cs.RO / 1 / 2609.09210
Identifying Habit, Physics, and Nuisance in Robot World Models
识别机器人世界模型中的习惯、物理和干扰
Abstract
Teleoperated demonstrations are often multimodal even when the underlying dynamics are nearly deterministic given the executed action. We argue that this multimodality typically mixes three factors--operator habit in action selection, shared physics, and observation nuisance--and that entangled next-observation predictors absorb all three. We formalize the split with a structural causal model a=g(h,z,u), z'=f(z,a), o=r(z,c), and test it with complementary interventions: replacing or shuffling actions at fixed state sharply increases next-state error, whereas appearance and camera changes should not; habit-aware reverse scoring improves ranking of feasible pasts without rewriting the dynamics. The associated adaptation rule is to freeze a shared physics readout and update only a thin interface. On StackCube, DROID, and RH20T this rule improves low-shot transfer relative to training from scratch, retains cleaner dynamics under corrupted adaptation data, and extends from proprioception to pixel observations with multi-view and multi-step checks. We do not equate latent actions with operator habit, and we do not target large-scale video generation benchmarks.
Chinese Translation
遥操作演示通常是多模态的，即使在执行的动作几乎是确定性的情况下。我们认为，这种多模态性通常混合了三个因素——操作者在动作选择中的习惯、共享的物理特性和观察干扰——而纠缠的下一个观察预测器吸收了这三者。我们通过结构因果模型 a=g(h,z,u), z'=f(z,a), o=r(z,c) 正式化了这种划分，并通过互补干预进行测试：在固定状态下替换或打乱动作会显著增加下一个状态的误差，而外观和相机的变化则不应如此；考虑习惯的反向评分在不重写动态的情况下改善了可行过去的排名。相关的适应规则是冻结共享物理读出，仅更新一个薄接口。在 StackCube、DROID 和 RH20T 上，这一规则相较于从头训练改善了低样本迁移，在受到干扰的适应数据下保持了更清晰的动态，并且从本体感知扩展到像素观察，结合多视角和多步骤检查。我们不将潜在动作等同于操作者习惯，也不针对大规模视频生成基准。
cs.RO / 2 / 2609.09213
Geometry Conditioning in an Embodied SLM: Training Controls and Robustness Diagnostics in a 0.8B Hybrid Model
具身SLM中的几何条件化：在0.8B混合模型中的训练控制与鲁棒性诊断
Abstract
We study how physical-state inputs affect a 0.8B hybrid language model adapted for manipulation with 6.2M trainable parameters. Six conditions are trained on three LIBERO-Spatial tasks and evaluated over three seeds and 540 held-out rollouts. Conditioning recurrent decay gates on geometric increments yields 28.9% success, compared with 36.7% when those increments are shuffled during training and 24.4% without explicit object/goal geometry. Both geometry policies receive correct inputs at evaluation. A token adapter using the same increments scores 27.8%; differences vary across seeds and remain inconclusive. Token-clock conditioning scores 11.1%, including one seed that fails to converge. In separate robustness tests, a state-only relative-coordinate policy retains 7/10 success under frame relabeling, whereas all four tested visual policies fall to at most 3/20 after a 5 cm object displacement. These results show no reliable advantage from training-time geometric alignment under this recipe and illustrate the gap between coordinate invariance and physical-layout generalization. Episode records, seed-level analyses, and figure-generation code accompany the paper.
Chinese Translation
我们研究了物理状态输入如何影响一个适用于操控的0.8B混合语言模型，该模型具有620万可训练参数。六种条件在三个LIBERO-Spatial任务上进行训练，并在三个种子和540个保留回合中进行评估。将递归衰减门与几何增量条件化结合，成功率为28.9%；而在训练过程中对这些增量进行随机打乱时，成功率为36.7%，在没有明确的物体/目标几何时，成功率为24.4%。在评估时，两种几何策略均接收到正确的输入。使用相同增量的令牌适配器得分为27.8%；不同种子之间的差异各异，结果尚不明确。令牌时钟条件化得分为11.1%，其中一个种子未能收敛。在单独的鲁棒性测试中，仅基于状态的相对坐标策略在帧重标定下保持7/10的成功率，而所有四种测试的视觉策略在5厘米物体位移后最多降至3/20。这些结果表明，在这种训练方案下，几何对齐并未带来可靠的优势，并展示了坐标不变性与物理布局泛化之间的差距。论文附带了情节记录、种子级分析和图形生成代码。
cs.RO / 3 / 2609.09217
Design and Attitude Control of an Underwater Quadruped Robot
水下四足机器人设计与姿态控制
Abstract
Legged robots are versatile on land, but their use in underwater environments remains limited. Extending quadruped locomotion to water enables amphibious mobility with applications in inspection, environmental monitoring and disaster response. This paper presents the design, modeling, and experimental validation of a reproducible underwater quadruped robot. The robot is built around custom waterproof motor housings machined from polyoxymethylene plastic, which use off-the-shelf O-rings and dynamic shaft seals. A simplified model is derived to describe the dynamics of this underwater legged system, capturing how drag forces on spherical end effectors transmit torque to the floating base. Building on this model, a closed-loop attitude controller is developed using an error formulation defined on the special orthogonal group SO(3). The controller is evaluated both in simulation and experimentally in a water tank, where the robot tracks desired orientation setpoints in roll, pitch and yaw.
Chinese Translation
腿式机器人在陆地上具有多功能性，但其在水下环境中的应用仍然有限。将四足机器人运动扩展到水中，实现两栖机动性，可应用于巡检、环境监测和灾害响应等领域。本文介绍了一种可复现的水下四足机器人的设计、建模及实验验证。该机器人基于定制的防水电机外壳构建，外壳采用聚甲醛塑料加工，配备现成的O型圈和动态轴封。文中推导了一个简化模型，用于描述该水下腿式系统的动力学，捕捉了球形末端执行器上的阻力如何传递扭矩至浮动基座。在此模型基础上，设计了一种基于特殊正交群SO(3)上的误差表达的闭环姿态控制器。该控制器在仿真和水槽实验中均进行了评估，机器人能够跟踪滚转、俯仰和偏航方向上的期望姿态设定点。
cs.RO / 4 / 2609.09234
Learning to Fly: Stable Vision-Guided UAV Servoing with Compact Target-Centric Cues and Reinforcement Learning
学习飞行：基于紧凑目标中心线索和强化学习的稳定视觉引导无人机伺服控制
Abstract
Vision-guided reinforcement learning for Unmanned Aerial Vehicles (UAVs) remains challenging due to unstable policy optimisation, aggressive exploration, and the cost of high-dimensional visual perception. In this work, we investigate long-horizon UAV visual servoing using compact target-centric cues combined with low-dimensional sensor measurements. Rather than learning directly from RGB images, lightweight target segmentation provides image-space offsets and relative depth, which are combined with quadrotor velocity and projected-gravity measurements into a compact 12D policy observation. We compare Direct PPO with three matched-budget curriculum strategies: a Visual curriculum that progressively expands target placement difficulty, a Dynamics curriculum that gradually relaxes action constraints and smoothing, and a Joint curriculum that combines both progressions. All strategies reach comparable nominal performance, with complementary advantages across tracking metrics. Observation ablations show that proprioceptive measurements are critical for stable flight and image-space cues for target alignment, while explicit depth is not necessary for strong performance in the evaluated setting. Against tuned classical visual-servo controllers, learned policies show greater robustness to strong control and visual perturbations, while the Visual curriculum exhibits the smallest degradation under unseen target motion. Overall, the results demonstrate that compact target-centric representations can support robust long-horizon aerial visual servoing and that visual curriculum training can improve robustness to dynamic distribution shifts despite limited gains in nominal performance.
Chinese Translation
视觉引导的无人机（UAV）强化学习仍然面临不稳定的策略优化、激进的探索以及高维视觉感知成本等挑战。在本研究中，我们探讨了使用紧凑的目标中心线索结合低维传感器测量的长时间跨度无人机视觉伺服控制。我们并非直接从RGB图像中学习，而是通过轻量级目标分割提供图像空间偏移和相对深度，这些信息与四旋翼的速度和投影重力测量结合形成紧凑的12维策略观测。我们将直接的PPO与三种匹配预算的课程策略进行比较：一种视觉课程，逐步增加目标放置的难度；一种动力学课程，逐渐放宽动作约束和光滑度；以及一种联合课程，结合了这两种进展。所有策略在跟踪指标上都达到了可比的名义性能，并在各个指标上具有互补优势。观测消融实验表明，身体感知测量对于稳定飞行至关重要，而图像空间线索对于目标对齐也很重要，而在评估设置中，显式深度并不是强性能所必需的。与调优的经典视觉伺服控制器相比，学习的策略对强控制和视觉扰动表现出更大的鲁棒性，而视觉课程在未见目标运动下表现出最小的性能下降。总体而言，结果表明，紧凑的目标中心表示可以支持鲁棒的长时间跨度空中视觉伺服控制，并且视觉课程训练可以提高对动态分布变化的鲁棒性，尽管名义性能的提升有限。
cs.RO / 5 / 2609.09250
No Free Checker: A Survey of Verifiers for Robot Policies
没有免费的检查器：机器人策略验证器的调查
Abstract
A verifier for robot policies reads a candidate behavior and returns a score for how well it did, used both to evaluate vision-language-action policies and to train them. Verifiers range from success detectors and reward models to runtime monitors, safety filters, and temporal-logic specifications. We survey roughly 150 verifiers and compare them along two properties. Availability is how much a verdict costs, how early in a rollout the verdict arrives, and how often a verdict can be asked for. Availability rises as verdicts get cheaper, earlier, and denser. Credibility is how much a high score tells us about the task. Credibility falls as the judgment becomes gameable and self-serving. We group the verifiers by who supplies the judgment: human verifiers, rule-based and formal verifiers, learned and pretrained verifiers, and model-intrinsic verifiers. Across the four families, we find that credibility falls as availability rises. Regardless of who supplies the judgment, there is no free checker. We then examine what validates a verifier itself, and how much a high score tells us. Three measures appear in the literature: agreement with human labels, the performance of the policy it trains, and behavior under reward hacking. We close with nine metrics that make a verifier claim checkable, and coordinates for the verifiers still to be built.
