的模型进行推广。我们介绍了CogArena，这是一个程序生成的13种范式基准，围绕多方法框架构建，用于确定何时认知任务评分应在五个理论驱动的分组中获得维度标签。在55个开放权重模型中，几乎所有范式相关性均为正相关，并且一个共同轴解释了约一半的方差。组内优势较小，对评分敏感，并且在模型家族之间存在不确定性。在一个单独冻结的、完全交叉的研究中，涉及来自六个家族的12个模型，目标支架显示出小的匹配组优势，但没有支架特定的对比在多重性校正后存活，选择性也未能改善对保留家族的预测。冻结确认标准未能成立。事后替代措辞的重复实验产生了较小的正估计，并再次失败。综合来看，这些结果支持了一个边界结论。与理论对齐的提示产生了小的电池内对角倾向，但目前的证据并未建立稳定的五维特征。CogArena提供了一种工作流程，将行为特征、协方差、匹配干预和超出家族的预测结合在一起，然后再将认知标签附加到模型评分上。
cs.CL / 6 / 2607.25069
DS@GT ARC at CheckThat! 2026: LLM-Based Trace Ranking and Grouped Reward Modeling for Multilingual Numerical Claim Verification
DS@GT ARC 在 CheckThat! 2026：基于大语言模型的追踪排名和分组奖励建模用于多语言数值声明验证
Abstract
Automated verification of numerical claims is a challenging problem, as it requires both language understanding and quantitative reasoning. This paper describes our system for CLEF 2026 CheckThat! Task 2, which focuses on ranking reasoning traces generated by large language models (LLMs) and predicting a final verdict for numerical claims in English and Arabic. We explore two approaches. The first approach fine-tunes an LLM-based verifier using LoRA to score each reasoning trace independently as a binary classification problem, and selects the final verdict using Best-of-N selection. We further experiment with adaptive sub-claim decomposition to break complex claims into simpler parts before verification. The second approach uses a lightweight TF-IDF reward model with handcrafted numeric and temporal overlap features to score traces, and aggregates scores by verdict group to determine the final prediction. For Arabic, we compare a general multilingual model against AraBERT, a language-specific model pretrained on Arabic text. Our results show that the LLM-based approach outperforms the lightweight reward model on most metrics, particularly Recall@5, while the reward-based approach shows stronger performance on the Conflicting class. Sub-claim decomposition did not improve performance, suggesting that claim splitting introduces noise rather than aiding reasoning. For Arabic, AraBERT outperforms the multilingual baseline across most metrics.
Chinese Translation
自动化验证数值声明是一项具有挑战性的问题，因为它需要语言理解和定量推理。本文描述了我们在 CLEF 2026 CheckThat! 任务 2 中的系统，该系统专注于对由大型语言模型（LLMs）生成的推理追踪进行排名，并预测英语和阿拉伯语数值声明的最终裁决。我们探索了两种方法。第一种方法使用 LoRA 微调基于 LLM 的验证器，将每个推理追踪独立地作为二元分类问题进行评分，并通过最佳选择（Best-of-N）选取最终裁决。我们进一步实验了自适应子声明分解，以在验证之前将复杂声明拆分为更简单的部分。第二种方法使用轻量级的 TF-IDF 奖励模型，结合手工制作的数值和时间重叠特征来评分追踪，并通过裁决组聚合评分以确定最终预测。对于阿拉伯语，我们将通用多语言模型与 AraBERT（一种在阿拉伯文本上预训练的语言特定模型）进行了比较。我们的结果表明，基于 LLM 的方法在大多数指标上优于轻量级奖励模型，特别是在 Recall@5 上，而基于奖励的方法在冲突类上表现更强。子声明分解未能提高性能，这表明声明拆分引入了噪声而不是帮助推理。对于阿拉伯语，AraBERT 在大多数指标上优于多语言基线。
cs.CL / 7 / 2607.25094
Evaluating Communicative Belief Updates in Large Language Models via Implicature Recognition and Cancellation
通过隐含意义识别与取消评估大型语言模型中的交际信念更新
Abstract
Human language is driven by unspoken beliefs and belief updates, making these critical to model for successful communication between large language models (LLMs) and their users. In this paper, we evaluate the ability of LLMs to recognize unspoken beliefs made through implicatures and to understand their updates through implicature cancellation: the pragmatic phenomenon whereby an utterance's implied meaning is weakened or negated. We create the first expert-annotated implicature cancellation dataset, [DatasetName], crowdsourced for human judgements of implicatures and their corresponding cancellations. We find that LLM belief update understanding lags behind that of humans, especially in more naturally-occurring scenarios. Additional control experiments suggest that successes in LLM belief updates may stem in part from a reliance on prior beliefs, and that failures in belief updates may depend on their type and on their form. Overall, our study suggests that current LLMs have not yet reached human-level understanding of unspoken beliefs and belief updates. Code and data are available at https://github.com/cesare-spinoso/ImplicatureX.
Chinese Translation
人类语言受到未言明信念和信念更新的驱动，这对于大型语言模型（LLMs）与其用户之间成功沟通至关重要。本文评估了LLMs识别通过隐含意义传达的未言明信念的能力，以及通过隐含意义取消理解其更新的能力：这一语用现象指的是话语的隐含意义被削弱或否定。我们创建了第一个专家注释的隐含意义取消数据集[DatasetName]，该数据集通过众包方式收集了关于隐含意义及其相应取消的人工判断。我们的研究发现，LLM对信念更新的理解落后于人类，尤其是在更自然发生的场景中。额外的控制实验表明，LLM在信念更新方面的成功可能部分源于对先前信念的依赖，而信念更新的失败可能与其类型和形式有关。总体而言，我们的研究表明，目前的LLMs尚未达到人类对未言明信念和信念更新的理解水平。代码和数据可在https://github.com/cesare-spinoso/ImplicatureX获取。
cs.CL / 8 / 2607.25129
Deep Label-Wise Attentive Temporal Convolutional Networks Improve Medical Coding
深度标签注意力时间卷积网络提升医疗编码
Abstract
Medical coding is the task of assigning a set of diagnosis and procedure codes for a hospitalization using recorded notes. It requires aggregating information from different parts of the text and focus to different sections for each individual code, making it a very difficult problem even for professional human coders. We model the task as a multi-label text classification problem. To overcome the mentioned difficulties, we propose a deep neural model consisting of a multi-layer temporal convolution network (TCN) followed by label-wise attention. While multi-layer TCN helps extract a global document representation with the ability to learn relations over very long sequences, label-specific attention mechanism allows the model to focus on different aspects of the same document for each individual label. Our method achieves significantly better F-1 scores (9% increase) compared to the previous state-of-the-art model, with a remarkable increase in recall score (28% increase), which we believe is the more important metric for a clinical decision support setting.
Chinese Translation
医疗编码是将住院期间的诊断和手术代码分配给一组记录笔记的任务。这需要从文本的不同部分汇总信息，并针对每个单独的代码关注不同的部分，这使得即使是专业的人类编码员也面临着非常困难的问题。我们将该任务建模为一个多标签文本分类问题。为了克服上述困难，我们提出了一种深度神经模型，该模型由多层时间卷积网络（Temporal Convolutional Network, TCN）和标签级注意力机制组成。多层 TCN 有助于提取具有学习非常长序列关系能力的全局文档表示，而标签特定的注意力机制则允许模型针对每个单独标签关注同一文档的不同方面。与之前的最先进模型相比，我们的方法在 F-1 分数上显著提高（增加 9%），并且在召回率上有显著提升（增加 28%），我们认为这是临床决策支持环境中更重要的指标。
cs.CL / 9 / 2607.25182
TabRank: Chain-of-Thought Distillation for Table Re-Rankers
TabRank：用于表格重排序的思维链蒸馏
Abstract
The ability to retrieve relevant tables for answering questions is a key task for structured information retrieval. Multi-stage retrieval systems rely heavily on rerankers to refine candidate lists produced by efficient first-stage retrievers. As a result, neural rerankers and LLM-based reranking methods have become increasingly important due to their superior capacity for semantic understanding and reasoning compared to conventional sparse or dense retrieval models. Recently, Large Reasoning Models (LRMs) equipped with explicit chain-of-thought (CoT) reasoning have shown strong improvements in ranking quality in unstructured passage retrieval. In this work, we present TabRank, a framework for training reasoning rerankers for Tabular Retrieval. We first present a comprehensive dataset of 6728 reasoning traces for tabular reranking on the Natural Questions Tables dataset. We then explore two variants of training a compact reasoning model on these reasoning traces: explicit CoT distillation and conditioning the student reranker on the teacher's reasoning trace within the prompt. We stress-test TabRank on several out-of-distribution generalization settings on diverse domains and multi-table scenarios. Our approach significantly improves performance across a variety of table retrieval datasets, increasing Acc@10 by 30.5% on HybridQA, 15.2% on SQA, 52.9% on TabFact, and 13.1% on TATQA subsets of the Multi-Table QA Benchmark compared to the base model. Notably, TabRank generalizes effectively to multi-table reasoning. Our code, data and models are available at https://github.com/AdarshSingh7647/TabRanker
Chinese Translation
检索相关表格以回答问题的能力是结构化信息检索中的一项关键任务。多阶段检索系统在很大程度上依赖重排序器来优化由高效的第一阶段检索器生成的候选列表。因此，神经重排序器和基于大语言模型（LLM）的重排序方法因其在语义理解和推理方面的优越能力而变得越来越重要，相较于传统的稀疏或密集检索模型。最近，配备显式思维链（CoT）推理的大型推理模型（LRM）在非结构化段落检索中的排名质量上显示出显著改善。在本研究中，我们提出了TabRank，一个用于表格检索的推理重排序器训练框架。我们首先提供了一个包含6728个推理轨迹的综合数据集，用于在自然问题表格数据集上进行表格重排序。然后，我们探索了在这些推理轨迹上训练紧凑推理模型的两种变体：显式CoT蒸馏和在提示中将学生重排序器的条件设置为教师的推理轨迹。我们在多个分布外泛化设置以及多表场景中对TabRank进行了压力测试。与基础模型相比，我们的方法在多种表格检索数据集上显著提高了性能，在HybridQA上提高了30.5%的Acc@10，在SQA上提高了15.2%，在TabFact上提高了52.9%，在Multi-Table QA Benchmark的TATQA子集上提高了13.1%。值得注意的是，TabRank在多表推理中有效泛化。我们的代码、数据和模型可在https://github.com/AdarshSingh7647/TabRanker获取。
cs.CL / 10 / 2607.25184
A scaling law of contextual persistence in human language
人类语言中的上下文持久性缩放法则
Abstract
Human language exhibits lawful structure at the level of words (frequency, vocabulary growth) and word pairs (co-occurrence across distance). Here we show that the arrangement of words in sequence -- a central determinant of meaning -- obeys a comparable law. Using large language models as probabilistic probes, we measured the reduction in target perplexity conferred by prior context at distance d beyond that of the same words scrambled; this difference, the contextual persistence function P(d), isolates the influence of arrangement. Across ten corpora spanning six language families and written and spoken modalities, P(d) decayed approximately as 1/d ($P(d) \propto d^{-\alpha}$, mean $\alpha = 1.04$; median $r^2 = 0.96$). The effect vanished in scrambled and synthetic controls, replicated across independent probes, and did not appear in genomic or protein sequences under domain-native models. An exponent near 1 distributes contextual influence approximately uniformly across logarithmic timescales. The results establish a scaling law of contextual persistence in human language.