Chinese Translation
机器人策略的验证器读取候选行为并返回一个评分，以评估其表现，既用于评估视觉-语言-动作策略，也用于训练这些策略。验证器的类型包括成功检测器、奖励模型、运行时监控、安全过滤器和时序逻辑规范。我们调查了大约150个验证器，并从两个属性进行比较。可用性是指判决的成本、判决在执行过程中的到达时间以及可以请求判决的频率。随着判决变得更便宜、更早和更密集，可用性提高。可信度是指高分在多大程度上反映了任务的实际情况。随着判断变得可操控和自利，可信度下降。我们根据谁提供判断将验证器分为四类：人工验证器、基于规则和形式的验证器、学习和预训练的验证器，以及模型内在验证器。在这四个类别中，我们发现随着可用性的提高，可信度下降。无论谁提供判断，都不存在免费的检查器。接着，我们考察了验证器本身的有效性，以及高分在多大程度上反映了真实情况。文献中出现了三种衡量标准：与人工标签的一致性、训练的策略的表现，以及在奖励操控下的行为。最后，我们提出了九个指标，使验证器的声明可检查，并为尚待构建的验证器提供了坐标。
cs.RO / 6 / 2609.09380
AccelMPC: High-Rate, Low-Power FPGA-Accelerated Model Predictive Control for Tiny Drones
AccelMPC：高频率、低功耗的FPGA加速模型预测控制用于微型无人机
Abstract
Unlocking the potential of tiny aerial robots requires order of magnitude improvements in the performance of embedded edge control. In particular, although recent cached model predictive control (MPC) solvers can handle the fast system dynamics and complex constraints required for agile drone flight, their computational demands remain prohibitive for resource-constrained robots, forcing prior implementations to operate at reduced control rates. AccelMPC overcomes this challenge through an end-to-end co-design approach that jointly optimizes the solver algorithm, numerical representation, hardware mapping, and physical integration. AccelMPC pairs a co-designed FPGA-accelerated alternating direction method of multipliers (ADMM)-based MPC solver with a custom 6g PCB, providing high-bandwidth communication for deployment on a 35g Crazyflie. Hardware experiments demonstrate 1 kHz onboard constrained MPC with dynamic obstacles, up to 15.6x faster solve times and 195.4x improvement in energy-delay product over state-of-the-art embedded microcontroller-based solvers, all while scaling to optimization problems with over 20,000 optimization variables and a comparable number of constraints. We release our PCB design files, firmware, and FPGA solver code open source.
Chinese Translation
释放微型空中机器人潜力需要在嵌入式边缘控制性能上实现数量级的提升。特别是，尽管最近的缓存模型预测控制（MPC）求解器能够处理灵活无人机飞行所需的快速系统动态和复杂约束，但其计算需求对于资源受限的机器人仍然过于苛刻，迫使先前的实现只能以降低的控制频率运行。AccelMPC通过端到端的协同设计方法克服了这一挑战，该方法联合优化了求解器算法、数值表示、硬件映射和物理集成。AccelMPC将协同设计的基于交替方向乘子法（ADMM）的FPGA加速MPC求解器与定制的6g PCB相结合，为35g的Crazyflie提供高带宽通信。硬件实验展示了在动态障碍物下1 kHz的机载约束MPC求解，相比于最先进的基于嵌入式微控制器的求解器，其求解时间快至15.6倍，能量延迟乘积提升195.4倍，同时能够扩展到超过20,000个优化变量和相当数量的约束的优化问题。我们将PCB设计文件、固件和FPGA求解器代码开源发布。
cs.RO / 7 / 2609.09403
A Decade of Bayesian Optimization for Controller Tuning and Robot Learning: Tutorial, Review, and Future Prospects
十年贝叶斯优化在控制器调优和机器人学习中的应用：教程、综述与未来展望
Abstract
In the past decade, Bayesian optimization (BO) has emerged as a powerful and adaptable framework for automatic controller tuning and robot learning. This article offers a comprehensive overview of the state-of-the-art in BO, designed to support both researchers and practitioners in understanding recent advancements, practical applications, and future research directions. We begin by adopting a practitioner's perspective, illustrating how to effectively set up BO through a representative controller tuning example. We position BO within the broader context of learning paradigms, ranging from deep reinforcement learning to data-driven control, and highlight scenarios where BO is most advantageous. Next, we discuss the diverse range of BO methods that have been developed to tackle complex problems and specific applications. This article provides a unified perspective on the current landscape of BO, emphasizing its relevance to control systems and robotics, and it highlights future prospects by identifying key research challenges and promising avenues for advancing BO in the field. This includes addressing a significant gap in the BO landscape: the lack of standardized benchmark problems specifically for control-related applications. To foster future research and ensure rigorous evaluation, we start an effort towards a lightweight benchmark suite for control engineering and robotics. We also present metrics and best practices to facilitate direct comparisons between new BO algorithms and established state-of-the-art methods.
Chinese Translation
在过去的十年中，贝叶斯优化（Bayesian Optimization, BO）作为一种强大且灵活的框架，已在自动控制器调优和机器人学习中崭露头角。本文提供了BO的最新进展的全面概述，旨在帮助研究人员和从业者理解近期的进展、实际应用及未来的研究方向。我们首先从实践者的角度出发，通过一个典型的控制器调优示例，展示如何有效地设置BO。我们将BO置于更广泛的学习范式背景中，涵盖从深度强化学习到数据驱动控制的范围，并强调BO最具优势的场景。接下来，我们讨论了为解决复杂问题和特定应用而开发的多种BO方法。本文提供了对当前BO领域的统一视角，强调其在控制系统和机器人领域的相关性，并通过识别关键研究挑战和有前景的研究方向，突出未来的展望。这包括解决BO领域中的一个重要空白：缺乏专门针对控制相关应用的标准化基准问题。为了促进未来的研究并确保严格的评估，我们开始着手开发一个轻量级的控制工程和机器人基准套件。我们还提出了指标和最佳实践，以便于新BO算法与现有最先进方法之间的直接比较。
cs.RO / 8 / 2609.09492
Actuator Dynamics Curricula for Narrow-Viability Tasks in Legged Robot Learning
腿部机器人学习中的窄适应性任务的驱动器动态课程
Abstract
Reinforcement learning has produced capable controllers across a broad range of legged-robot tasks, but a subset of these tasks fail to converge under standard training: those for which most exploration trajectories terminate before producing useful gradient signal. To address such tasks we introduce the \emph{Actuator Dynamics Curriculum}, a procedure that initializes joint stiffness at a high value and anneals it toward the system-identified value as completed episode lengths grow. Using a cart-pole system as a representative example, we show that higher closed-loop joint natural frequency under critical damping enlarges the viability kernel of the underlying Markov Decision Process, increasing the fraction of initial states from which the task is feasible. We validate the kernel monotonicity on the cart-pole and apply the curriculum to a quadrupedal-to-handstand transition on the Boston Dynamics Spot, a narrow-viability task where training under fixed identified stiffness plateaus at a policy that never completes the transition. The trained policy executes the transition in simulation across 10 seeds and transfers to hardware. More broadly, our results suggest that simulated actuator dynamics is a useful axis along which to design curricula for tasks in which exploration is bottlenecked by termination conditions rather than by reward signal.
Chinese Translation
强化学习在广泛的腿部机器人任务中产生了有效的控制器，但其中一部分任务在标准训练下无法收敛：这些任务的大多数探索轨迹在产生有用的梯度信号之前就已终止。为了解决这些任务，我们引入了 extit{驱动器动态课程}，该程序在初始阶段将关节刚度设置为高值，并随着完成的回合长度的增加逐渐降低至系统识别的值。以小车摆系统作为代表性示例，我们展示了在临界阻尼下更高的闭环关节固有频率扩大了基础马尔可夫决策过程的适应性核，从而增加了任务可行的初始状态的比例。我们在小车摆上验证了核的单调性，并将该课程应用于波士顿动力公司的Spot机器人进行四足到倒立的过渡，这是一个窄适应性任务，其中在固定识别刚度下的训练停滞于一个从未完成过渡的策略。经过训练的策略在10个种子下的仿真中成功执行了过渡，并转移到硬件上。更广泛地说，我们的结果表明，模拟驱动器动态是设计课程的一个有用方向，特别是在探索受到终止条件而非奖励信号瓶颈的任务中。
cs.RO / 9 / 2609.09503
Agentic AI-enabled Semantic Commissioning of a Cognitive Digital Twin for Reconfigurable Manufacturing
基于代理的人工智能驱动的认知数字双胞胎的语义委托用于可重构制造
Abstract
Rapid bespoke commissioning of the Cognitive Digital Twin (CDT) is a major challenge in reconfigurable manufacturing. Traditional digital twin (DT) construction methods primarily focus on geometric reconstruction, often neglecting the deep semantic integration and functional interoperability necessary for autonomous reasoning. This paper proposes an agent-based, AI-driven workflow to automate end-to-end CDT debugging. The system utilises LangGraph as a multi-agent orchestration engine to achieve dual-path synthesis: the semantic path extracts technical specifications from unstructured documents using Retrieval Augmented Generation (RAG), while the functional path autonomously discovers and binds to real-time industrial telemetry data using Model Context Protocol (MCP). Experimental validation in a robotic machining cell demonstrates that the system achieves a mean average accuracy (mAP) of 97.2% in perception and reduces the deployment cycle from several weeks to an average of 2 hours, marking a paradigm shift from manual scripting to autonomous orchestration.
Chinese Translation
在可重构制造中，快速定制认知数字双胞胎（Cognitive Digital Twin, CDT）的委托是一个主要挑战。传统的数字双胞胎（Digital Twin, DT）构建方法主要关注几何重建，往往忽视了自主推理所需的深层语义集成和功能互操作性。本文提出了一种基于代理的、人工智能驱动的工作流程，以自动化端到端的CDT调试。该系统利用LangGraph作为多代理编排引擎，实现双路径合成：语义路径使用检索增强生成（Retrieval Augmented Generation, RAG）从非结构化文档中提取技术规范，而功能路径则通过模型上下文协议（Model Context Protocol, MCP）自主发现并绑定实时工业遥测数据。在一个机器人加工单元中的实验验证表明，该系统在感知方面实现了97.2%的平均准确率（mean average accuracy, mAP），并将部署周期从几周缩短到平均2小时，标志着从手动脚本编写到自主编排的范式转变。
cs.RO / 10 / 2609.09597
Compact Visuotactile World Models for Lifting: Prediction, Reward Alignment, and Force Constraints
紧凑的视觉触觉世界模型用于提升：预测、奖励对齐与力约束
Abstract
Accurate contact prediction is useful for robotic manipulation only if it supports effective decisions. We investigate this connection using a compact, randomly initialized visuotactile world model, trajectory-level uncertainty calibration, and behavior-initialized actor-critic learning in imagination. On 160 MuJoCo Lift episodes, adding touch reduces endpoint-force prediction error from 1.058 to 0.228 N and interval-peak error from 2.724 to 0.523 N across three training seeds. However, tactile persistence achieves lower errors of 0.095 and 0.498 N, respectively. Two exploratory control rounds comprise 680 executions on 40 independent test initial conditions. A matched reward revision on fresh test environments increases in-distribution 10 cm lifting success from 20.0% to 93.3%, while success within an 8 N per-finger budget reaches only 33.3%, compared with 70.0% for force feedback. Calibration margins reduce force violations at the cost of task completion. In a separate study of public GelSight recordings, a force regressor achieves 0.04234 N error, but frame-level calibration covers only 15.80% of complete trajectories; trajectory-level calibration raises this to 87.36% at nominal 90% coverage. Together, these findings distinguish improvements in sensing and task reward from improvements in force-constrained control. The evidence is limited to public sensing records and simulator execution, without a demonstrated transfer between them.
Chinese Translation
准确的接触预测只有在支持有效决策时对机器人操作才有用。我们通过一个紧凑的、随机初始化的视觉触觉世界模型、轨迹级不确定性校准和行为初始化的演员-评论家学习进行想象，探讨了这一联系。在160个MuJoCo Lift实验中，添加触觉将末端力预测误差从1.058 N降低到0.228 N，区间峰值误差从2.724 N降低到0.523 N，跨越三个训练种子。然而，触觉持续性分别实现了更低的误差0.095 N和0.498 N。两个探索性控制回合包括在40个独立测试初始条件下的680次执行。在新测试环境中进行的匹配奖励修订使得10 cm的提升成功率从20.0%提高到93.3%，而在每指8 N的预算内成功率仅为33.3%，相比之下，力反馈的成功率为70.0%。校准边际降低了力违规，但以任务完成为代价。在对公共GelSight录音的单独研究中，一个力回归模型达到了0.04234 N的误差，但帧级校准仅覆盖了15.80%的完整轨迹；而轨迹级校准将这一比例提高到87.36%，在名义上达到90%的覆盖率。综合来看，这些发现将感知和任务奖励的改善与力约束控制的改善区分开来。证据仅限于公共感知记录和模拟器执行，尚未证明两者之间的转移。
cs.RO / 11 / 2609.09612
MuJoCable: Reduced-Order Surface-Routed Cable Transmission for Tendon-Driven Robots
MuJoCable：用于腱驱动机器人的降阶表面导向电缆传输
Abstract
Tendon transmissions reduce distal inertia and add compliance, yet routing, slack, and friction govern motion and force transfer. Mainstream rigid-body robotics simulators such as MuJoCo do not jointly resolve moving noncircular contact, unilateral tension, and segment friction. We present MuJoCable, which adds a reduced-order, configuration-dependent cable transmission to MuJoCo. Its routing algorithm jointly optimizes an ordered path across moving analytic and mesh surfaces. A unilateral axial law, directional Capstan propagation, and nodal virtual work map this path to segment tensions and body forces. The warm-started engine plugin applies these forces during simulation and exposes route and load states for design. Pulley benchmarks recover analytical transmission relations with a Capstan-ratio error below 0.5%. On the underactuated 18-joint SpiRobs, MuJoCable reveals friction-driven load growth and proximal redistribution of joint rotation that the native tendon does not represent. Hardware tests on SpiRobs and a tendon-route-coupled finger reproduce observed motion sequences. By making physical threading executable, MuJoCable brings transmission sources of the simulation-to-reality gap into route, cable, and actuator design before fabrication.
Chinese Translation
腱传动减少了远端惯性并增加了柔顺性，但导向、松弛和摩擦支配着运动和力的传递。主流的刚体机器人仿真器如MuJoCo无法共同解决移动的非圆形接触、单边张力和段摩擦。我们提出了MuJoCable，它在MuJoCo中增加了一种降阶的、依赖于配置的电缆传输。其导向算法联合优化了在移动的解析和网格表面上的有序路径。单边轴向定律、方向性Capstan传播和节点虚功将该路径映射到段张力和体力。在仿真过程中，预热启动的引擎插件应用这些力，并暴露出设计所需的路径和负载状态。滑轮基准测试恢复了分析传输关系，Capstan比率误差低于0.5%。在欠驱动的18关节SpiRobs上，MuJoCable揭示了摩擦驱动的负载增长和关节旋转的近端重新分配，而原生腱并未表现出这些特征。在SpiRobs和一个与腱导向耦合的手指上的硬件测试重现了观察到的运动序列。通过使物理穿线可执行，MuJoCable在制造之前将仿真与现实之间的传输源引入到导向、电缆和驱动器设计中。
cs.RO / 12 / 2609.09630
JEPA Policy: Diffusion-Free Imitation Learning via Paired Action and Future Representation Prediction
JEPA策略：通过配对动作和未来表征预测实现无扩散模仿学习
Abstract
Standard behavior cloning supervises actions without explicitly constraining the future representation paired with each demonstrated action chunk. We introduce JEPA Policy, a diffusion-free framework that uses the action chunk and its observed future representation as paired training targets. Action and future-representation tokens interact in a shared Transformer and are refined through two forward passes. Future prediction can therefore shape the representation used to generate actions. Dual-branch and gradient-routing controls attribute the gain to this shared topology rather than to an auxiliary prediction head alone. Across nine simulated tasks, JEPA Policy improves mean success over the action-only MIP baseline and outperforms Diffusion Policy under the evaluated configurations, while adding 0.29 ms to MIP's model latency. A five-task, 630-episode physical-robot study produces the same pooled ranking. Further audits find no complete representation collapse under action supervision and identify a task-conditioned failure-ranking signal in future-prediction error. These results support paired future-representation supervision as a practical approach to low-latency visuomotor imitation without iterative generative sampling.
Chinese Translation
标准行为克隆在监督动作时并未明确限制与每个示范动作块配对的未来表征。我们引入了JEPA策略，这是一种无扩散框架，利用动作块及其观察到的未来表征作为配对训练目标。动作和未来表征标记在共享的Transformer中相互作用，并通过两次前向传递进行优化。因此，未来预测可以塑造用于生成动作的表征。双分支和梯度路由控制将增益归因于这一共享拓扑，而不仅仅是辅助预测头。在九个模拟任务中，JEPA策略在动作仅基线MIP上提高了平均成功率，并在评估的配置下优于扩散策略，同时将MIP的模型延迟增加了0.29毫秒。在五个任务、630集的物理机器人研究中，产生了相同的汇总排名。进一步审计发现，在动作监督下没有完全的表征崩溃，并在未来预测误差中识别出任务条件的失败排名信号。这些结果支持配对未来表征监督作为一种低延迟视觉运动模仿的实用方法，而无需迭代生成采样。
cs.RO / 13 / 2609.09650
A Risk-Sensitive and Uncertainty-Aware Decision-Making and Control Framework for Safe and Robust Autonomous Driving
一种风险敏感与不确定性感知的安全稳健自主驾驶决策与控制框架
Abstract
Reinforcement learning (RL) has demonstrated considerable potential for autonomous driving decision-making. However, its deployment in urban autonomous driving, particularly at highly interactive unsignalized intersections, remains challenging, as learned policies may struggle to maintain both safety and robust decision-making in complex traffic situations. Conventional safety-filtering approaches typically employ fixed conservative constraints, which may improve safety at the cost of excessive intervention and degraded traffic efficiency. To address these limitations, we propose a Risk-sensitive and Uncertainty-aware Decision-making and Control (RUDC) framework for safe and robust autonomous driving. RUDC couples risk-sensitive distributional RL with ensemble-based policy uncertainty quantification, jointly accounting for tail risks in return distributions and uncertainty in learned policies. An uncertainty-aware high-order control barrier function (HOCBF)-based safety correction mechanism adaptively adjusts constraint strictness according to policy uncertainty, while a learnable residual predictor compensates for CBF model mismatches and discretization errors. Extensive simulations at unsignalized intersections demonstrate that RUDC achieves a favorable balance among safety, efficiency, and robustness, outperforming representative safe RL baselines under both nominal and challenging OOD and long-tail scenarios while satisfying real-time requirements.
Chinese Translation
强化学习（RL）在自主驾驶决策中展现了相当大的潜力。然而，其在城市自主驾驶中的应用，特别是在高度互动的无信号交叉口，仍然面临挑战，因为学习到的策略可能难以在复杂交通情况下保持安全性和稳健的决策能力。传统的安全过滤方法通常采用固定的保守约束，这可能在提高安全性的同时导致过度干预和交通效率下降。为了解决这些局限性，我们提出了一种风险敏感与不确定性感知的决策与控制框架（RUDC），旨在实现安全和稳健的自主驾驶。RUDC将风险敏感的分布式强化学习与基于集成的策略不确定性量化相结合，联合考虑收益分布中的尾部风险和学习策略中的不确定性。基于不确定性感知的高阶控制屏障函数（HOCBF）安全修正机制根据策略不确定性自适应调整约束的严格性，同时可学习的残差预测器补偿CBF模型的不匹配和离散化误差。在无信号交叉口的大量仿真中，RUDC在安全性、效率和稳健性之间实现了良好的平衡，优于代表性的安全强化学习基线，在名义和具有挑战性的OOD及长尾场景下均满足实时要求。
cs.RO / 14 / 2609.09692
CT-SAFR: Safe and Interpretable Chain-of-Thought Reasoning for Autonomous Robots: A Multi-Layered Verification Framework for Trustworthy AI-Driven Robotic Decision Making
CT-SAFR：自主机器人的安全与可解释的思维链推理：可信赖的人工智能驱动机器人决策的多层验证框架
Abstract
Chain-of-Thought (CoT) prompting enables LLMs to perform explicit, step-by-step reasoning, creating opportunities for sophisticated autonomous robots. However, recent research reveals that reasoning models verbalize their actual decision processes only 25-39% of the time, with faithfulness degrading 44% on complex tasks. This paper presents CT-SAFR (Chain-of-Thought Safety and Faithfulness for Robotics), a multi-layered verification framework achieving 94.2% hallucination detection (n = 500, 95% CI: 91.8-95.9%) with sub-500ms latency. Through a warehouse robot case study, this work demonstrates 87% reduction in unsafe reasoning outputs (p < 0.001) and provides recommendations for responsible deployment of reasoning-capable autonomous robots.