Chinese Translation
人类语言在词汇（频率、词汇增长）和词对（跨距离的共现）层面上表现出规律性的结构。本文展示了词语在序列中的排列——意义的一个核心决定因素——遵循类似的法则。我们使用大型语言模型作为概率探针，测量了在距离 d 处，先前上下文所带来的目标困惑度的降低，超出了相同词语被打乱后的困惑度；这一差异，即上下文持久性函数 P(d)，隔离了排列的影响。在跨越六个语言家族的十个语料库中，无论是书面还是口语形式，P(d) 的衰减大致呈现为 1/d（$P(d) ext{propto} d^{-eta}$，平均 $eta = 1.04$；中位数 $r^2 = 0.96$）。在打乱和合成对照组中，该效应消失，并在独立探针中得到了重复验证，而在基因组或蛋白质序列中则未出现该效应。接近 1 的指数在对数时间尺度上大致均匀地分布了上下文影响。研究结果确立了人类语言中上下文持久性的缩放法则。
cs.CL / 11 / 2607.25186
MyoCardBench: A Real-World Data Benchmark for Evaluating Large Language Models in Clinically Authentic Cardiovascular Care Scenarios
MyoCardBench：用于评估大型语言模型在临床真实心血管护理场景中的表现的真实世界数据基准
Abstract
Background: Most medical large language model (LLM) benchmarks focus on examination knowledge or isolated tasks and may not reflect the longitudinal, multimodal, and safety-critical workflow of cardiovascular care. Objective: To develop MyoCardBench, a real-world benchmark spanning the cardiovascular care continuum, and assess LLM performance across clinical dimensions and specialist tasks. Methods: MyoCardBench includes 2,263 items from 13 task-specific datasets derived from de-identified cardiovascular records and examination data. Sixteen cardiology physicians conducted annotation and reference construction, followed by cross-review from two senior cardiologists. Seven LLMs generated 15,841 outputs under standardized zero-shot settings. Open-ended tasks were evaluated using key-point coverage and holistic clinical quality, while CardioEthics was scored by accuracy. Results: GPT-5.4 achieved the highest macro-average (62.55) and item-weighted mean (62.19), followed by Gemini 3.1 Pro (59.95) and Qwen 3.6 27B (59.72). GPT-5.4 ranked first in all three dimensions. CardioAuxReport performed best (86.38), whereas CardioECGRead (17.25) and CardioEthics (17.34) were lowest. The largest gaps between holistic clinical quality and key-point coverage occurred in CardioComm (52.71), CardioEmergRescue (52.05), and CardioTreatPlan (48.80). Conclusions: To our knowledge, MyoCardBench is the largest real-world, multi-task benchmark for LLM evaluation across the cardiovascular care continuum and offers the broadest coverage of clinically authentic cardiology scenarios reported to date. It provides a rigorous framework for identifying model strengths, clinically important omissions, and priorities for future development.
Chinese Translation
背景：大多数医学大型语言模型（LLM）基准测试侧重于考试知识或孤立任务，可能无法反映心血管护理的纵向、多模态和安全关键工作流程。目的：开发MyoCardBench，一个涵盖心血管护理连续体的真实世界基准，并评估LLM在临床维度和专业任务中的表现。方法：MyoCardBench包含来自13个特定任务数据集的2263个项目，这些数据集源自去标识化的心血管记录和检查数据。十六位心脏病医生进行了注释和参考构建，随后由两位资深心脏病专家进行了交叉审查。七个LLM在标准化的零样本设置下生成了15841个输出。开放式任务通过关键点覆盖和整体临床质量进行评估，而CardioEthics则通过准确性评分。结果：GPT-5.4获得了最高的宏平均（62.55）和项目加权平均（62.19），其次是Gemini 3.1 Pro（59.95）和Qwen 3.6 27B（59.72）。GPT-5.4在所有三个维度中排名第一。CardioAuxReport表现最佳（86.38），而CardioECGRead（17.25）和CardioEthics（17.34）得分最低。整体临床质量与关键点覆盖之间的最大差距出现在CardioComm（52.71）、CardioEmergRescue（52.05）和CardioTreatPlan（48.80）。结论：据我们所知，MyoCardBench是迄今为止最大的真实世界多任务基准，用于评估心血管护理连续体中的LLM，并提供了最广泛的临床真实心脏病场景覆盖。它为识别模型优势、临床重要遗漏和未来开发优先事项提供了严格的框架。
cs.CL / 12 / 2607.25202
A Cross-lingual Comparison of Human and Classification Model Entrainment Behavior in Code-switched Speech Settings
跨语言比较人类与分类模型在代码切换语境中的共鸣行为
Abstract
Conversational entrainment is well-studied in monolingual and written contexts, but remains underexplored in spoken code-switching (CSW). We present a novel cross-lingual analysis of entrainment in Mandarin-English, Hindi-English, and Spanish-English dialogue and show that, while lexical entrainment generalizes across language pairs, entrainment over acoustic-prosodic and CSW style aspects exhibits context-specific variation. We build on these findings by asking whether classification models capture these human behavioral patterns. Applying feature importance and ablation analyses, we find that classical and Transformer-based classifiers detect entrainment reasonably well but consistently prioritize features other than those most salient to human entraining behavior. Our approach introduces a human-grounded framework for evaluating model decision-making in multilingual stylistic contexts, and suggests future challenges for developing conversational agents capable of producing naturalistic code-switched speech.
Chinese Translation
对话共鸣在单语和书面语境中得到了充分研究，但在口语代码切换（CSW）中仍然未被深入探讨。我们提出了一种新颖的跨语言分析，研究普通话-英语、印地语-英语和西班牙语-英语对话中的共鸣现象，并表明，尽管词汇共鸣在语言对之间具有普遍性，但在声学-韵律和代码切换风格方面的共鸣则表现出特定于语境的变化。基于这些发现，我们进一步探讨分类模型是否能够捕捉到这些人类行为模式。通过特征重要性和消融分析，我们发现经典和基于Transformer的分类器能够合理地检测共鸣，但始终优先考虑与人类共鸣行为最显著的特征无关的特征。我们的方法引入了一个以人为基础的框架，用于评估多语言风格语境中模型决策的有效性，并为开发能够产生自然代码切换语音的对话代理提出了未来的挑战。
cs.CL / 13 / 2607.25228
Interpretable Column Annotation with LLM-Symbolized Decision Process Materialization
可解释的列注释与LLM符号化决策过程的实现
Abstract
Column annotation (CA), including column type annotation (CTA) and column property annotation (CPA), aims to identify the meanings of table columns and the semantic relationships among them. Recent CA methods usually use various neural models to learn column representations and directly map them to label categories, thereby (1) sacrificing model interpretability and adaptivity, and (2) overlooking rich label semantics and ultimately limiting accuracy. To address these limitations, we propose SymCA, an LLM-empowered interpretable CA framework that materializes column annotation as a global-to-local symbolic decision process. SymCA consists of two components: (1) global skeleton induction, which constructs a semantic skeleton over the label space, and (2) local substrate evolution, which evolves predictive substrates within the skeleton. Specifically, to exploit label semantics while preserving an interpretable decision process, the global skeleton induction module leverages LLMs to generate candidate hypernym-inspired tree-structured semantic skeletons and employs a Minimum Bayes Risk (MBR)-based consensus strategy to select a robust skeleton against generation variance. Since different internal nodes require different evidence to distinguish among their child nodes, the local substrate evolution module materializes each internal node as an executable and evolvable predictive substrate. Over multiple evolution rounds, each substrate trains an interpretable random forest classifier with the current operator set, leverages the LLM to propose node-specific operator modifications, and uses an exploration-exploitation strategy to prioritize promising substrates. Extensive experiments demonstrate that SymCA is accurate, robust, and interpretable, outperforming the strongest baselines by an average of 6.42% in Micro-F1 and 11.03% in Macro-F1.
Chinese Translation
列注释（CA），包括列类型注释（CTA）和列属性注释（CPA），旨在识别表列的含义及其之间的语义关系。近期的CA方法通常使用各种神经模型来学习列表示，并直接将其映射到标签类别，从而（1）牺牲了模型的可解释性和适应性，以及（2）忽视了丰富的标签语义，最终限制了准确性。为了解决这些局限性，我们提出了SymCA，一个由LLM驱动的可解释CA框架，将列注释实现为一个从全局到局部的符号决策过程。SymCA由两个组成部分构成：（1）全局骨架诱导，构建一个标签空间上的语义骨架，以及（2）局部基质演化，在骨架内演化预测基质。具体而言，为了在保留可解释决策过程的同时利用标签语义，全局骨架诱导模块利用LLM生成候选的受超义词启发的树状语义骨架，并采用基于最小贝叶斯风险（MBR）的共识策略来选择一个对生成方差具有鲁棒性的骨架。由于不同的内部节点需要不同的证据来区分其子节点，局部基质演化模块将每个内部节点实现为一个可执行和可演化的预测基质。在多个演化轮次中，每个基质使用当前操作符集训练一个可解释的随机森林分类器，利用LLM提出节点特定的操作符修改，并使用探索-开发策略来优先考虑有前景的基质。大量实验表明，SymCA准确、鲁棒且可解释，在Micro-F1和Macro-F1上平均超越最强基线6.42%和11.03%。
cs.CL / 14 / 2607.25236
VisualPatchWorld: Code World Models as Latent Structured Representations for Planning
VisualPatchWorld：作为潜在结构化表示的代码世界模型用于规划
Abstract
Different research lines use the term world model in different ways, yet they share a common aim: to capture how the world evolves under action in a form that supports perception, simulation, and planning. Two prominent realizations are neural predictors that learn dynamics in continuous vector spaces, and hand-built physics engines that expose explicit state and physical laws. Neural predictors scale from data but leave the form of the dynamics implicit; physics engines are inspectable and editable but difficult to construct at scale. We introduce VisualPatchWorld (VPW), which represents world dynamics as code. VPW first selects a qualitative dynamical form with short active probes, then fits that form's free parameters from recorded state-action traces by minimizing multi-step prediction error. The resulting programs can be rolled forward like a simulator, inspected in source form, and used inside model-predictive control; image-derived scene graphs can supply the live state at replan time. Across comparisons with prior code-based world models, VPW attains 69.0% mean planning success and exceeds the strongest code baseline by 23.5 points. The largest gains arise when choosing the correct qualitative dynamics is essential. Under the same planner, the induced models approach ground-truth engine success on navigation and grasp-rich control; a residual gap remains for contact-rich pushing, and checking a shortlist of promising plans in the engine closes most of that gap. These results establish a practical route toward automatically constructed code world models that are useful for planning. Code is available at https://github.com/HKBU-KnowComp/VisualPatchWorld/.