Chinese Translation
思维链（Chain-of-Thought, CoT）提示使大型语言模型（LLMs）能够进行明确的逐步推理，为复杂的自主机器人创造了机会。然而，近期研究表明，推理模型仅在25-39%的情况下能够准确表达其实际决策过程，且在复杂任务中其忠实度下降了44%。本文提出了CT-SAFR（机器人领域的思维链安全与忠实度），这是一个多层验证框架，能够以94.2%的幻觉检测率（n = 500, 95% CI: 91.8-95.9%）实现低于500毫秒的延迟。通过一个仓库机器人案例研究，本研究展示了不安全推理输出减少87%（p < 0.001），并为负责任地部署具有推理能力的自主机器人提供了建议。
cs.RO / 15 / 2609.09745
PccDiffuser: Multi-solution Motion Planning for Continuum Robots
PccDiffuser：连续机器人多解运动规划
Abstract
We present the PccDiffuser, a conditional diffusion framework for continuum robots that learns a multimodal distribution over complete configuration-space paths and samples multiple candidate solutions in parallel, which are subsequently converted into an executable trajectory by time allocation considering actuator constraints. Under the piecewise constant-curvature model, we use exponential co-ordinates to describe the robot kinematics, and use graph neural network to encode a variable number of environment obstacles. Analytical differential kinematics is incorporated in the denoising process to improve terminal accuracy and whole-body clearance. On a mixed test set comprising workspace with zero to four obstacles, PccDiffuser achieved a success rate of 91\%. Compared with existing sampling- and optimisation-based benchmarks, it delivered both a higher success rate and greater computational efficiency, with the latter advantage becoming more substantial when sampling more candidate solutions. Experiments on a three-section tendon-driven continuum robot further demonstrate consecutive planning, multi-solution planning, and whole-body obstacle avoidance.
Chinese Translation
我们提出了PccDiffuser，这是一种针对连续机器人的条件扩散框架，能够学习完整配置空间路径的多模态分布，并以并行方式采样多个候选解，这些解随后通过考虑执行器约束的时间分配转换为可执行的轨迹。在分段常曲率模型下，我们使用指数坐标描述机器人运动学，并利用图神经网络对环境障碍物的可变数量进行编码。分析性微分运动学被纳入去噪过程，以提高终端精度和全身间隙。在一个包含零到四个障碍物的混合测试集上，PccDiffuser实现了91%的成功率。与现有的基于采样和优化的基准相比，它不仅提供了更高的成功率，还具有更大的计算效率，后者在采样更多候选解时优势更加明显。在一个三段腱驱动的连续机器人上的实验进一步展示了连续规划、多解规划和全身避障能力。
cs.RO / 16 / 2609.09752
HiRAD: A Flexible Large-Scale AGV Routing System
HiRAD：一种灵活的大规模AGV路径规划系统
Abstract
Automatic Guided Vehicles (AGVs) substantially boost warehouse throughput, but routing large-scale AGV fleets remains challenging. Classical Multi-Agent Pathfinding solvers suffer from exploding combinatorial complexity and super-quadratic runtime, while relying on idealized grid or piecewise-linear motion models that mismatch real-world kinematics. Recent Reinforcement Learning (RL) solutions improve flexibility via decentralized agent policies but depend on discretized spatiotemporal representations, require millions of episodes to converge, and incur full-map observation at every step, which leads to large models, slow convergence, and high inference latency that violates real-time industrial control constraints. To address these bottlenecks, we propose HiRAD, a hierarchical RL framework for continuous-space AGV routing with real-time guarantees: (1) a step-level spatiotemporal representation that translates continuous motion into a differentiable RL problem, (2) a hierarchical strategy that splits heading choice from velocity control to reduce the action space, and (3) an asynchronous event-driven decision pipeline that lowers inference complexity from O(n^2) to O(n) and cuts per-step latency by as much as 71 percent. Across random graphs and two warehouse maps, HiRAD reduces makespan by 45 percent to 63 percent and shortens end-to-end runtime.
Chinese Translation
自动导引车（AGV）显著提升了仓库的吞吐量，但大规模AGV车队的路径规划仍然面临挑战。经典的多智能体路径规划求解器受到组合复杂性爆炸和超二次运行时间的困扰，同时依赖于理想化的网格或分段线性运动模型，这与现实世界的运动学不匹配。近期的强化学习（RL）解决方案通过去中心化的智能体策略提高了灵活性，但依赖于离散的时空表示，需要数百万个回合才能收敛，并且在每一步都需要全图观察，这导致模型庞大、收敛缓慢以及高推理延迟，违反了实时工业控制的约束。为了解决这些瓶颈，我们提出了HiRAD，一个具有实时保证的连续空间AGV路径规划的层次化RL框架：（1）一个步级时空表示，将连续运动转化为可微分的RL问题；（2）一个层次化策略，将航向选择与速度控制分开，以减少动作空间；（3）一个异步事件驱动的决策管道，将推理复杂度从O(n^2)降低到O(n)，并将每步延迟减少多达71%。在随机图和两个仓库地图上，HiRAD将完工时间减少了45%至63%，并缩短了端到端的运行时间。
cs.RO / 17 / 2609.09808
GTA-2: A Multi-VLM Framework for Synthesizing Robot Manipulation Skills via Grounded Task Axes
GTA-2：通过基础任务轴合成机器人操作技能的多VLM框架
Abstract
Robotic manipulation tasks are often decomposed into behaviors or skills. However, one often needs to predefine these behaviors for specific tasks or try to cover a wide range of tasks using generic skills. As a result, these behaviors can remain too coarse to expose the geometric, control, and scene-dependent decisions required for execution. We introduce Grounded Task Axes v2 (GTA-2), a modular multi-VLM framework that constructs executable, task-bespoke manipulation skills from reusable object-centric task-axis components. Rather than predicting actions end-to-end or composing fixed task-level primitives, GTA-2 represents each skill as semantic subtasks comprising task-relevant keypoints and axes, controller compositions, and scene-dependent parameters. Four specialized VLM agents separately decompose the task, construct an abstract task-axis skill, assign controller parameters, and ground the required visual features from RGB-D observations. This abstraction-to-grounding factorization enables zero-shot skill generation without task-specific robot demonstrations, policy training, or fine-tuning. It also keeps intermediate decisions explicit, allowing targeted human feedback to refine an incorrect stage while preserving correct components. We evaluate GTA-2 on 14 real-robot manipulation tasks against a VLA policy pi_{0.5} and two Code-as-Policies baselines using task-axis controllers or conventional robot primitives. GTA-2 achieves an average zero-shot success rate of 73.9%, exceeding the strongest baseline by 31.4 percentage points, while targeted refinement raises GTA-2's average success rate to 90.7%. Project page: https://gta2-project.github.io/
Chinese Translation
机器人操作任务通常被分解为行为或技能。然而，通常需要为特定任务预定义这些行为，或者尝试使用通用技能覆盖广泛的任务。因此，这些行为可能过于粗略，无法揭示执行所需的几何、控制和场景依赖决策。我们引入了基础任务轴版本2（GTA-2），这是一个模块化的多VLM框架，能够从可重用的以对象为中心的任务轴组件构建可执行的、任务定制的操作技能。GTA-2并不是通过端到端预测动作或组合固定的任务级原语，而是将每个技能表示为由与任务相关的关键点和轴、控制器组合以及场景依赖参数组成的语义子任务。四个专门的VLM代理分别分解任务、构建抽象任务轴技能、分配控制器参数，并从RGB-D观测中获取所需的视觉特征。这种从抽象到具体的分解使得在没有任务特定机器人演示、策略训练或微调的情况下实现零-shot技能生成成为可能。它还保持了中间决策的显式性，允许针对性的人工反馈来细化不正确的阶段，同时保留正确的组件。我们在14个真实机器人操作任务上评估了GTA-2，比较了VLA策略pi_{0.5}和两个使用任务轴控制器或传统机器人原语的Code-as-Policies基线。GTA-2实现了73.9%的平均零-shot成功率，超过了最强基线31.4个百分点，而针对性细化将GTA-2的平均成功率提高到90.7%。项目页面：https://gta2-project.github.io/
cs.RO / 18 / 2609.09918
ViBe: Visual Behavior Adaptation for Perceptive Humanoid Whole-Body Control
ViBe：用于感知类人全身控制的视觉行为适应
Abstract
Motion tracking provides a scalable recipe for humanoid whole-body control. By design, the resulting trackers lack exteroceptive feedback hence reacting to the environment remains the responsibility of a higher-level planner. Existing perceptive controllers train geometry-only encoders from scratch, trading semantics for sim-to-real ease, and typically rely on teacher-student distillation for a task of interest. We present ViBe, a post-training framework for adapting motion trackers to perceptive control tasks. We leverage pre-trained visual encoders with a multi-query extractor module to learn task-relevant perceptive feedback. This feedback is grafted onto the tracker's input via low-rank adapters, enabling parameter-efficient fine-tuning. Given a task reward and a reference dataset, this modular controller can be adapted directly via policy optimization. Across four tasks, ViBe shows zero-shot sim-to-real transfer spanning perceptive walking on curbs and parkour, Repose Cube, omni-object loco-manipulation, and dodgeball, with visually robust performance across outdoor, low-light, and RGB distractor conditions. Finally, we solve a goal-oriented Repose Cube task with a deliberately simple planner, demonstrating the efficacy of perceptive controllers, adapted by our approach.
Chinese Translation
运动跟踪为类人全身控制提供了一种可扩展的方案。由于设计原因，所产生的跟踪器缺乏外部感知反馈，因此对环境的反应仍然是高层规划者的责任。现有的感知控制器从头开始训练仅基于几何的编码器，牺牲了语义以便于从模拟到现实的转移，并通常依赖于教师-学生蒸馏来完成特定任务。我们提出了ViBe，一种用于将运动跟踪器适应于感知控制任务的后训练框架。我们利用预训练的视觉编码器和多查询提取模块来学习与任务相关的感知反馈。该反馈通过低秩适配器接入跟踪器的输入，从而实现参数高效的微调。在给定任务奖励和参考数据集的情况下，该模块化控制器可以通过策略优化直接进行适应。在四个任务中，ViBe展示了零样本的模拟到现实转移，涵盖了在人行道和跑酷上的感知行走、静态立方体、全物体运动操控和躲避球，并在户外、低光和RGB干扰条件下表现出视觉鲁棒性。最后，我们使用一个故意简单的规划器解决了一个以目标为导向的静态立方体任务，展示了通过我们的方法适应的感知控制器的有效性。
cs.RO / 19 / 2609.09941
HaWMPO: Hallucination-Aware World Model-based Policy Optimization for Generalist Robot Policy
HaWMPO：基于幻觉感知的世界模型政策优化用于通用机器人政策
Abstract
Generalist robot policies have demonstrated strong generalization across robotic manipulation tasks, yet their success rates remain limited in com- plex long-horizon scenarios. Recent methods improve Visual-Language-Action (VLA) policies through online reinforcement learning on real robots, but such training relies on costly physical interactions, suffers from low sample efficiency, and may introduce hardware and safety risks. World models offer a promising alternative by enabling policy optimization with imagined rollouts. However, long-horizon rollouts generated by world models often suffer from prediction hal- lucinations, producing biased state transitions that can mislead policy learning. To address this issue, we propose Hallucination-aware World Model-based Pol- icy Optimization (HaWMPO), a closed-loop reinforcement learning pipeline for VLA policy post-training with world models. Specifically, HaWMPO introduces an action-conditioned hallucination-aware model to estimate the reliability of gen- erated image sequences, and incorporates hallucination scores into group relative policy optimization through a Reward-Soft mechanism, suppressing unreliable ac- tion chunks during training. On the LIBERO benchmark, HaWMPO achieves the best average success rate, with gains of 15.0% over the base model and 2.8% over the strongest baseline; real-world experiments on a G1 robot further validate its effectiveness, raising the average success rate on two manipulation tasks from 67.5% to 80.0%.
Chinese Translation
通用机器人政策在机器人操控任务中表现出强大的泛化能力，但在复杂的长时间场景中的成功率仍然有限。最近的方法通过在真实机器人上进行在线强化学习来改进视觉-语言-动作（VLA）政策，但这种训练依赖于昂贵的物理交互，样本效率低下，并可能引入硬件和安全风险。世界模型提供了一种有前景的替代方案，通过想象的回放实现政策优化。然而，世界模型生成的长时间回放通常会受到预测幻觉的影响，产生偏倚的状态转移，可能误导政策学习。为了解决这个问题，我们提出了幻觉感知的世界模型政策优化（HaWMPO），这是一个用于VLA政策后训练的闭环强化学习管道，基于世界模型。具体而言，HaWMPO引入了一种动作条件的幻觉感知模型，以估计生成的图像序列的可靠性，并通过奖励-软机制将幻觉评分纳入群体相对政策优化，在训练过程中抑制不可靠的动作片段。在LIBERO基准测试中，HaWMPO实现了最佳的平均成功率，相较于基础模型提高了15.0%，相较于最强基线提高了2.8%；在G1机器人上的真实世界实验进一步验证了其有效性，将两个操控任务的平均成功率从67.5%提高到80.0%。
cs.RO / 20 / 2609.10021
RoboDrop: Curating VLA Post-Training Data via Local Gradient Compatibility
RoboDrop：通过局部梯度兼容性策划 VLA 后训练数据
Abstract
Vision--language--action (VLA) models acquire broad generalization through large-scale pretraining, yet adapting them to a new task and robot embodiment still requires post-training on newly collected data. Unlike pretraining, post-training targets task- and embodiment-specific adaptation, making it particularly sensitive to data quality. In practice, collected robot datasets often contain heterogeneous errors, including execution mistakes, sensor drift, and timestamp misalignment, which can impair post-training and policy performance. Manual inspection is costly, while existing data-cleaning methods are typically tailored to particular corruption types. To address these challenges, we introduce \textsc{RoboDrop}, a data-curation framework that audits supervision using local gradient compatibility measured along the training trajectory as a proxy for its effect on post-training performance. During a one-epoch warm-up run, RoboDrop scores each candidate sample online by comparing its gradient with those of task-semantic and visually matched validation samples. The resulting sample scores are aggregated at the episode level, and a simple automatic post-processing rule converts them into filtering decisions. We evaluate RoboDrop on controlled observation--action corruptions, naturally suboptimal demonstrations in simulation, and real-robot datasets containing non-expert collection errors. Across these settings, RoboDrop more accurately distinguishes unreliable demonstrations than prior methods, while post-training on the curated data consistently yields stronger downstream policies, with average real-robot rollout success rising from $35.0\%$ to $67.5\%$. These results establish training-trajectory-aware, context-conditioned supervision auditing as an effective approach to robust VLA post-training.
Chinese Translation
视觉-语言-动作（VLA）模型通过大规模预训练获得广泛的泛化能力，但将其适应于新任务和机器人体现仍需在新收集的数据上进行后训练。与预训练不同，后训练的目标是任务和体现特定的适应性，这使其对数据质量特别敏感。在实践中，收集的机器人数据集通常包含异质性错误，包括执行错误、传感器漂移和时间戳不对齐，这些错误可能会损害后训练和策略性能。人工检查成本高昂，而现有的数据清理方法通常针对特定的损坏类型。为了解决这些挑战，我们提出了 extsc{RoboDrop}，一种数据策划框架，通过沿训练轨迹测量的局部梯度兼容性来审计监督，作为其对后训练性能影响的代理。在一次周期的热身运行中，RoboDrop 在线评分每个候选样本，通过将其梯度与任务语义和视觉匹配的验证样本的梯度进行比较。生成的样本评分在回合级别上进行聚合，简单的自动后处理规则将其转换为过滤决策。我们在受控观察-动作损坏、仿真中的自然次优演示以及包含非专家收集错误的真实机器人数据集上评估了 RoboDrop。在这些设置中，RoboDrop 比以前的方法更准确地区分不可靠的演示，而在策划数据上进行后训练则持续产生更强的下游策略，真实机器人回合成功率从 $35.0\%$ 提升至 $67.5\%$。这些结果确立了基于训练轨迹感知的、上下文条件的监督审计作为一种有效的鲁棒 VLA 后训练方法。
cs.RO / 21 / 2609.10024
AXON: A ROS 2 RMW with Shared-Memory/QUIC Transport and QKD/ML-KEM Key Establishment
AXON：一种具有共享内存/QUIC传输和QKD/ML-KEM密钥建立的ROS 2 RMW
Abstract
Robot Operating System 2 (ROS 2) standardizes application code against a middleware interface (RMW) whose reference implementations are built on the Data Distribution Service (DDS). We present AXON, an alternative ROS 2 RMW implementation that separates transport policy by deployment scope. A Rust core and C++ adapter use POSIX shared-memory rings for same-host communication, QUIC for remote communication, and a daemon for discovery and graph synchronization. We then describe two fail-closed TLS 1.3 key-establishment configurations for remote traffic. The classic configuration offers only the hybrid X25519MLKEM768 group, preventing negotiation of a classical-only group. The qkd configuration imports a 256-bit key obtained through the ETSI GS QKD 014 API as a pairwise external PSK and offers no Diffie-Hellman group. Its default messages10 strategy additionally protects remote application messages with AES-256-GCM, rotating KME material after ten outgoing messages and using a fresh nonce per envelope; session relies on QUIC protection alone. The external-PSK path requires a narrow extension to rustls, now bundled with AXON. We define the threat model, distinguish peer authentication in the two configurations, and delimit the implementation-level validation from ROS 2 conformance, comparative performance, and physical-QKD validation.
Chinese Translation
机器人操作系统2（ROS 2）通过中间件接口（RMW）标准化应用代码，其参考实现基于数据分发服务（DDS）。我们提出了AXON，一种替代的ROS 2 RMW实现，它根据部署范围分离传输策略。一个Rust核心和C++适配器使用POSIX共享内存环进行同主机通信，使用QUIC进行远程通信，并通过守护进程进行发现和图同步。接下来，我们描述了两种用于远程流量的故障关闭TLS 1.3密钥建立配置。经典配置仅提供混合X25519MLKEM768组，防止协商仅经典组。qkd配置通过ETSI GS QKD 014 API导入256位密钥作为成对外部预共享密钥（PSK），并不提供Diffie-Hellman组。其默认消息策略额外使用AES-256-GCM保护远程应用消息，在发送十条消息后轮换KME材料，并为每个信封使用新随机数；会话仅依赖于QUIC保护。外部PSK路径需要对rustls进行狭义扩展，该扩展现已与AXON捆绑在一起。我们定义了威胁模型，区分了两种配置中的对等认证，并将实现级验证与ROS 2一致性、比较性能和物理QKD验证区分开来。
cs.RO / 22 / 2609.10033
What Symmetry Buys a Learned Motion Planner
学习型运动规划器所获得的对称性
Abstract
Learning-based motion planners pay at training what classical planners pay per query. Trained in world coordinates, they relearn the same motion at every position and orientation. Existing work restores the missing rigid-body equivariance in the training data, in the inference operator, or in the weights, and each carries a cost. We ask how much of that equivariance the planning query supplies for free. A start s and a goal g determine a frame in closed form, with origin at their midpoint and first axis along g-s. Expressing trajectory and obstacles in that frame removes three translations and two rotations of SE(3), at initialisation, for one cross product per query and with no constraint on the architecture. A single rotation about the start-goal axis remains, and no continuous rule removes it. On a cluttered 3D benchmark, holding architecture, data and budget fixed, the frame raises the held-out collision-free rate from 14.60% to 51.10%, where a straight segment from start to goal scores 15.6% and the world-frame model does not beat it. We build all three mechanisms for the residual rotation and each is worth under a point, though the equivariant backbone reaches any given level two to three times sooner. What the representation supplies therefore dominates what any mechanism enforces, and the standard diagnostic does not see the difference: two models with indistinguishable non-equivariance residuals differ by 28 points. Calibrated against a non-symmetry intervention, the frame is not even the largest effect available, since local geometry is worth +40.0 where the frame is worth +36.5.