Chinese Translation
不同的研究方向对世界模型一词的使用各有不同，但它们的共同目标是：以支持感知、模拟和规划的形式捕捉世界在行动下的演变。两种突出的实现方式是学习连续向量空间中动态的神经预测器，以及暴露显式状态和物理法则的手工构建物理引擎。神经预测器从数据中进行扩展，但将动态的形式隐式化；物理引擎可供检查和编辑，但在大规模构建时困难重重。我们提出了VisualPatchWorld（VPW），它将世界动态表示为代码。VPW首先通过短期的主动探测选择一种定性动态形式，然后通过最小化多步预测误差，从记录的状态-动作轨迹中拟合该形式的自由参数。生成的程序可以像模拟器一样向前滚动，以源代码形式进行检查，并在模型预测控制中使用；图像衍生的场景图可以在重新规划时提供实时状态。在与先前基于代码的世界模型的比较中，VPW达到了69.0%的平均规划成功率，比最强的代码基线高出23.5个百分点。当选择正确的定性动态至关重要时，获得的增益最大。在相同的规划器下，诱导模型在导航和丰富抓取控制方面接近真实引擎的成功；在接触丰富的推送任务中仍存在一定差距，而在引擎中检查一份有前景的计划短名单可以弥补大部分差距。这些结果为自动构建对规划有用的代码世界模型提供了一条实用的路径。代码可在 https://github.com/HKBU-KnowComp/VisualPatchWorld/ 获取。
cs.CL / 15 / 2607.25270
Where Steering Signals Come From: Activation Source Selection in Activation Steering
引导信号来源：激活引导中的激活源选择
Abstract
Activation steering controls language models by adding vectors or features to hidden states at inference time, but the upstream source of these steering signals is often treated as a secondary detail. We study this source choice as activation source selection: the combination of source context and activation readout policy used to collect the hidden states from which a steering signal is built. Holding the downstream intervention fixed, we show across three instruction-tuned models and four steering task families that changing only the source activations substantially changes steering success. We further find that effective steering is not explained simply by whether the desired behavior appears in the source text. Instead, strong signals come from execution-boundary states, where the model is about to produce or continue the target behavior. This pre-/post-realization distinction explains why answer-based sources sometimes work: their useful component aligns with execution-boundary directions rather than target appearance alone. Building on this view, we introduce tail subtraction, which removes shared prompt and continuation semantics from boundary states and yields cleaner, more stable steering signals. Overall, our results suggest that steering depends on representations of what the model is about to do, not merely on what has already appeared.
Chinese Translation
激活引导通过在推理时向隐藏状态添加向量或特征来控制语言模型，但这些引导信号的上游来源通常被视为次要细节。我们将这种来源选择研究为激活源选择：即用于收集构建引导信号的隐藏状态的源上下文与激活读取策略的组合。在固定下游干预的情况下，我们在三个经过指令调优的模型和四个引导任务家族中展示，仅改变源激活会显著改变引导成功。我们进一步发现，有效的引导并不仅仅由期望行为是否出现在源文本中来解释。相反，强信号来自执行边界状态，即模型即将产生或继续目标行为的状态。这种前后实现的区分解释了为什么基于答案的源有时有效：它们的有用成分与执行边界方向对齐，而不仅仅是目标出现。基于这一观点，我们引入了尾部减法，它从边界状态中去除共享的提示和延续语义，从而产生更干净、更稳定的引导信号。总体而言，我们的结果表明，引导依赖于模型即将执行的内容的表征，而不仅仅是已经出现的内容。
cs.CL / 16 / 2607.25291
CoSA: Accelerating Long-Context Inference via Proxy-Kernel Co-Designed Sparse Attention
CoSA：通过代理内核共同设计的稀疏注意力加速长上下文推理
Abstract
The quadratic cost of self-attention makes long-context inference prohibitively expensive, and proxy-based block-sparse attention has become a practical remedy. Existing methods typically rely on a proxy to predict a binary sparse mask and a kernel to consume this mask and perform sparse attention computation. Such an approach is effective under moderate budgets. However, as the budget tightens, the estimated proxy inevitably drops some salient blocks, while the kernel can only apply the sparse mask mechanically, leading to an evident drop in model accuracy. We propose CoSA, a two-stage training-free Sparse Attention under proxy-kernel CO-design, which couples a Kernel-Aware Proxy (KAP) with an Ordered-Skipping Kernel (OSK). In the first stage, the KAP selects blocks under a moderate budget and produces an ordered mask that prescribes the order in which KV pages are visited in the kernel inner loop. In the second stage, the OSK applies this mask and skips more blocks under a tightened budget given online-softmax statistics. Across mainstream LLM backbones and long-context benchmarks, CoSA attains higher accuracy at lower budgets. Impressively, CoSA achieves a 4.93$\times$ attention speedup and reduces end-to-end Time-to-First-Token by 2.53$\times$ under a context length of 128K with negligible performance degradation.
Chinese Translation
自注意力的二次成本使得长上下文推理变得极其昂贵，而基于代理的块稀疏注意力已成为一种实用的解决方案。现有方法通常依赖于代理来预测二进制稀疏掩码，并使用内核来消耗该掩码并执行稀疏注意力计算。这种方法在适度预算下是有效的。然而，随着预算的收紧，估计的代理不可避免地会丢失一些显著的块，而内核只能机械地应用稀疏掩码，导致模型准确性明显下降。我们提出了CoSA，一种在代理-内核共同设计下的两阶段无训练稀疏注意力，它将内核感知代理（Kernel-Aware Proxy, KAP）与有序跳过内核（Ordered-Skipping Kernel, OSK）结合在一起。在第一阶段，KAP在适度预算下选择块，并生成一个有序掩码，规定内核内部循环中KV页面的访问顺序。在第二阶段，OSK应用此掩码，并在给定在线软最大统计信息的情况下跳过更多块。通过主流的LLM骨干网络和长上下文基准，CoSA在较低预算下实现了更高的准确性。令人印象深刻的是，CoSA在上下文长度为128K的情况下实现了4.93倍的注意力加速，并将端到端的首次令牌时间减少了2.53倍，且性能降级微乎其微。
cs.CL / 17 / 2607.25305
Toward a systematic method for identifying language areas
朝向一种系统化的方法以识别语言区域
Abstract
Macroareas are geographical areas used in typological research for grouping variables of interest. In linguistic typology, languages in a given macroarea are considered to have potential for contact, in contrast to those outside the area, where contact is less likely. Along with language family membership, macroareas are used as controls for models in linguistic typology, in an attempt to address the problem of autocorrelation - the observation that historical developments or typological patterns may be due to contact between neighboring languages and/or inheritance from a common ancestral language. Macroareas are therefore a central aspect of research that seeks to separate universal properties of language from local (or language-specific) properties. Existing macroareas largely depend on expert determinations of what constitutes a geographical area of potential contact, and to date have mainly aligned with continents or landmasses (Hammarstr\"om and Donohue 2014; Nichols, Witzlack-Makarevich, and Bickel 2013). While there are various historical and theoretical reasons for these groupings, there as of yet has been no systematic approach to identifying such areas for a given region. This paper attempts to address such a gap and move beyond macroarea to identification of language areas of relatively arbitrary size, presenting a simple geographical clustering method for identifying groupings over any area. The method produces a set of worldwide macroareas that largely align with existing groupings, as well as local groupings for a well-known sprachbund.
Chinese Translation
宏区域是用于类型学研究的地理区域，用于对感兴趣的变量进行分组。在语言类型学中，某一宏区域内的语言被认为具有接触的潜力，相较之下，宏区域外的语言则接触的可能性较小。除了语言家族归属外，宏区域还被用作语言类型学模型的控制变量，以试图解决自相关问题——即历史发展或类型模式可能是由于邻近语言之间的接触和/或来自共同祖先语言的继承。因此，宏区域是研究旨在将语言的普遍特性与地方（或语言特有）特性分离的核心方面。现有的宏区域主要依赖于专家对潜在接触地理区域的判断，并且迄今为止主要与大陆或陆块相一致（Hammarström 和 Donohue 2014；Nichols, Witzlack-Makarevich 和 Bickel 2013）。虽然这些分组有各种历史和理论原因，但至今尚未有系统的方法来识别特定区域的此类区域。本文试图填补这一空白，并超越宏区域，识别相对任意大小的语言区域，提出了一种简单的地理聚类方法，用于识别任何区域内的分组。该方法生成了一组全球宏区域，基本上与现有分组一致，并为一个著名的语言联盟提供了地方分组。
cs.CL / 18 / 2607.25308
CAST: Game Solvers as Turn-Level Teachers for LLM Agents
CAST：作为回合级教师的游戏求解器为大型语言模型代理提供指导
Abstract
Training large language models (LLMs) to act in long-horizon games is a promising step toward generalist decision-making, yet reinforcement learning with verifiable rewards (RLVR) relies on sparse final rewards that reveal little about which decisions determine success. Denser process signals could supply this missing turn-level credit, but existing sources are hard to keep both cheap and accurate. We observe that changes in a game solver's state value reveal whether an action advances the state toward success. Building on this insight, we propose CAST (Credit Assignment from Solver Teachers), which converts these value changes into solver advantages and injects them into RLVR as turn-level signals. We further show that, under a soft-optimal solver assumption, maximizing the solver advantage is equivalent to on-policy distillation from the solver, requiring only scalar values rather than teacher logits. Across Sokoban, Minesweeper, and Rush Hour, CAST outperforms all trained baselines on every game under both in-domain and unseen-difficulty evaluation and achieves the highest average zero-shot performance on ALFWorld and WebShop. Our code is available at https://github.com/Wloner0809/CAST.
Chinese Translation
训练大型语言模型（LLMs）在长时间跨度的游戏中进行决策是朝向通用决策能力迈出的重要一步，但依赖可验证奖励的强化学习（RLVR）依赖于稀疏的最终奖励，这些奖励对哪些决策决定了成功几乎没有揭示。更密集的过程信号可以提供这种缺失的回合级信用，但现有来源难以同时保持低成本和高准确性。我们观察到，游戏求解器状态值的变化揭示了某个动作是否将状态推进至成功。基于这一洞察，我们提出了CAST（来自求解器教师的信用分配），该方法将这些值变化转化为求解器优势，并将其作为回合级信号注入到RLVR中。我们进一步表明，在软最优求解器假设下，最大化求解器优势等同于从求解器进行的在线策略蒸馏，仅需标量值而非教师的logits。在Sokoban、扫雷和高峰时段游戏中，CAST在所有游戏的领域内和未见难度评估中均优于所有训练基线，并在ALFWorld和WebShop上实现了最高的平均零-shot性能。我们的代码可在https://github.com/Wloner0809/CAST获取。
cs.CL / 19 / 2607.25335
Every Time I Hire a Linguist, Inference Costs Go Down: On Linguistic Rules as Effective Prompt Compressors
每次我雇佣语言学家时，推理成本降低：语言规则作为有效的提示压缩器
Abstract
Prompt compression shortens LLM input to reduce inference cost, yet existing methods score token importance through LM forward passes. It remains questionable whether such nuanced, costly token selection is necessary. Compression requires identifying informative content, a problem that linguistic research has long addressed through cues that can be operationalized as deterministic rules. We therefore ask: can \textbf{linguistic rules alone} serve as effective prompt compressors, without LM-based scoring at compression time? To address this, we conduct offline evolutionary search over lexical, syntactic, semantic, and discourse seeds to find competitive rule combinations. The resulting linguistic compressor requires no LM forward pass at deployment and uses only CPU-side processing for compression. We evaluate it with a dual-path protocol to balance compression quality and reconstruction fidelity. Across short passages, multi-document reasoning, and dialogue-memory QA datasets, evolved compressors achieve performance similar to that of recent advanced prompt-compression strategies. Performance is strongest under light-to-moderate compression and degrades as compression becomes more aggressive, while the Direct and Reconstruction paths exhibit distinct patterns. Evolutionary analysis reveals that effective compression fuses signals across linguistic levels and, as the compression ratio increases, rules shift from token pruning to sentence extraction.