Chinese Translation
基于学习的运动规划器在训练阶段的代价与经典规划器在查询阶段的代价相同。它们在世界坐标系中进行训练，在每个位置和方向上重新学习相同的运动。现有的研究通过恢复训练数据、推理操作符或权重中缺失的刚体等变性来解决这一问题，但每种方法都有其成本。我们探讨规划查询在多大程度上免费提供了这种等变性。起点 s 和目标 g 确定了一个封闭形式的框架，其原点位于它们的中点，第一轴沿着 g-s 方向。将轨迹和障碍物表示在该框架中，可以在初始化时消除三个平移和两个 SE(3) 旋转，每个查询只需一个叉乘，并且对架构没有约束。仅剩下围绕起点-目标轴的单一旋转，而没有任何连续规则可以消除它。在一个拥挤的 3D 基准测试中，固定架构、数据和预算的情况下，该框架将保留的无碰撞率从 14.60% 提高到 51.10%，而从起点到目标的直线段得分为 15.6%，世界框架模型未能超越它。我们构建了三种机制来处理剩余旋转，每种机制的价值均低于一个点，尽管等变骨干网络能够在两到三倍的速度内达到任何给定水平。因此，表示所提供的内容主导了任何机制所强加的内容，而标准诊断无法识别这种差异：两个具有不可区分的非等变残差的模型相差 28 分。与非对称干预进行校准后，该框架甚至不是可用的最大影响，因为局部几何的价值为 +40.0，而框架的价值为 +36.5。
cs.RO / 23 / 2609.10050
Grounding Generated Video Plans in Simulation Towards Versatile Dexterous Controllers
在仿真中将生成的视频计划与灵活的灵巧控制器相结合
Abstract
Generated hand-object interaction (HOI) videos provide a controllable way to propose manipulation motions. Simulation-based HOI tracking can translate such kinematic references into feasible low-level control, but its scalability is limited by the lack of reliable reference motions. We therefore combine generated videos with simulation-based HOI grounding: during training, generated videos provide diverse motion references for learning a multi-object, multi-trajectory HOI tracker, and at deployment, the video model produces motion plans that are executed by the learned tracker. In particular, we propose a method that enables scalable reference generation by HOI reconstruction with minimal manual intervention and successfully grounds more than 1,500 generated videos in simulation, achieving success rates over 25 percentage points higher than those of baselines during simulation-based training. In real-world closed-loop experiments, it achieves diverse grasps, including functional grasps, non-prehensile manipulation, and post-grasp object-pose tracking. Videos and code are available at https://boyuan-an.github.io/GALATEA/.
Chinese Translation
生成的手-物体交互（HOI）视频提供了一种可控的方式来提出操作动作。基于仿真的HOI跟踪可以将这些运动学参考转换为可行的低级控制，但其可扩展性受到可靠参考动作缺乏的限制。因此，我们将生成的视频与基于仿真的HOI基础结合起来：在训练过程中，生成的视频为学习多物体、多轨迹的HOI跟踪器提供了多样化的运动参考，而在部署时，视频模型生成的运动计划由学习到的跟踪器执行。特别地，我们提出了一种方法，通过HOI重建以最小的人工干预实现可扩展的参考生成，并成功地在仿真中为超过1,500个生成视频提供基础，其成功率比基线在基于仿真的训练中高出超过25个百分点。在真实世界的闭环实验中，它实现了多样化的抓取，包括功能性抓取、非抓取操作和抓取后物体姿态跟踪。视频和代码可在 https://boyuan-an.github.io/GALATEA/ 获取。
cs.RO / 24 / 2609.10082
Automatic Reproducible Camera Intrinsic Calibration
自动可重复的相机内参标定
Abstract
Accurate camera intrinsic calibration is fundamental to robot perception, and the accuracy depends on the quality of the collected images. However, existing target-based calibration methods often require the practitioner to manually filter out high-quality images and to specify an appropriate radial distortion order. This paper presents a fully automatic intrinsic calibration pipeline that determines both from the collected data. We adopt an iterative rejection scheme that estimates parameters on a candidate image set and removes views whose mean residual exceeds a multiple of the median. Crucially, this process runs independently under each candidate distortion order, so that the retained image set is consistent with the residual scale of that order. Further, the distortion order is selected on held-out images, with the intrinsics and distortion fixed and only the board pose re-estimated, ensuring that an added coefficient is supported by independent observations. Finally, we integrate both steps into an interactive calibration tool that supports full-pipeline data inspection and parameter estimation. Experiments on our own camera data and five public real-world datasets show that image filtering reduces the held-out reprojection error by 25\%, the order selection further by 5\%, achieving the lowest held-out mean among four compared configurations without manual image selection. We will release the code and data to facilitate future research.
Chinese Translation
准确的相机内参标定是机器人感知的基础，其准确性依赖于所收集图像的质量。然而，现有的基于目标的标定方法通常需要操作者手动筛选高质量图像，并指定合适的径向畸变阶数。本文提出了一种完全自动化的内参标定流程，该流程能够从收集的数据中自动确定这两个参数。我们采用了一种迭代拒绝方案，该方案在候选图像集上估计参数，并移除均方残差超过中位数倍数的视图。关键是，这一过程在每个候选畸变阶数下独立运行，从而确保保留的图像集与该阶数的残差尺度一致。此外，畸变阶数是在保留图像上选择的，内参和畸变保持不变，仅重新估计板的姿态，确保新增系数由独立观测支持。最后，我们将这两个步骤整合到一个交互式标定工具中，支持全流程数据检查和参数估计。在我们自己的相机数据和五个公共真实世界数据集上的实验表明，图像筛选将保留的重投影误差降低了25%，而阶数选择进一步降低了5%，在不进行手动图像选择的情况下，达到了四种对比配置中最低的保留均值。我们将发布代码和数据，以促进未来的研究。
cs.RO / 25 / 2609.10137
Assembling Two Parts in One Hand
一只手中组装两个部件
Abstract
A hallmark of human dexterity is the cooperative use of fingers, where different fingers take on distinct yet coordinated roles to accomplish fine manipu- lation, such as capping a pen with the hand that holds it. We study this finger-level coordination through in-hand assembly: mating two rigid objects within a single dexterous hand, with no second arm and no fixture. We present a reinforcement learning formulation to solve this problem in a unified framework, which is driven by a goal relative pose between the two parts. Finger coordination is shaped by a function-based auxiliary reward and regularized toward a single human reference pose, while domain randomization and a fusion of historical proprioception and object observation confer robustness to occlusion-induced estimation noise. The same recipe solves three different assembly tasks (Bottle, Syringe, and Marker). Trained purely in simulation, the policies transfer zero-shot to hardware with a single camera, demonstrating robustness to state-estimation errors caused by oc- clusion. Our experiments also reveal that in-hand assembly places demands on hand morphology and can serve as a benchmark for modern robotic hand systems. Videos and code are available at https://ltbgbird.github.io/in-hand-assembly-page/.
Chinese Translation
人类灵巧性的一个标志是手指的协同使用，不同的手指承担不同但协调的角色，以完成精细操作，例如用握住的手为笔盖上盖。我们通过手内组装研究这种手指级协调：在单个灵巧手中配合两个刚性物体，不使用第二只手臂和夹具。我们提出了一种强化学习的框架来解决这个问题，该框架由两个部件之间的目标相对姿态驱动。手指协调通过基于函数的辅助奖励进行塑造，并朝向单一的人类参考姿态进行正则化，同时领域随机化和历史本体感知与物体观察的融合增强了对遮挡引起的估计噪声的鲁棒性。同样的方法解决了三种不同的组装任务（瓶子、注射器和记号笔）。在纯模拟中训练的策略能够零-shot 转移到仅有一台摄像头的硬件上，展示了对遮挡引起的状态估计误差的鲁棒性。我们的实验还揭示，手内组装对手的形态提出了要求，并可以作为现代机器人手系统的基准。视频和代码可在 https://ltbgbird.github.io/in-hand-assembly-page/ 获取。
cs.RO / 26 / 2609.10166
Future-Aware Flow Planning for Safe UAV Target Following
面向未来的流动规划以实现安全的无人机目标跟踪
Abstract
UAV target following in cluttered environments is inherently predictive: current-state followers can lag behind turns, choose blocked corridors, or trade tracking for unsafe near-horizon motion. We propose a future-aware flow planning framework for state-informed UAV target following. Predicted target futures guide clean UAV trajectory generation as horizon-aligned residual signals, while risk-scored executable-prefix repair is embedded inside the sampling loop. On fixed ID/OOD receding-horizon benchmarks, the planner improves the intended safety--tracking trade-off rather than dominating every metric: it matches zero measured ID collision rate with the highest ID safe-tracking time, and gives the lowest OOD macro collision rate and final tracking error among the displayed methods, while Future-MPC remains smoother and stronger on some thresholded OOD success metrics under its hand-designed objective. Ablations show that future adaptation improves candidate generation before safety repair, and simulator-facing stress tests probe interface, sensing, and controller-execution effects. These results support horizon-aligned future adaptation and embedded prefix repair as complementary ingredients for safe UAV target following under the tested simulation conditions.
Chinese Translation
在复杂环境中，无人机（UAV）目标跟踪本质上是预测性的：当前状态的跟踪者可能在转弯时滞后，选择被阻塞的通道，或者在不安全的近地平线运动中权衡跟踪。我们提出了一种面向未来的流动规划框架，用于基于状态的信息进行无人机目标跟踪。预测的目标未来通过与地平线对齐的残差信号引导干净的无人机轨迹生成，同时在采样循环中嵌入了风险评分的可执行前缀修复。在固定的ID/OOD（外部分布）递归地平线基准测试中，该规划器改善了预期的安全性与跟踪之间的权衡，而不是主导每个指标：它在测得的ID碰撞率为零的情况下，匹配了最高的ID安全跟踪时间，并在所展示的方法中提供了最低的OOD宏观碰撞率和最终跟踪误差，而Future-MPC在一些阈值化的OOD成功指标上保持更平滑和更强的表现。消融实验表明，未来适应性在安全修复之前改善了候选生成，而面向模拟器的压力测试探讨了接口、传感和控制执行的影响。这些结果支持面向地平线的未来适应性和嵌入式前缀修复作为在测试的模拟条件下实现安全无人机目标跟踪的互补要素。
cs.RO / 27 / 2609.10169
Multi-Robot Scanner for Automated Full-Body Dermoscopic Imaging
用于自动全身皮肤镜成像的多机器人扫描仪
Abstract
This paper outlines the specifications and design approach used to construct a full body imaging scanner capable of capturing skin lesions at a dermatoscopic level using cameras mounted on the end-effectors of four UR10 manipulators. The system possesses a view-planning algorithm capable of appropriately selecting the best camera position to acquire images of moles, a high-level controller to allow the manipulators to work simultaneously and a collision-detector that halts the manipulators when they make contact with an object or a person. We evaluate the system through real-patient full-body scans, comparing acquired images against contact dermoscopy and an existing total-body photography system (Vectra) across clinically relevant lesion features, and quantify true optical resolving power using a USAF 1951 resolution target, yielding a smallest resolvable feature size of 22.1 microns for our scanner compared to 8.8 microns for contact dermoscopy. Results show the scanner consistently outperforms Vectra across most clinically relevant features and achieves comparable performance to contact dermoscopy for the majority of features assessed. By acquiring dermatoscopic-quality images automatically and without contact, and without requiring a separate manual dermoscopic examination, the scanner closes part of the gap between total-body photography and handheld dermoscopy, suggesting potential for future integration into screening workflows.
Chinese Translation
本文概述了构建能够在皮肤镜水平捕捉皮肤病变的全身成像扫描仪的规格和设计方法，该扫描仪使用安装在四个UR10机械手末端执行器上的摄像头。该系统具备视图规划算法，能够适当地选择最佳摄像头位置以获取痣的图像，配备高层控制器以允许机械手同时工作，并且具备碰撞检测器，当机械手与物体或人接触时会停止工作。我们通过对真实患者进行全身扫描来评估该系统，将获取的图像与接触式皮肤镜和现有的全身摄影系统（Vectra）在临床相关的病变特征上进行比较，并使用USAF 1951分辨率目标量化真实的光学分辨率，结果显示我们的扫描仪的最小可分辨特征尺寸为22.1微米，而接触式皮肤镜为8.8微米。结果表明，该扫描仪在大多数临床相关特征上始终优于Vectra，并且在大多数评估特征上与接触式皮肤镜的性能相当。通过自动获取皮肤镜质量的图像且无需接触，也不需要单独的手动皮肤镜检查，该扫描仪缩小了全身摄影与手持皮肤镜之间的部分差距，暗示其在未来整合到筛查工作流程中的潜力。
cs.RO / 28 / 2609.10215
Adaptive Shared Control with Online Bounded-Rational Human Behavior Estimation
基于在线有限理性人类行为估计的自适应共享控制
Abstract
This work considers adaptive shared human-robot control for nonlinear control-affine systems, where the assumption of a fully rational human is relaxed and the robot adapts its assistance to observed boundedly rational human behavior. We use a level-k bounded-rationality model of the two-player game to construct a finite bank of candidate human and robot policies through alternating best-response computations, with the associated value functions and policies approximated using adaptive dynamic programming. During the shared-control interaction, state-transition residuals compare the measured system evolution with the trajectories predicted by the candidate human policies. The residuals are accumulated using a forgetting factor and mapped to a probabilistic human-behavior model over the finite candidate bank. Rather than selecting a single candidate or averaging stored robot policies, the robot computes a distribution-aware one-step best response by minimizing an expected cooperative cost over the complete estimated human behavior distribution. For a quadratic terminal-value approximation and Euler state propagation, this response admits a closed-form solution expressed in terms of the expected human input. The proposed methods are evaluated in simulations of a benchmark nonlinear system stabilization task, and of a planar manipulator shared control setup. The reported results show decreasing Kullback-Leibler divergence between the estimated and simulated human behavior distributions, and a lower accumulated running cost for the robot agent over the shared control interaction period, than the maximum-probability and probability-weighted alternative policies baseline.
Chinese Translation
本研究考虑了针对非线性控制仿射系统的自适应共享人机控制，其中放宽了完全理性人类的假设，机器人根据观察到的有限理性人类行为调整其辅助策略。我们使用二人博弈的水平-k 有限理性模型，通过交替的最佳响应计算构建有限的候选人类和机器人策略库，并利用自适应动态规划对相关的价值函数和策略进行近似。在共享控制交互过程中，状态转换残差将测量的系统演变与候选人类策略预测的轨迹进行比较。残差通过遗忘因子进行累积，并映射到有限候选库上的概率人类行为模型。机器人并不是选择单一候选策略或对存储的机器人策略进行平均，而是通过最小化整个估计人类行为分布下的期望合作成本，计算一个分布感知的一步最佳响应。对于二次终端值近似和欧拉状态传播，该响应具有以期望人类输入表示的封闭形式解。所提出的方法在基准非线性系统稳定化任务和一个平面操纵器共享控制设置的仿真中进行了评估。报告的结果显示，估计的人类行为分布与模拟的人类行为分布之间的Kullback-Leibler散度逐渐减小，并且在共享控制交互期间，机器人代理的累计运行成本低于最大概率和概率加权替代策略基准。
cs.RO / 29 / 2609.10230
CougarTail & CUB: A General-Purpose Mast and Central Utility Board for Cylindrical Underwater Enclosures
CougarTail与CUB：用于圆柱形水下密闭容器的通用支架和中央电源板
Abstract
Cylindrical watertight enclosures are widely used across various underwater systems, from unmanned underwater vehicles (UUVs), to remotely operated vehicles (ROVs), to various sensor platforms. However, electronics are typically built on rectangular PCBs arranged in horizontal stacks, which inefficiently occupy the circular cross-section volume that is critical for both payload capacity and buoyancy management. This paper presents CUB (Central Utility Board) and CougarTail, a general-purpose system designed to address this gap. CUB is a circular PCB sized for 4-inch-diameter enclosures that consolidates a Raspberry Pi Compute Module 5 (CM5) and an STM32 microcontroller, while also providing power management features and auxiliary connections. Mounted coaxially, CUB reduces the electronics stack of our CougUV from 200 mm of tube length and 709 g to 25 mm and 156 g, returning that length and mass budget to payload and buoyancy trim. CougarTail is an open-source companion sensor mast that houses a GPS antenna and two dual-band (2.4 and 5 GHz) omnidirectional PCB antennas. Both components are validated through bench testing and integration on a CougUV platform, our small open-sourced torpedo UUVs.
Chinese Translation
圆柱形防水密闭容器广泛应用于各种水下系统，从无人水下航行器（UUV）到遥控水下航行器（ROV），再到各种传感器平台。然而，电子设备通常基于矩形印刷电路板（PCB）构建，并以水平堆叠的方式排列，这种布局在占用对于有效载荷能力和浮力管理至关重要的圆形横截面体积方面效率较低。本文提出了CUB（中央电源板）和CougarTail，这是一种旨在填补这一空白的通用系统。CUB是一个直径为4英寸的圆形PCB，整合了Raspberry Pi计算模块5（CM5）和STM32微控制器，同时提供电源管理功能和辅助连接。CUB以同轴方式安装，减少了我们CougUV的电子堆叠长度，从200毫米和709克减少到25毫米和156克，将这部分长度和质量预算返还给有效载荷和浮力调整。CougarTail是一个开源的伴随传感器支架，内置GPS天线和两个双频（2.4和5 GHz）全向PCB天线。两个组件通过基准测试和在CougUV平台上的集成得到了验证，CougUV是我们的小型开源鱼雷UUV。
cs.RO / 30 / 2609.10243
FolDeX: A Physical-World Benchmark for Long-Horizon Robotic Manipulation of Deformable Objects
FolDeX：用于长时间跨度可变形物体机器人操作的物理世界基准
Abstract
Embodied AI, including vision-language-action and world-action models, must operate reliably in the physical world. Yet methods that perform well in simulation can degrade substantially on real robots, especially in long-horizon deformable-object manipulation, where policies must track changing states and execute reliable multi-stage bimanual interactions. Existing real-robot benchmarks mainly focus on short-horizon rigid-object tasks and offer limited coverage of long-horizon deformable manipulation. We introduce FolDeX, a physical-world benchmark built entirely from real-robot data, with garment folding as its primary task. Since real-robot data collection is costly, FolDeX studies how heterogeneous physical experience can be reused efficiently. The benchmark is organized around four research axes: leveraging human intervention and recovery data collected during deployment; transferring data across tasks, including across garment categories and from rigid to deformable-object manipulation; reusing data across scenes with changes in lighting, background, and layout; and transferring data across robotic embodiments. FolDeX provides 2,000+ hours of real-robot data spanning 20+ tasks and 10+ embodiments. We also establish a fair real-robot evaluation platform for externally submitted policies, with standardized tasks, held-out physical objects, controlled initializations, and a unified execution protocol. The platform is publicly accessible at https://ai.midea.com/#/fold-challenge. We hope FolDeX will serve as a unified testbed for heterogeneous real-robot data reuse and reliable long-horizon deformable manipulation.
Chinese Translation
具身人工智能，包括视觉-语言-行动和世界-行动模型，必须在物理世界中可靠地运行。然而，在模拟中表现良好的方法在真实机器人上可能会显著下降，特别是在长时间跨度的可变形物体操作中，策略必须跟踪变化的状态并执行可靠的多阶段双手交互。现有的真实机器人基准主要集中在短时间跨度的刚性物体任务上，对长时间跨度的可变形操作覆盖有限。我们引入了FolDeX，这是一个完全基于真实机器人数据构建的物理世界基准，主要任务是衣物折叠。由于真实机器人数据收集成本高，FolDeX研究如何高效地重用异质物理经验。该基准围绕四个研究轴心组织：利用在部署过程中收集的人类干预和恢复数据；跨任务转移数据，包括跨衣物类别和从刚性到可变形物体操作的转移；在照明、背景和布局变化的场景中重用数据；以及跨机器人体现转移数据。FolDeX提供了超过2000小时的真实机器人数据，涵盖20多个任务和10多个体现。我们还建立了一个公平的真实机器人评估平台，供外部提交的策略使用，具有标准化任务、保留的物理对象、受控初始化和统一的执行协议。该平台可公开访问，网址为https://ai.midea.com/#/fold-challenge。我们希望FolDeX能作为异质真实机器人数据重用和可靠长时间跨度可变形操作的统一测试平台。
cs.RO / 31 / 2609.10283
SwingBot: Learning Whole-Body Brachiation for Humanoid Robots
SwingBot：为类人机器人学习全身摆荡
Abstract
Brachiation enables primates to move across overhead supports when ground paths are blocked, suggesting a complementary locomotion mode for robots operating in cluttered or hazardous environments. Bringing this capabil?ity to high-DoF humanoid robots is difficult because the controller must discover a long-horizon release-swing-capture sequence, coordinate alternating contacts with whole-body momentum, and act without reliable measurements of segment?relative displacement or hook-contact state. We present SwingBot, a learning framework for continuous humanoid brachiation with passive wrist hooks. Swing?Bot makes the task trainable by organizing learning around the structure of brachi?ation: biomimetic keyframes make rare release-swing-capture transitions reach?able during early exploration, and recurrent privileged-state estimation provides compact position and contact latents for deployment. Hardware experiments demonstrate continuous bar traversal and robustness to payload, external distur?bances and different bar spacings, showing that this formulation offers a practical route to whole-body robotic brachiation.