Chinese Translation
提示压缩通过缩短大型语言模型（LLM）的输入以降低推理成本，然而现有方法通过语言模型的前向传递来评估标记的重要性。这样的细致且成本高昂的标记选择是否必要仍然存在疑问。压缩需要识别信息内容，这是语言学研究长期以来通过可以操作化为确定性规则的线索所解决的问题。因此，我们提出了一个问题： extbf{仅靠语言规则}能否作为有效的提示压缩器，而无需在压缩时进行基于语言模型的评分？为了解决这个问题，我们对词汇、句法、语义和话语种子进行离线进化搜索，以寻找具有竞争力的规则组合。所得到的语言压缩器在部署时不需要语言模型的前向传递，仅使用CPU端处理进行压缩。我们通过双路径协议评估其在压缩质量和重建保真度之间的平衡。在短段落、多文档推理和对话记忆问答数据集上，进化压缩器的性能与近期先进的提示压缩策略相似。在轻度到中度压缩下，性能最强，随着压缩变得更加激进，性能下降，而直接路径和重建路径则表现出不同的模式。进化分析表明，有效的压缩在语言层面之间融合信号，并且随着压缩比的增加，规则从标记修剪转向句子提取。
cs.CL / 20 / 2607.25337
Temporal-Distance JEPA: Plan-Aware Representation Learning for Latent World Model Predictive Control
时间距离JEPA：面向计划的潜在世界模型预测控制的表征学习
Abstract
Joint-Embedding Predictive Architectures (JEPAs) learn world models by predicting in representation space rather than reconstructing pixels, making them a natural backbone for latent model predictive control from offline demonstration logs. JEPA-style training optimizes short-horizon latent prediction, whereas planning requires a multi-step ranking of imagined futures by goal progress. Prior JEPA planners often inherit that ranking from embedding geometry, typically latent Euclidean distance, which arises as a byproduct of representation learning rather than as a progress cost mined from the logs. We propose temporal-distance JEPA (TD-JEPA), which retains the LeWM encoder--predictor backbone and mines a directed temporal cost from reward-free trajectories: same-trajectory step order supplies positive targets, cross-trajectory pairs act as heuristic negatives, and a rollout-consistency term matches the planner horizon. The mined supervision serves two roles: as the deployed planning cost when progress is topological, and as a representation signal that improves Euclidean planning when contact geometry dominates. Under locked evaluation, deploying the mined cost raises Two-Room success to 100.0% versus LeWM's 97.4%, while shared Euclidean planning on the same temporally trained checkpoint raises OGB-Cube by 14.2 points over LeWM and improves Push-T. Against LeWM and the concurrent RC-aux baseline under locked evaluation, TD-JEPA matches or exceeds both methods on every environment. Ablations show that the directed head, cross-trajectory negatives, and rollout consistency each contribute. TD-JEPA narrows the train--plan gap for JEPA world-model planners by discovering temporal progress structure in offline logs and co-designing cost form with plan-time deployment. Code is available at https://github.com/HKBU-KnowComp/TD-JEPA.
Chinese Translation
联合嵌入预测架构（JEPA）通过在表征空间中进行预测而非重构像素来学习世界模型，使其成为从离线演示日志中进行潜在模型预测控制的自然基础。JEPA风格的训练优化短期潜在预测，而规划则需要通过目标进展对想象的未来进行多步排序。之前的JEPA规划器通常从嵌入几何中继承这种排序，通常是潜在的欧几里得距离，这是一种表征学习的副产品，而不是从日志中挖掘的进展成本。我们提出了时间距离JEPA（TD-JEPA），它保留了LeWM编码器-预测器骨干，并从无奖励轨迹中挖掘出有向时间成本：相同轨迹的步骤顺序提供正目标，跨轨迹对作为启发式负样本，而回滚一致性项与规划者的时间范围相匹配。挖掘的监督承担了两个角色：当进展是拓扑时作为部署的规划成本，以及当接触几何主导时作为改善欧几里得规划的表征信号。在锁定评估下，部署挖掘的成本使Two-Room的成功率提高至100.0%，而LeWM为97.4%；在同一时间训练的检查点上共享的欧几里得规划使OGB-Cube提高了14.2分，改善了Push-T。在锁定评估中，TD-JEPA在每个环境中与LeWM和同时进行的RC-aux基线相匹配或超越。消融实验表明，有向头、跨轨迹负样本和回滚一致性各自都有贡献。TD-JEPA通过在离线日志中发现时间进展结构并与计划时间部署共同设计成本形式，缩小了JEPA世界模型规划器的训练-规划差距。代码可在https://github.com/HKBU-KnowComp/TD-JEPA获取。
cs.CL / 21 / 2607.25375
Inspect India Evals: An Open Benchmarking Framework for Evaluating Large Language Models in the Indian Linguistic and Cultural Context
Inspect India Evals：一个用于评估大型语言模型在印度语言和文化背景下的开放基准框架
Abstract
India is a vast nation of over 1.4 billion people, varied by hundreds of diverse and locally specific traditions and cultures and 22 officially recognized languages. Large language models (LLMs) are now being deployed on a massive scale throughout the mainland as well as in remote villages. However, the common benchmarks - MMLU, BIG-Bench, and TruthfulQA are almost exclusively English- and Western-centric. They do not identify those safety, fairness, and accuracy failures unique to the Indian context. That is the gap Inspect India Evals seeks to fill. It is an open-source framework built on top of UK AISI's Inspect AI platform. It has six benchmarks: Multilingual MMLU across sixteen Indian languages, BharatBBQ (our adaptation of BBQ for Indian social bias), a safety evaluation for Digital Public Infrastructure, a multilingual safety test using harmful prompts in Indian languages, a multi-turn jailbreak resistance test, and an Indian cultural knowledge benchmark scored using LLM-as-judge rubrics. In this study, we tested five open-weight models ranging from 8B to 32B parameters. Sarvam-M 24B and Gemma 2 27B came out on top, both scoring 80% on the composite India Fairness Index, with Sarvam-M even beating larger 32B models on Indian cultural knowledge and DPI safety compliance. All models scored 100% refusal on Multilingual Safety, whereas DPI safety varied from 20% to 100%. The framework is public. It's built to work with the UK AISI registry. Anyone can reproduce or extend this work.
Chinese Translation
印度是一个拥有超过14亿人口的广阔国家，拥有数百种多样且地方特有的传统和文化，以及22种官方认可的语言。大型语言模型（LLMs）现在正在整个大陆以及偏远村庄大规模部署。然而，常见的基准测试——MMLU、BIG-Bench和TruthfulQA几乎完全以英语和西方为中心，未能识别出在印度背景下独特的安全性、公平性和准确性缺陷。这正是Inspect India Evals试图填补的空白。该框架是一个开源框架，建立在英国AISI的Inspect AI平台之上。它包含六个基准：跨十六种印度语言的多语言MMLU、BharatBBQ（我们对BBQ在印度社会偏见的适应）、数字公共基础设施的安全评估、使用印度语言的有害提示进行的多语言安全测试、多轮越狱抵抗测试，以及使用LLM作为评判标准的印度文化知识基准。在本研究中，我们测试了五个开放权重模型，参数范围从8B到32B。Sarvam-M 24B和Gemma 2 27B表现最佳，均在综合印度公平指数上得分80%，而Sarvam-M在印度文化知识和DPI安全合规性方面甚至超过了更大的32B模型。所有模型在多语言安全测试中均得分100%的拒绝率，而DPI安全性则在20%到100%之间波动。该框架是公开的，旨在与英国AISI注册处协同工作。任何人都可以复制或扩展这项工作。
cs.CL / 22 / 2607.25380
Memory for Large Language Models
大型语言模型的记忆
Abstract
Memory has evolved into a foundational architectural dimension in large language models (LLMs), shifting from an implicit byproduct of computation to a spectrum of explicit, controllable mechanisms. While recent advances introduce diverse strategies---spanning transient attention, recurrent state dynamics, parameter-efficient adaptations, and scalable lookup storage---this rapid evolution has led to a highly fragmented research landscape. In this survey, we present a systematic, architecture-centric taxonomy of memory in LLMs. Our framework characterizes memory along three orthogonal axes: representation (implicit versus explicit), update dynamics (offline versus online), and persistence (short-term versus long-term). We further formalize the granular mechanisms dictating memory writing, routing, state transitions, and consolidation. This unified perspective elucidates the conceptual boundaries between computation-coupled and independently addressable memory, effectively bridging disparate architectural paradigms. Additionally, we critically analyze hybrid memory architectures, system-level efficiency trade-offs, and multi-dimensional evaluation methodologies. By consolidating these scattered advancements into a cohesive framework, this survey charts the trajectory of memory-centric LLM design and provides a principled foundation for future innovations in scalable and adaptive language modeling.
Chinese Translation
记忆已经发展成为大型语言模型（LLMs）中的一个基础架构维度，从计算的隐性副产品转变为一系列显性、可控的机制。尽管最近的进展引入了多样化的策略——涵盖瞬态注意力、递归状态动态、参数高效的适应和可扩展的查找存储——但这种快速演变导致了高度分散的研究格局。在本次调查中，我们提出了一个系统的、以架构为中心的LLMs记忆分类法。我们的框架沿着三个正交轴对记忆进行特征描述：表示（隐性与显性）、更新动态（离线与在线）和持久性（短期与长期）。我们进一步形式化了决定记忆写入、路由、状态转换和整合的细粒度机制。这一统一视角阐明了计算耦合记忆与独立可寻址记忆之间的概念边界，有效地桥接了不同的架构范式。此外，我们还对混合记忆架构、系统级效率权衡和多维评估方法进行了批判性分析。通过将这些分散的进展整合为一个连贯的框架，本次调查描绘了以记忆为中心的LLM设计的轨迹，并为未来在可扩展和自适应语言建模方面的创新提供了原则性的基础。
cs.CL / 23 / 2607.25507
Phase Structure in Rotary Attention: A Spectral Framework for Semantic Continuity and Execution-Boundary Governance
旋转注意力中的相位结构：语义连续性与执行边界治理的谱框架
Abstract
Transformer language models are usually analyzed through vector geometry, yet ordered context and rotary position encoding introduce explicit phase structure into query-key interactions. This paper develops a bounded spectral framework for examining rotary phase alignment, hidden-state continuity, and semantic drift without treating language models as literal physical wave systems. It first identifies ordered hidden-state sequences, rather than vocabulary indices, as valid domains for spectral decomposition. It then derives the Rotary Position Embedding (RoPE) attention score as a sum of magnitude-weighted cosine terms and proves a local stability lemma: uniformly bounded phase displacement limits degradation of the corresponding pre-softmax score. To extend phase analysis beyond native RoPE coordinates, the paper defines complex modal coordinates over fixed orthonormal direction pairs and introduces a weighted coherence functional for hidden-state trajectories. These constructions support a strict distinction between representational continuity and execution-boundary admissibility. Internal coherence may describe preservation of task-relevant relations, but it cannot authorize a consequential transition. Positioned against existing geometric, spectral, phase-modulation, representation-analysis, and mechanistic-interpretability accounts, the framework contributes a theoretical and methodological program for determining when spectral structure explains continuity and when governance must remain an external predicate over execution.
Chinese Translation
变换器语言模型通常通过向量几何进行分析，但有序上下文和旋转位置编码在查询-键交互中引入了明确的相位结构。本文发展了一个有界谱框架，用于研究旋转相位对齐、隐藏状态连续性和语义漂移，而不将语言模型视为字面上的物理波动系统。首先，本文识别有序的隐藏状态序列，而不是词汇索引，作为谱分解的有效领域。然后，推导出旋转位置嵌入（Rotary Position Embedding, RoPE）注意力得分，作为幅度加权余弦项的总和，并证明了一个局部稳定性引理：均匀有界的相位位移限制了相应的预-softmax得分的降解。为了将相位分析扩展到原生RoPE坐标之外，本文在固定的正交方向对上定义了复模态坐标，并引入了一个加权一致性泛函用于隐藏状态轨迹。这些构造支持表示连续性与执行边界可接受性之间的严格区分。内部一致性可能描述任务相关关系的保持，但它不能授权一个有意义的转变。与现有的几何、谱、相位调制、表示分析和机械可解释性理论相对比，该框架为确定何时谱结构解释连续性以及何时治理必须保持为执行的外部谓词提供了理论和方法论的程序。
cs.CL / 24 / 2607.25579
IRIS: Reusable Identity Representations from Frozen LLMs for Entity Alignment
IRIS：来自冻结大型语言模型的可重用身份表示用于实体对齐
Abstract
Entity alignment (EA) identifies entities across knowledge graphs (KGs) that refer to the same real-world object. Conventional EA methods mainly exploit explicit graph structures and textual fields, which often provide insufficient semantic understanding to recognize the same entity under heterogeneous descriptions and distinguish it from semantically similar entities. Although large language models (LLMs) offer deeper entity understanding, existing LLM-based EA methods largely use this capability for auxiliary generation or candidate-conditioned decisions. Consequently, such understanding is not distilled into a stable and directly comparable identity space, leaving alignment tied to specific KG pairs or candidate sets and requiring repeated processing as the matching context changes. To address these limitations, we propose IRIS (Identity Representations from Internal States), a training-free framework that constructs for each entity an iris-like signature encoding its distinctive and stable identity characteristics. IRIS derives these signatures by eliciting identity-oriented contextual representations from a frozen LLM, thereby forming a shared space in which each entity is encoded once and can be aligned across different KGs through direct similarity comparison, without pair-dependent representation construction or candidate-wise LLM inference. Across four established EA benchmarks and two frozen LLM backbones, the best IRIS variants achieve Hits@1 scores of 100.00, 99.38, 98.31, and 97.99 on D-Y-15K V2, DBP-WIKI, ICEWS-WIKI, and ICEWS-YAGO, respectively.
Chinese Translation
实体对齐（EA）旨在识别知识图谱（KGs）中指代同一现实世界对象的实体。传统的EA方法主要利用显式的图结构和文本字段，这往往无法提供足够的语义理解，以识别在异构描述下的相同实体，并将其与语义上相似的实体区分开来。尽管大型语言模型（LLMs）提供了更深层次的实体理解，但现有的基于LLM的EA方法主要将这种能力用于辅助生成或候选条件决策。因此，这种理解并未提炼成一个稳定且可直接比较的身份空间，使得对齐依赖于特定的KG对或候选集，并且随着匹配上下文的变化需要重复处理。为了解决这些局限性，我们提出了IRIS（来自内部状态的身份表示），这是一个无训练的框架，为每个实体构建一个类似虹膜的签名，编码其独特且稳定的身份特征。IRIS通过从冻结的LLM中引出以身份为导向的上下文表示来推导这些签名，从而形成一个共享空间，在该空间中每个实体仅被编码一次，并且可以通过直接相似性比较在不同KGs之间进行对齐，而无需依赖于配对的表示构建或候选的LLM推理。在四个已建立的EA基准和两个冻结的LLM骨干网络上，最佳的IRIS变体在D-Y-15K V2、DBP-WIKI、ICEWS-WIKI和ICEWS-YAGO上分别达到了100.00、99.38、98.31和97.99的Hits@1分数。
cs.CL / 25 / 2607.25581
Evaluation of forced alignment of code-mixed speech: the case of Hindi-English
混合语言语音的强制对齐评估：以印地语-英语为例
Abstract
Code-mixed speech poses unique challenges to forced alignment: expanded inventories, orthographic errors, and speaker variation. We evaluate forced alignment of Hindi-English code-mixed speech using the Montreal Forced Aligner. We address 2 problems: (1) free variation involving native vs non-native pairs and (2) phonemic boundary detection for mid-utterance English words. Bootstrapping strategies substantially outperform unmodified lexicons. Acoustic models trained on sentence-level code-mixed data achieve a mean error of 4.15ms, ie. ten times lower than monolingual Hindi (38.18ms) or isolated English (37.58ms) alternatives. Principled lexicon design and code-mixed training data are both essential for reliable alignment of bilingual speech.
Chinese Translation
混合语言语音对强制对齐提出了独特的挑战：扩展的音素库、拼写错误和说话者变异。我们使用蒙特利尔强制对齐工具评估印地语-英语混合语言语音的强制对齐。我们解决了两个问题：（1）涉及母语与非母语对的自由变异，以及（2）中间话语中英语单词的音位边界检测。引导策略显著优于未修改的词典。基于句子级混合语言数据训练的声学模型实现了平均误差为4.15毫秒，即比单语印地语（38.18毫秒）或孤立英语（37.58毫秒）替代方案低十倍。原则性的词典设计和混合语言训练数据对于双语语音的可靠对齐都是至关重要的。
cs.CL / 26 / 2607.25590
PILA: Plug-and-Play Insertion for LLM-native Advertising
PILA：即插即用的LLM原生广告插入
Abstract
How to monetize large language models (LLMs) by naturally integrating sponsored content into their responses, known as LLM-native advertising, has recently emerged as a critical problem. However, existing solutions entangle advertising with content generation inside a single model, which is incompatible with modern API-only or workflow-based LLM applications and inevitably compromises the original response quality. To address this, we propose PILA, which reformulates ad insertion as a conditional response rewriting problem and decouples it from the upstream service as a lightweight sidecar module. PILA is model-agnostic and can be seamlessly integrated with existing LLM services without modifying the base model or its workflow. It also exposes a controllable trade-off between user-side naturalness and ad-side exposure, offering a practical interface for downstream pricing and deployment. Experiments across diverse upstream models show that \pila consistently improves ad effectiveness while preserving response quality, highlighting its promise as a practical solution for LLM-native advertising.
Chinese Translation
如何通过自然地将赞助内容融入大型语言模型（LLMs）的响应中来实现变现，即LLM原生广告，最近已成为一个关键问题。然而，现有解决方案将广告与内容生成纠缠在单一模型中，这与现代仅基于API或工作流的LLM应用不兼容，并不可避免地影响原始响应的质量。为了解决这个问题，我们提出了PILA，它将广告插入重新表述为条件响应重写问题，并将其与上游服务解耦，作为一个轻量级的侧车模块。PILA是模型无关的，可以与现有的LLM服务无缝集成，而无需修改基础模型或其工作流。它还提供了用户端自然性与广告端曝光之间的可控权衡，为下游定价和部署提供了实用接口。针对多种上游模型的实验表明，PILA始终提高了广告效果，同时保持响应质量，突显了其作为LLM原生广告实用解决方案的潜力。
cs.CL / 27 / 2607.25630
A Human-in-the-Loop Corpus for LLM-Based Simplification of Scientific Summaries
基于人类参与的语料库用于大型语言模型（LLM）驱动的科学摘要简化
Abstract
Interdisciplinary research is accelerating, yet scientific papers remain difficult to understand outside their home fields. We study large language model (LLM)-based simplification of scientific texts and present a human-in-the-loop workflow that transforms expert summaries into more accessible versions for non-specialists. Using SciSummNet as the source corpus, we first generate baseline simplifications with GPT-4o-mini. In Phase 1, readers from STEM fields outside computer science identify difficult sentences and phrases and compare the original and GPT-simplified summaries in terms of comprehensibility, naturalness, and simplicity. In Phase 2, computer science experts use this feedback to create expert-edited reference simplifications. We release the resulting corpus together with human judgments and automatic evaluation results. The Phase 1 judgments show a clear preference for the GPT-generated summaries in terms of comprehensibility and simplicity, while qualitative analysis of the Phase 2 edits highlights the importance of preserving domain-specific terminology and the strength of scientific claims. The resulting resource supports the training and benchmarking of simplification systems for cross-disciplinary scientific communication.
Chinese Translation
跨学科研究正在加速，但科学论文在其本领域之外仍然难以理解。我们研究了基于大型语言模型（LLM）的科学文本简化，并提出了一种人类参与的工作流程，将专家摘要转化为更易于非专业人士理解的版本。以SciSummNet作为源语料库，我们首先使用GPT-4o-mini生成基线简化版本。在第一阶段，来自计算机科学以外的STEM领域的读者识别困难的句子和短语，并在可理解性、自然性和简洁性方面比较原始摘要和GPT简化的摘要。在第二阶段，计算机科学专家利用这些反馈创建专家编辑的参考简化版本。我们发布了生成的语料库以及人类评判和自动评估结果。第一阶段的评判显示，在可理解性和简洁性方面，GPT生成的摘要明显更受欢迎，而第二阶段编辑的定性分析突出了保留领域特定术语和科学主张强度的重要性。该资源支持跨学科科学交流的简化系统的训练和基准测试。
cs.CL / 28 / 2607.25633
Construction-Driven Injection: Linguistically-Grounded Edit-Based Code-Mixing Fingerprints for Large Language Models
构建驱动的注入：基于语言学的编辑型代码混合指纹用于大型语言模型
Abstract
Large language models (LLMs) are costly intellectual assets that remain exposed to unauthorized redistribution and commercial misuse. Injected fingerprints, i.e., trigger--target pairs embedded in model behavior, offer a practical, black-box-verifiable ownership signal, but existing methods decouple the two stages of the fingerprint life cycle: how a fingerprint is constructed and how it is injected. Existing fingerprinting frameworks suffer from two limitations. Natural-language fingerprints are prone to accidental activation, and garbled fingerprints are easily filtered by perplexity-based detection. Furthermore, decoupling construction from injection leaves the latter unaware of the trigger's linguistic structure, missing the opportunity for targeted optimization. We argue that fingerprint construction should drive injection, and present a unified fingerprinting framework that jointly optimizes both stages. First, LCF constructs code-mixing fingerprints by combining low-resource languages under a semantic-density substitution rule and grammar-biased mixing, yielding triggers whose perplexity sits far below garbled baselines while avoiding the accidental-activation failures of natural-language triggers. Second, LCFEdit injects each fingerprint with a null-space projection derived from high-resource multilingual representations that preserves knowledge, augmented by a cross-lingual alignment step that steers the weight update toward the fingerprint language's representation subspace. This construction-aware injection ensures that the update is linguistically informed and therefore more stable. Extensive evaluations on imperceptibility, detectability, and harmlessness demonstrate persistent ownership verification with negligible impact on utility.
Chinese Translation
大型语言模型（LLMs）是昂贵的知识资产，但仍然面临未经授权的再分发和商业滥用的风险。注入的指纹，即嵌入模型行为中的触发器-目标对，提供了一种实用的、黑箱可验证的所有权信号，但现有方法将指纹生命周期的两个阶段解耦：指纹的构建和注入。现有的指纹框架存在两个局限性。自然语言指纹容易意外激活，而混乱的指纹则容易被基于困惑度的检测过滤。此外，将构建与注入解耦使得后者无法了解触发器的语言结构，错失了针对性优化的机会。我们认为指纹构建应驱动注入，并提出一个统一的指纹框架，联合优化这两个阶段。首先，LCF通过在语义密度替代规则和语法偏向混合下结合低资源语言来构建代码混合指纹，从而生成困惑度远低于混乱基线的触发器，同时避免自然语言触发器的意外激活失败。其次，LCFEdit通过从高资源多语言表示中导出的零空间投影注入每个指纹，保留知识，并通过跨语言对齐步骤增强，指导权重更新朝向指纹语言的表示子空间。这种关注构建的注入确保了更新在语言学上是知情的，因此更加稳定。对不可感知性、可检测性和无害性的广泛评估表明，所有权验证持续有效，对效用的影响微乎其微。
cs.CL / 29 / 2607.25667
MyMentorLLM: A psychotherapy GenAI environment with multimodal voice/text patients, trainees and experts for deliberate practice
MyMentorLLM：一个具有多模态语音/文本患者、培训者和专家的心理治疗生成AI环境，用于有意识的练习
Abstract
Psychotherapists need repeated training and supervision by experts; however, scalability is problematic. Here we present MyMentorLLM, a multimodal voice- and text-based simulation environment for deliberate practice, used to generate 2,100 complete Cognitive Behavioural Therapy (CBT) training sessions. Each session links a DSM-5-TR-grounded patient (with major depressive, generalised anxiety or borderline personality disorder), a therapist-in-training and an expert supervisor. As an initial implementation, we adopted CBT because its structured procedures and competency-based supervision facilitate standardised simulation and evaluation. Sessions were analysed for emotional dynamics, therapeutic competence and diagnostic accuracy. Simulated patients expressed disorder-congruent emotional profiles, which trainee therapists mirrored as in real human counselling. The quality of supervision differed across LLMs: while most models overestimated trainees' competences, native speech-to-speech was closest to human scores. Supervisors' feedback led to better diagnoses in simulated psychotherapists in 5 out of 7 LLMs, and symptom identification accuracy increased with model size. This work shows that simulation of deliberate practice is possible for CBT training, although patient fidelity, calibration of supervisors, and harmful feedback should be evaluated together.
Chinese Translation
心理治疗师需要专家的重复培训和监督；然而，规模化存在问题。在此，我们呈现MyMentorLLM，一个基于多模态语音和文本的模拟环境，用于有意识的练习，生成2100个完整的认知行为疗法（CBT）培训课程。每个课程链接了一位基于《精神疾病诊断与统计手册》第五版（DSM-5-TR）的患者（患有重度抑郁、广泛性焦虑或边缘性人格障碍）、一名培训中的治疗师和一名专家监督者。作为初步实施，我们采用了CBT，因为其结构化程序和基于能力的监督促进了标准化的模拟和评估。课程分析了情感动态、治疗能力和诊断准确性。模拟患者表现出与其障碍相符的情感特征，培训治疗师在真实人类咨询中反映了这一点。不同的LLM（大语言模型）之间监督质量存在差异：虽然大多数模型高估了培训者的能力，但原生的语音到语音模型与人类评分最接近。监督者的反馈使得模拟心理治疗师在7个LLM中的5个中得到了更好的诊断，且随着模型规模的增大，症状识别的准确性也有所提高。这项工作表明，CBT培训的有意识练习模拟是可能的，尽管患者的真实性、监督者的校准和有害反馈应当一起评估。
cs.CL / 30 / 2607.25716
SpeechLLM Meets Federated Learning for End-to-End ASR: English and Italian Case Studies
SpeechLLM与联邦学习相结合的端到端自动语音识别：英语和意大利语案例研究
Abstract
Federated learning (FL) enables privacy-preserving training of automatic speech recognition (ASR) systems across distributed data sources, yet its application to large-scale speech language models (SpeechLLMs) remains unexplored. This paper presents the first systematic study of federated training for SpeechLLM-based end-to-end ASR systems. We design a communication-efficient federated optimization strategy tailored to the unique challenges of SpeechLLM architectures, addressing high-dimensional parameter spaces, gradient communication overhead, and computational constraints in distributed settings. Through extensive empirical evaluation on monolingual ASR tasks in English and Italian, we demonstrate the effectiveness and stability of our federated approach compared to centralized training baselines across diverse acoustic conditions and speaking styles. Additionally, we conduct a comprehensive ablation study analyzing the impact of different speech encoder architectures on monolingual English ASR performance within the federated framework, providing insights into optimal model configurations for decentralized training. Our results achieve competitive word error rates while reducing communication costs, establishing practical foundations for federated SpeechLLM deployment in real-world multilingual scenarios.
Chinese Translation
联邦学习（FL）使得在分布式数据源上进行隐私保护的自动语音识别（ASR）系统训练成为可能，但其在大规模语音语言模型（SpeechLLMs）中的应用尚未得到探索。本文首次系统性地研究了基于SpeechLLM的端到端ASR系统的联邦训练。我们设计了一种针对SpeechLLM架构独特挑战的高效通信联邦优化策略，解决了高维参数空间、梯度通信开销以及分布式环境中的计算限制。通过对英语和意大利语的单语ASR任务进行广泛的实证评估，我们展示了与集中式训练基线相比，我们的联邦方法在不同声学条件和说话风格下的有效性和稳定性。此外，我们还进行了全面的消融研究，分析了不同语音编码器架构在联邦框架内对单语英语ASR性能的影响，为去中心化训练的最佳模型配置提供了见解。我们的结果在降低通信成本的同时实现了具有竞争力的词错误率，为在现实世界多语言场景中部署联邦SpeechLLM奠定了实用基础。
cs.CL / 31 / 2607.25765
WorkSurface-Bench: Benchmarking Enterprise Agents on Multi-Surface Knowledge Routing
WorkSurface-Bench：多表面知识路由中企业代理的基准测试
Abstract
Enterprise agents often need to integrate heterogeneous knowledge sources: documents for narrative facts, tables for computation, and dependency graphs for file relationships. Existing benchmarks typically evaluate retrieval or tool use without distinguishing whether an agent first selects the appropriate knowledge sources. We introduce WorkSurface-Bench, a benchmark for evaluating this capability as surface routing. It contains 1,151 atomic tasks derived from persona-scoped Workspace-Bench-Lite workspaces, spanning document, table, graph, and cross-surface questions. Its reference answers are auditable: table answers are reproduced through executed DuckDB queries, document answers are grounded in verified text spans, and graph answers are traced to source dependency annotations. We evaluate four model backbones across six controlled agent settings, yielding 27,624 protocol-error-free trajectories. Under gold-constrained tool access, agents achieve 98.7-99.8 Route F1, while Answer remains only 56.1-75.3 percent, showing that correct surface selection is necessary but insufficient for task completion. Matched interventions further show that surface hints improve Answer for three of four models, whereas removing irrelevant tools primarily improves routing and efficiency. In an independent three-annotator audit, all 200 sampled tasks pass all six quality criteria by majority vote, with 192 receiving unanimous judgments on every criterion. We release the dataset, construction pipeline, scoring code, and agent harness at https://github.com/haolpku/WorkSurface-Bench.
Chinese Translation
企业代理通常需要整合异构知识源：用于叙述事实的文档、用于计算的表格以及用于文件关系的依赖图。现有的基准测试通常评估检索或工具使用，而不区分代理是否首先选择了适当的知识源。我们引入了WorkSurface-Bench，这是一个用于评估这一能力（即表面路由）的基准测试。它包含1,151个原子任务，这些任务源自以角色为范围的Workspace-Bench-Lite工作空间，涵盖文档、表格、图形和跨表面问题。其参考答案是可审计的：表格答案通过执行DuckDB查询重现，文档答案基于经过验证的文本片段，图形答案追溯到源依赖注释。我们在六个受控代理设置中评估了四个模型骨干，产生了27,624条无协议错误的轨迹。在黄金约束工具访问下，代理的Route F1达到98.7-99.8，而Answer仅为56.1-75.3，显示出正确的表面选择是任务完成所必需的，但不足以保证完成。匹配的干预进一步表明，表面提示提高了四个模型中三个的Answer，而移除无关工具主要改善了路由和效率。在一个独立的三位注释者审计中，所有200个抽样任务通过了六项质量标准的多数投票，192个任务在每个标准上获得了一致的判断。我们在https://github.com/haolpku/WorkSurface-Bench发布了数据集、构建管道、评分代码和代理工具。
cs.CL / 32 / 2607.25814
Evaluation of Adversarial Robustness in Arabic Language Models
阿拉伯语模型的对抗鲁棒性评估
Abstract
The emergence of the recent outstanding capabilities of Arabic Language Models has opened doors for exposing their vulnerabilities. One of the major security risks associated with such Natural Language Processing models is adversarial attacks. These attacks can deceive the model into the wrong prediction, raising critical model security and safety concerns. This study aims to assess the robustness of five state-of-the-art Arabic Language Models under a distinct set of Arabic adversarial attacks applied at various levels of granularity and using different example generation strategies. We also explore a defense technique based on adversarial training to enhance model robustness. The results show that insertion of diacritics can reduce the accuracy of some models by 92% while maintaining a low perturbation distance. For word-level attacks, manipulating Arabic conjunctions preserves high semantic similarity scores, low perturbation distance, and leads to an accuracy degradation of up to 58%. For sentence-level attacks, paraphrasing proves its effectiveness by an average reduction of 76% in the victim models' performance. While adversarial training improves overall resilience, with MARBERT being the most robust and AraBERT showing the greatest relative gains, challenges persist, particularly against character-level noise. These findings highlight both the potential and limitations of current defense strategies in morphologically rich languages like Arabic.
Chinese Translation
阿拉伯语模型近期卓越能力的出现揭示了其脆弱性。与此类自然语言处理模型相关的主要安全风险之一是对抗攻击。这些攻击可能会误导模型做出错误预测，从而引发严重的模型安全和安全性问题。本研究旨在评估五种最先进的阿拉伯语模型在不同层次的阿拉伯对抗攻击下的鲁棒性，这些攻击采用了不同的示例生成策略。我们还探索了一种基于对抗训练的防御技术，以增强模型的鲁棒性。结果表明，插入元音符号可以使某些模型的准确率降低92%，同时保持较低的扰动距离。对于词级攻击，操控阿拉伯连词能够保持高语义相似度分数和低扰动距离，并导致准确率下降高达58%。对于句子级攻击，改写证明其有效性，受害模型的性能平均降低76%。虽然对抗训练提高了整体韧性，其中MARBERT表现出最强的鲁棒性，而AraBERT则显示出最大的相对增益，但在字符级噪声方面仍然存在挑战。这些发现突显了在形态丰富的语言（如阿拉伯语）中当前防御策略的潜力和局限性。
cs.CL / 33 / 2607.25852
AngelSpec: Towards Real-World High Performance Inference with Speculative Decoding
AngelSpec：朝着现实世界高性能推理的投机解码
Abstract
Speculative decoding accelerates large language model inference without changing the target distribution, but no single drafting structure performs best across real-world workloads. Autoregressive multi-token prediction (MTP) is a lightweight, stable proposal mechanism, whereas block-parallel diffusion amortizes drafting latency over much longer candidate sequences; the better choice depends strongly on the output distribution. We present AngelSpec, a unified training framework for MTP and block-parallel speculative decoding that addresses this heterogeneity at three levels. At the training level, rather than fitting one universal drafter to a uniform data mixture, we co-specialize structure and data: the MTP drafter is trained on diverse conversational data for high-entropy open-ended chat, and the block-diffusion drafter on code and mathematics data for longer predictable continuations. At the architecture level, we propose DFly, a block-diffusion framework combining a hybrid target-conditioning backbone with a predecessor-conditioned autoregressive head, improving target-feature utilization and intra-block dependency modeling while keeping generation parallel. At the inference level, both acceptance length and verification cost vary with domain, request, online load, and hardware, so DFly treats verification as a shared batch-level resource: it reallocates compute toward high-confidence prefixes across requests and combines expected utility with a profiled cost model to adapt verification depth online. Across the Hy3 series, DFly raises the average accepted length on Hy3-A21B by roughly 30% and attains the highest average throughput at every tested concurrency from 4 to 64, a 1.98-2.40x speedup over autoregressive decoding and 10.5-11.8% higher throughput than DFlash. We release AngelSpec to support training and extending these methods.
Chinese Translation
投机解码在不改变目标分布的情况下加速大型语言模型的推理，但没有单一的草拟结构能够在现实世界的工作负载中表现最佳。自回归多标记预测（MTP）是一种轻量级、稳定的提议机制，而块并行扩散则在更长的候选序列上摊销草拟延迟；更好的选择在很大程度上取决于输出分布。我们提出了AngelSpec，这是一个统一的MTP和块并行投机解码的训练框架，解决了这一异质性问题，分为三个层面。在训练层面，我们并不是将一个通用的草拟器拟合到统一的数据混合中，而是共同专业化结构和数据：MTP草拟器在多样的对话数据上进行训练，以应对高熵的开放式聊天，而块扩散草拟器则在代码和数学数据上进行训练，以实现更长的可预测延续。在架构层面，我们提出了DFly，一个块扩散框架，结合了混合目标条件的主干和前驱条件的自回归头，改善了目标特征的利用率和块内依赖建模，同时保持生成的并行性。在推理层面，接受长度和验证成本因领域、请求、在线负载和硬件而异，因此DFly将验证视为共享的批量级资源：它在请求之间重新分配计算资源，以高置信度前缀为中心，并结合期望效用与配置成本模型在线调整验证深度。在Hy3系列中，DFly将Hy3-A21B的平均接受长度提高了约30%，并在从4到64的每个测试并发下达到了最高的平均吞吐量，相较于自回归解码实现了1.98-2.40倍的加速，并比DFlash高出10.5-11.8%的吞吐量。我们发布AngelSpec以支持这些方法的训练和扩展。
cs.CL / 34 / 2607.25857
Shieldstral
Shieldstral
Abstract
We introduce Shieldstral, a 3B-parameter policy-adaptive multimodal safety classifier that matches or outperforms models nearly 7$\times$ its size on text safety benchmarks and sets a new state of the art on multimodal safety classification. Shieldstral formulates content moderation as a binary question-answering task. This simple formulation unifies diverse moderation tasks into a single yes/no problem, enabling heterogeneous safety datasets with divergent taxonomies to be consolidated under one training framework. We present the data construction recipe, covering curation and generation of approximately 54.1M samples and a fine-grained evaluation set to evaluate policy adaptability. Together, these enable a small adaptive model to match or outperform much larger models.
Chinese Translation
我们介绍了Shieldstral，这是一种具有30亿参数的政策自适应多模态安全分类器，在文本安全基准测试中，其性能与接近其规模7倍的模型相当或更优，并在多模态安全分类上设定了新的最先进水平。Shieldstral将内容审核形式化为一个二元问答任务。这种简单的形式化将多样的审核任务统一为一个是/否问题，使得具有不同分类法的异构安全数据集能够在一个训练框架下整合。我们展示了数据构建的配方，涵盖了约5410万样本的策划和生成，以及用于评估政策适应性的细粒度评估集。这些共同使得一个小型自适应模型能够与更大模型的性能相匹配或超越。
cs.CL / 35 / 2607.25881
AI's Capability in Assisting Scientific Research in Physics, Astrophysics, and Cosmology II: Project Planning and Proposal Evaluation
人工智能在物理学、天体物理学和宇宙学中辅助科学研究的能力 II：项目规划与提案评估
Liu, Jia, Krishnaraj, Veena, Vovk, Kateryna, Aizawa, Kosuke, Bayer, Adrian E., Blot, Linda, Cowell, Jessica, Garg, Suyog, Grée, Jonathan, Hell, Anamaria, Horowitz, Ben, Ichikawa, Masaya, Iemoto, Kanyuni, Kondo, Keigo, Lorsin, Zacharie, McCarthy, Kevin, Robinson, Jamie, Ruiz-Granda, Miguel, Thiele, Leander, Vovk, Ievgen, Zhou, Mingshen
Abstract
We investigate how well large language models (LLMs) can assist scientific project planning and proposal evaluation. One-page project plans were independently generated for eight expert-conceived research projects in physics, astrophysics, and cosmology by human researchers and three contemporary LLMs (ChatGPT, Claude, and DeepSeek; mid-2025 models, used with their default tool access). The resulting 32 proposals were blindly evaluated by four human reviewers and two newer frontier LLMs (Claude Opus 4.8 and ChatGPT Pro 5.5) using a four-aspect evaluation rubric. Reviewers were also asked to identify whether each proposal was written by a human or an AI. Human reviewers rated human- and AI-written proposals similarly overall, whereas both AI reviewers scored AI-written proposals about one point higher (on a five-point scale) than human-written proposals. Human reviewers correctly identified human- and AI-written proposals 72% and 79% of the time, respectively, while both AI reviewers correctly classified all 32 proposals (100%). These results suggest that current LLMs can produce project plans comparable to human-written ones in the eyes of human reviewers, but that AI reviewers show a systematic preference for AI-generated proposals. Our results suggest caution when deploying LLMs widely in proposal preparation and evaluation.
Chinese Translation
我们研究了大型语言模型（LLMs）在科学项目规划和提案评估中的辅助能力。人类研究人员和三种当代LLMs（ChatGPT、Claude和DeepSeek；2025年中期模型，使用其默认工具访问）独立生成了八个专家构思的物理学、天体物理学和宇宙学研究项目的一页项目计划。最终生成的32个提案由四位人类评审员和两种新兴前沿LLMs（Claude Opus 4.8和ChatGPT Pro 5.5）使用四个方面的评估标准进行盲评。评审员还被要求识别每个提案是由人类还是人工智能撰写的。人类评审员对人类和AI撰写的提案的总体评分相似，而两位AI评审员则对AI撰写的提案的评分比人类撰写的提案高出约一分（满分五分）。人类评审员正确识别出人类和AI撰写的提案的比例分别为72%和79%，而两位AI评审员则正确分类了所有32个提案（100%）。这些结果表明，当前的LLMs在评审员眼中能够生成与人类撰写的项目计划相当的内容，但AI评审员对AI生成的提案表现出系统性的偏好。我们的结果提示在提案准备和评估中广泛使用LLMs时需谨慎。
cs.CL / 36 / 2607.25933
Evaluating Multi-Turn Multimodal Diagnostic Reasoning on Challenging Real-World Clinical Cases
评估多轮多模态诊断推理在具有挑战性的真实世界临床案例中的表现
Yang, Rui, Xuan, Weihao, Lin, Yi, Bao, Zhuhan, Liew, Jonathan Chong Kai, Wong, Matthew Yu Heng, Lescano, Nicolás, Paripati, Nikita R., Pai, Emily Ling-Lin, Liu, Jiarui, Qi, Heli, Chang, Heng-Jui, Loo, Benny Kai Guo, Li, Huitao, Yu, Kunyu, Wang, Yufan, Hong, Chuan, Lu, Shijian, Teodoro, Douglas, Yokoya, Naoto, Koppel, Ross, Diab, Mona, Xu, Hua, Bates, David W., Liu, Nan, Peng, Yifan
Abstract
Clinical diagnostic evaluation should not only assess whether models can provide correct diagnoses, but also reflect the realities of clinical practice, including progressive disclosure of multimodal information, dynamic updating of diagnostic hypotheses, and continuous refinement of clinical reasoning. However, existing evaluations of multimodal large language models (MLLMs) typically rely on single-turn or isolated tasks, making it difficult to fully capture the complexity of real-world clinical diagnosis. To bridge this gap, we developed ClinMM-Bench, the largest multi-turn multimodal clinical diagnostic evaluation benchmark to date. ClinMM-Bench contains 1,089 challenging real-world clinical cases and 3,760 medical images across eight specialties. We systematically evaluated 15 representative MLLMs using a two-level evaluation framework that assessed both diagnostic accuracy and diagnostic reasoning quality. Results showed that proprietary models achieved the highest overall diagnostic accuracy, but the proportion of completely correct diagnoses remained limited across all models. In terms of diagnostic reasoning quality, current models can identify plausible diagnostic directions but still have considerable limitations in generating reliable diagnostic reasoning. Error analysis further identified five representative failure modes: information synthesis failure, knowledge mapping error, perception error, premature closure, and visual hallucination.
Chinese Translation
临床诊断评估不仅应评估模型是否能够提供正确的诊断，还应反映临床实践的现实，包括多模态信息的逐步披露、诊断假设的动态更新以及临床推理的持续完善。然而，现有的多模态大型语言模型（MLLMs）评估通常依赖于单轮或孤立任务，难以充分捕捉真实世界临床诊断的复杂性。为填补这一空白，我们开发了ClinMM-Bench，这是迄今为止最大的多轮多模态临床诊断评估基准。ClinMM-Bench包含1,089个具有挑战性的真实世界临床案例和3,760幅来自八个专业的医学图像。我们使用一个两级评估框架系统地评估了15个具有代表性的MLLMs，该框架评估了诊断准确性和诊断推理质量。结果显示，专有模型在整体诊断准确性上表现最佳，但所有模型的完全正确诊断比例仍然有限。在诊断推理质量方面，当前模型能够识别出合理的诊断方向，但在生成可靠的诊断推理方面仍存在相当大的局限性。错误分析进一步识别出五种典型的失败模式：信息综合失败、知识映射错误、感知错误、过早闭合和视觉幻觉。
cs.CL / 37 / 2607.25953
Polistemics: Evaluating LLMs as Information Mediators in Politics & Elections
政治信息的中介：评估大型语言模型在政治与选举中的作用
Abstract
As LLMs increasingly mediate the political information citizens rely on, there is still no standardized way to assess whether they do so responsibly. We introduce Polistemics, a theory-grounded benchmark for evaluating LLMs as mediators of political information in elections. Prior work has treated this task as reproduction rather than mediation, leaving its epistemic dimensions and interaction with imperfect information unaddressed. We ground the evaluation in Epistemic Modesty, a normative standard derived from citizens' epistemic agency, and test it across controlled settings that vary informational properties such as clarity, noise, and consistency. Applying the benchmark to three state-of-the-art LLMs on the 2025 German and Dutch elections, we find that high aggregate scores mask systematic failures. Models mediate reliably under clear evidence but break down under absent, vague, or contradictory information, while flattening the intensity of political language. These failures are likely driven by party priors, influenced by party labels and output language. Reliable mediation appears achievable, but no model delivers it consistently.
Chinese Translation
随着大型语言模型（LLMs）在公民依赖的政治信息中扮演越来越重要的中介角色，目前仍缺乏一种标准化的方法来评估它们是否以负责任的方式进行中介。我们提出了Polistemics，这是一个基于理论的基准，用于评估LLMs在选举中作为政治信息中介的表现。以往的研究将这一任务视为再现而非中介，未能解决其认识论维度及与不完美信息的互动。我们将评估基于认识谦逊（Epistemic Modesty），这一规范标准源于公民的认识主体性，并在控制环境中测试，变化的信息特性包括清晰度、噪声和一致性。将该基准应用于2025年德国和荷兰选举的三种最先进的LLMs，我们发现高的综合得分掩盖了系统性失败。在清晰证据下，模型能够可靠地进行中介，但在缺失、模糊或矛盾的信息下则会崩溃，同时削弱政治语言的强度。这些失败可能是由党派先验驱动的，受到党派标签和输出语言的影响。可靠的中介似乎是可以实现的，但没有任何模型能够始终如一地提供这种能力。
cs.CL / 38 / 2607.25959
Detecting Knowledge Inconsistencies Across Text, Tables, and Knowledge Graphs
跨文本、表格和知识图谱的知识不一致性检测
Abstract
Wikipedia and Wikidata are widely used for information access, LLM pre-training, and retrieval-augmented generation. Their knowledge is deeply connected but scattered across text, tables, and knowledge graphs. This raises a practical question: when these modalities disagree, how can we detect and explain the conflict? We study this problem as \emph{modality-level inconsistency detection}. We first introduce a taxonomy of cross-modal knowledge inconsistencies, covering information granularity differences, direct conflicts, temporal changes, and KG incompleteness. We then present \textsc{Kontrast}, an automatic framework that uses Text-to-SPARQL and LLM reasoning to compare table-based answers with KG evidence and categorize the resulting inconsistencies. Experiments on various Table-QA datasets show that cross-modal inconsistencies are common and informative. They reveal not only true knowledge conflicts, but also missing KG structure and temporal mismatches while being limited by Text-to-SPARQL errors and noise. Our analysis shows that text, tables, and KGs can complement and correct one another through systematic comparison. \textsc{Kontrast} provides a practical tool for large-scale knowledge auditing and establishes a benchmark for future work on cross-modal knowledge consistency. Code and data are available at https://github.com/ECLADATTA/KONTRAST.
Chinese Translation
维基百科和维基数据广泛用于信息获取、LLM（大规模语言模型）预训练和检索增强生成。它们的知识深度关联但分散在文本、表格和知识图谱中。这引发了一个实际问题：当这些模态不一致时，我们如何检测并解释这种冲突？我们将这个问题研究为 extit{模态级不一致性检测}。首先，我们介绍了一种跨模态知识不一致性的分类法，涵盖信息粒度差异、直接冲突、时间变化和知识图谱的不完整性。然后，我们提出了 extsc{Kontrast}，一个自动化框架，利用文本到SPARQL（Text-to-SPARQL）和LLM推理，将基于表格的答案与知识图谱证据进行比较，并对结果不一致性进行分类。在各种表格问答（Table-QA）数据集上的实验表明，跨模态不一致性是常见且富有信息的。它们不仅揭示了真实的知识冲突，还显示了缺失的知识图谱结构和时间不匹配，同时受到文本到SPARQL错误和噪声的限制。我们的分析表明，文本、表格和知识图谱可以通过系统比较相互补充和纠正。 extsc{Kontrast}为大规模知识审计提供了一个实用工具，并为未来的跨模态知识一致性研究建立了基准。代码和数据可在 https://github.com/ECLADATTA/KONTRAST 获取。
cs.CL / 39 / 2607.26015
Instruction-Tuned Models Locally Reuse Human Syntax More Than Humans Do
指令调优模型在局部上比人类更频繁地重用人类语法
Abstract
Syntactic convergence (the tendency of speakers to adapt in language towards the grammatical profiles of their interlocutors) is a well-documented feature of human dialogue widely considered to operate below conscious awareness. Whether large language models exhibit analogous syntactic convergence toward human users relative to human baselines and across a broad range of syntactic constructions remains an open question. Using substitution-paradigm data in which model generations replace one speaker's turns in pre-existing human dialogues, this study measures turn-adjacent reuse of context-free grammar (CFG) rules across sixteen open-weight Llama and Gemma models (1B-70B, pretrained and instruction-tuned) at 1,901 matched positions per model. Every model showed greater CFG-rule overlap with the preceding human turn than with a sampled unrelated human prime, and in every model this actual-versus-random difference was larger for lower-frequency rules. Each instruction-tuned model also showed greater natural-output overlap with the actual prime than the human response it replaced, and all eight matched architecture pairs exhibited greater actual-prime overlap after instruction tuning. However, relative to pretrained variants, instruction-tuned outputs overlapped more with unrelated primes, showed a smaller actual-versus-random increment, and had lower conditional rule-reuse odds once target rule-set size was held constant. In exploratory analyses, each model exhibited greater mean lexical and semantic similarity to the preceding turn than the matched human responses did. Instruction-tuned models additionally produced responses with greater mean semantic similarity than their pretrained counterparts in all eight architecture pairs, whereas the lexical similarity results were more heterogeneous.
Chinese Translation
句法趋同（说话者在语言中向其对话者的语法特征适应的倾向）是人类对话中一个经过充分记录的特征，广泛认为其在意识下运作。大型语言模型是否相对于人类基准和广泛的句法结构表现出类似的句法趋同仍然是一个未解的问题。通过使用替代范式数据，其中模型生成替代了预先存在的人类对话中的一个说话者的发言，本研究测量了在每个模型的1,901个匹配位置上，十六个开放权重的Llama和Gemma模型（1B-70B，经过预训练和指令调优）在上下文无关文法（CFG）规则的相邻重用。每个模型显示出与前一个人类发言的CFG规则重叠程度大于与一个随机无关的人类提示的重叠，并且在每个模型中，这种实际与随机的差异在低频规则上更大。每个指令调优模型还显示出与实际提示的自然输出重叠程度大于其替代的人类响应，并且所有八对匹配的架构对在指令调优后都表现出更大的实际提示重叠。然而，相较于预训练变体，指令调优输出与无关提示的重叠更大，实际与随机的增量较小，并且在目标规则集大小保持不变的情况下，条件规则重用的几率较低。在探索性分析中，每个模型表现出与前一个发言的平均词汇和语义相似度高于匹配的人类响应。指令调优模型在所有八对架构中产生的响应的平均语义相似度也高于其预训练的对应物，而词汇相似度的结果则更为异质。
cs.CL / 40 / 2607.26017
UniMem: Complementary Episodic-to-Parametric Memory for Boundary-Agnostic Task Streams
UniMem：用于边界无关任务流的互补性情节到参数记忆
Abstract
Memory is essential for LLM agents to accumulate task experience and reuse task-specific execution strategies. However, real-world deployment over boundary-agnostic and evolving task streams exposes a fundamental stability-plasticity dilemma. External retrieval-based memory can rapidly absorb new evidence, but it often fails to internalize recurring execution patterns and incurs inference-time retrieval overhead. Parametric memory enables stable and efficient execution once learned, but typically relies on explicit task boundaries and fixed parameter budgets. Inspired by the human brain, which balances plasticity and stability through complementary episodic storage and gradual consolidation, we propose UniMem, a self-routing framework for autonomous memory management. UniMem uses learnable routing tokens as memory controllers, enabling adaptive coordination between complementary memory pathways: novel or sparse tasks are retained in an episodic buffer for retrieval-augmented execution, while recurring and reliable patterns are consolidated into expandable parametric memory. By decoupling task identification from task execution with routing tokens and parametric memory blocks, UniMem expands memory on demand without task labels during deployment or uncontrolled parameter growth. Experiments on long-horizon streaming task sequences show that UniMem consistently outperforms baselines while maintaining execution fidelity, achieving an average gain of 4.0 EM points across three backbone models.
Chinese Translation
记忆对于大型语言模型（LLM）代理积累任务经验和重用任务特定执行策略至关重要。然而，在边界无关和不断演变的任务流中进行实际部署时，暴露出一个基本的稳定性-可塑性困境。基于外部检索的记忆能够快速吸收新证据，但往往无法内化重复的执行模式，并且在推理时会产生检索开销。参数记忆一旦学习后能够实现稳定和高效的执行，但通常依赖于明确的任务边界和固定的参数预算。受到人脑的启发，人脑通过互补的情节存储和逐步巩固来平衡可塑性和稳定性，我们提出了UniMem，一个用于自主记忆管理的自路由框架。UniMem使用可学习的路由令牌作为记忆控制器，实现互补记忆通路之间的自适应协调：新颖或稀疏的任务保留在情节缓冲区中以进行检索增强执行，而重复且可靠的模式则被巩固到可扩展的参数记忆中。通过将任务识别与任务执行解耦，使用路由令牌和参数记忆块，UniMem在部署期间无需任务标签即可按需扩展记忆，避免了不受控的参数增长。在长时间流式任务序列上的实验表明，UniMem在保持执行保真度的同时，始终优于基线模型，在三种主干模型中实现了平均4.0 EM点的提升。
cs.CL / 41 / 2607.26057
Pass the Baton: Trajectory-Relayed On-Policy Distillation
传递接力棒：基于轨迹的在线策略蒸馏
Abstract
On-policy distillation (OPD) grounds token-level supervision in the student's own trajectory, yet suffers from prefix failure: once the student commits to a wrong reasoning direction, all subsequent generation builds on this deviation, producing misdirected continuations that elicit unreliable supervision and waste compute. We identify a teacher-student continuation asymmetry on failed prefixes, where the teacher tends to redirect while the student continues along the original direction, and convert it into a label-free handoff trigger in Relay On-Policy Distillation (Relay-OPD). During training, Relay-OPD constructs relay trajectories by letting the teacher briefly take over at detected trigger points to produce a teacher leg, after which the student resumes and is optimized on the resulting trajectory. A limited relay budget concentrates intervention on critical early positions while limiting departure from the student policy. With a Qwen3-4B-Instruct-2507 teacher and Qwen3-0.6B/1.7B-Non-Thinking students on eight mathematical reasoning benchmarks, Relay-OPD achieves the best or second-best results on every benchmark, outperforming standard OPD by +5.73% and the strongest baseline FastOPD by +1.49% on average for 1.7B, with consistent gains at 0.6B. Training trajectory length is reduced by over 50%.
Chinese Translation
在线策略蒸馏（On-policy distillation, OPD）将标记级别的监督建立在学生自身的轨迹上，但存在前缀失败的问题：一旦学生在错误的推理方向上做出承诺，所有后续生成都将基于这一偏差，导致错误的延续，产生不可靠的监督并浪费计算资源。我们识别出在失败前缀上的教师-学生延续不对称性，其中教师倾向于重新引导，而学生则沿着原始方向继续，并将其转化为在接力在线策略蒸馏（Relay On-Policy Distillation, Relay-OPD）中的无标签交接触发器。在训练过程中，Relay-OPD通过让教师在检测到的触发点短暂接管来构建接力轨迹，以生成教师段，之后学生恢复并在生成的轨迹上进行优化。有限的接力预算将干预集中在关键的早期位置，同时限制学生策略的偏离。在八个数学推理基准上，使用Qwen3-4B-Instruct-2507作为教师，Qwen3-0.6B/1.7B-Non-Thinking作为学生，Relay-OPD在每个基准上都取得了最佳或第二最佳的结果，平均超越标准OPD +5.73%，超越最强基线FastOPD +1.49%，在0.6B上也保持了一致的提升。训练轨迹长度减少超过50%。