Chinese Translation
摆荡使得灵长类动物能够在地面路径被阻塞时，通过悬挂的支撑物移动，这为在杂乱或危险环境中操作的机器人提供了一种补充的运动模式。将这一能力引入高自由度（high-DoF）类人机器人是困难的，因为控制器必须发现一个长时间跨度的释放-摆动-捕捉序列，协调全身动量下的交替接触，并在缺乏可靠的关节相对位移或钩接触状态测量的情况下进行动作。我们提出了SwingBot，一个用于具有被动腕钩的连续类人摆荡的学习框架。SwingBot通过围绕摆荡的结构组织学习，使得任务可训练：生物仿生关键帧使得稀有的释放-摆动-捕捉过渡在早期探索中可达成，而递归特权状态估计为部署提供了紧凑的位置和接触潜变量。硬件实验展示了连续的横杆穿越能力，并对负载、外部干扰和不同横杆间距表现出鲁棒性，表明这一方法为全身机器人摆荡提供了一条实用的途径。
cs.RO / 32 / 2609.10286
Learning Terrain-Adaptive Humanoid Locomotion on Granular Terrain
在颗粒地形上学习适应性人形机器人运动
Abstract
Humanoid locomotion on granular terrain remains a significant challenge due to its complex foot-terrain interaction dynamics that are difficult to model. Existing approaches either ignore granular contact dynamics or incorporate simplified normal force models with heuristic tangential components. In this work, we present a physics-grounded granular contact model based on three-dimensional resistive force theory (3D RFT) and efficiently simulate granular terrain for reinforcement learning (RL) training. Unlike traditional rigid contact models and simplified granular contact models with ad-hoc heuristics, our contact solver produces physically accurate granular intrusion dynamics without resorting to heuristics. It captures realistic penetration and tangential drag during training, enabling the policy to learn behaviors that transfer reliably to real-world granular terrain where rigid contact models fail. To adapt to varying terrain conditions, we train a terrain-adaptive locomotion controller via teacher-student RL, using a variational autoencoder to encode terrain information into a compact latent representation. Simulation studies using material point method (MPM) with NVIDIA Newton demonstrate that our method generalizes to unseen granular terrains, achieves a significantly higher success rate than baselines, and demonstrates zero-shot terrain identification and adaptation. We further validate our approach through extensive hardware experiments across diverse real-world granular terrains including basalt, dry sand, and beach sand. To the best of our knowledge, this is the first demonstration of agile humanoid locomotion on real-world granular terrain. Project page: https://humanoid-gm-locomotion.github.io/HUMANOID-GM/
Chinese Translation
在颗粒地形上进行人形机器人运动仍然是一个重大挑战，因为其复杂的足部与地形的相互作用动态难以建模。现有的方法要么忽略颗粒接触动态，要么采用简化的法向力模型并结合启发式的切向分量。在本研究中，我们提出了一种基于三维阻力理论（3D RFT）的物理基础颗粒接触模型，并有效地模拟颗粒地形以进行强化学习（RL）训练。与传统的刚性接触模型和带有临时启发式的简化颗粒接触模型不同，我们的接触求解器在不依赖启发式的情况下产生物理上准确的颗粒侵入动态。它在训练过程中捕捉到真实的穿透和切向拖拽，使得策略能够学习到在现实世界的颗粒地形中可靠转移的行为，而在刚性接触模型失效的情况下表现良好。为了适应不同的地形条件，我们通过教师-学生强化学习训练一个适应性运动控制器，使用变分自编码器将地形信息编码为紧凑的潜在表示。使用NVIDIA Newton的材料点法（MPM）进行的仿真研究表明，我们的方法能够推广到未见过的颗粒地形，成功率显著高于基线，并展示了零-shot地形识别和适应能力。我们还通过在包括玄武岩、干沙和海滩沙等多种真实颗粒地形上的广泛硬件实验验证了我们的方法。根据我们所知，这是首次在真实颗粒地形上展示灵活的人形机器人运动。项目页面：https://humanoid-gm-locomotion.github.io/HUMANOID-GM/
cs.RO / 33 / 2609.10308
Deformable Object Manipulation under Partial Observability via Real-Time Full-Shape Estimation
基于实时全形状估计的部分可观测下可变形物体操控
Abstract
Manipulating deformable objects (DOs) is challenging due to their high-dimensional state space, underactuated dynamics, and partial observability. In this paper, we propose cRVAE, a lightweight conditional recurrent variational autoencoder that estimates the full DO state from only partial corner-node observations during inference. The resulting model is used as the forward model in a receding-horizon optimal control framework for obstacle-aware collaborative DO manipulation. In simulation on rope and fabric, cRVAE estimates the full DO state from the available corner-node measurements alone, matching the accuracy of a parameter-identified XPBD model. At inference it uses no physical parameters as model inputs and performs no online parameter identification. It also runs approximately 350 times faster on the rope and over 1500 times faster on the fabric per forward pass, keeping horizon-based planning within the 100 ms control budget where XPBD exceeds it already at short horizons. Full-shape estimation from corner sensing at in-loop speed is what makes the model deployable on hardware, which we demonstrate on a Unitree Go2 robot.
Chinese Translation
操控可变形物体（DOs）具有挑战性，因为它们具有高维状态空间、欠驱动动态和部分可观测性。本文提出了一种轻量级条件递归变分自编码器（cRVAE），该模型在推理过程中仅通过部分角节点观测来估计完整的DO状态。所得到的模型被用作回退视界最优控制框架中的前向模型，以实现障碍感知的协作DO操控。在对绳索和织物的仿真中，cRVAE仅通过可用的角节点测量估计完整的DO状态，其准确性与参数识别的XPBD模型相匹配。在推理过程中，它不使用物理参数作为模型输入，也不进行在线参数识别。它的前向传递在绳索上运行速度约为350倍，在织物上超过1500倍，保持在100毫秒的控制预算内进行基于视界的规划，而XPBD在短视界内已超出该限制。通过角传感器以循环速度进行的全形状估计使得该模型能够在硬件上部署，我们在Unitree Go2机器人上进行了验证。
cs.RO / 34 / 2609.10336
Odometer-Agnostic Drift Correction Using OpenStreetMap Lane Geometry
与里程计无关的漂移校正方法：基于OpenStreetMap车道几何
Abstract
Despite significant progress in odometry estimation, long-term drift remains a fundamental limitation of incremental pose integration, especially in large-scale or loop-free environments. Existing map-assisted methods can reduce drift, but often depend on dense maps, sensor-specific processing, or complex matching pipelines. We propose a lightweight open-source, odometry-agnostic correction method that aligns short trajectory segments to OpenStreetMap (OSM) lane centerlines. By formulating drift correction as a direct alignment between recent odometry and sparse lane geometry, the method enables efficient online operation without dense priors or expensive preprocessing. Experiments with LiDAR and visual odometry backends demonstrate consistent improvements, with particularly strong gains under severe drift.
Chinese Translation
尽管在里程计估计方面取得了显著进展，但长期漂移仍然是增量位姿集成的一个基本限制，尤其是在大规模或无环环境中。现有的地图辅助方法可以减少漂移，但通常依赖于密集地图、特定传感器的处理或复杂的匹配流程。我们提出了一种轻量级的开源、与里程计无关的校正方法，该方法将短轨迹段与OpenStreetMap (OSM) 车道中心线对齐。通过将漂移校正公式化为近期里程计与稀疏车道几何之间的直接对齐，该方法实现了高效的在线操作，而无需密集先验或昂贵的预处理。使用激光雷达和视觉里程计后端的实验表明，方法在严重漂移情况下表现出一致的改进，特别是在漂移严重时获得了显著提升。
cs.RO / 35 / 2609.10339
A Confidence-Aware Multimodal Fusion Framework for Industrial Human-Robot Collaboration
一种基于信心的多模态融合框架用于工业人机协作
Abstract
A confidence-aware multimodal fusion framework (CAMF) is proposed to realize reliable human intention prediction for industrial human-robot collaboration. This framework fuses four heterogeneous modalities including object 6D pose, gaze, skeletal motion and IMU-based hand motion. It embeds a confidence-trend-driven dynamic fusion mechanism into BiLSTM to adaptively balance bidirectional temporal features according to real-time modality reliability. A confidence-guided balanced learning strategy combined with a confidence freezing mechanism is further adopted to adjust network gradients dynamically, suppress noise from low-quality modalities and mitigate cross-modal learning bias. A physical platform based on the UR3 collaborative robot is built for experimental validation. Comparative results show that the proposed method reaches an intention recognition accuracy of 91.86% and outperforms existing multimodal fusion approaches in overall performance and stability. It also maintains satisfactory accuracy under low light and partial occlusion interference. In practical assembly tasks, the framework enables proactive and stable human-robot cooperation with strong environmental adaptability.
Chinese Translation
本文提出了一种基于信心的多模态融合框架（CAMF），旨在实现可靠的人类意图预测，以促进工业人机协作。该框架融合了四种异构模态，包括物体的六维姿态（6D pose）、注视（gaze）、骨骼运动（skeletal motion）和基于惯性测量单元（IMU）的手部运动。它将一个基于信心趋势的动态融合机制嵌入到双向长短期记忆网络（BiLSTM）中，以根据实时模态的可靠性自适应地平衡双向时间特征。进一步采用了一种基于信心的平衡学习策略，并结合信心冻结机制，动态调整网络梯度，抑制低质量模态的噪声，并减轻跨模态学习偏差。基于UR3协作机器人构建了一个物理平台进行实验验证。比较结果表明，所提出的方法在意图识别准确率上达到了91.86%，并在整体性能和稳定性上优于现有的多模态融合方法。该框架在低光照和部分遮挡干扰下也能保持令人满意的准确性。在实际装配任务中，该框架实现了主动且稳定的人机合作，具有较强的环境适应性。
cs.RO / 36 / 2609.10377
Data-Driven Risk Fields for Safer End-to-End Autonomous Driving
数据驱动的风险场以实现更安全的端到端自动驾驶
Abstract
Safety is a fundamental requirement for autonomous driving, yet existing end-to-end driving models still lack explicit risk-aware learning capacities. Existing rule-based risk models provide interpretable safety priors, yet their absolute risk scores depend on handcrafted functions, coefficients, and thresholds. Learning-based risk representations reduce part of this manual design, but their supervision often relies on occupancy-derived labels or heuristic cost values, which may not capture ego-conditioned planning risk. In this paper, we propose DRiF, a data-driven risk-field framework for safer end-to-end autonomous driving. DRiF learns a shared BEV feature with static map segmentation, dynamic risk prediction, and vehicle planning. For dynamic risk learning, DRiF converts rule-based safety priors into pairwise risk labels, and trains the risk field to preserve relative risk ordering instead of regressing handcrafted absolute scores. Experiments on Bench2Drive show that DRiF achieves competitive overall performance, with consistent improvements in driving score, success rate, and collision-related metrics. These results establish relative risk supervision as an effective way to connect explicit safety structure with end-to-end planning. The data and code will be publicly available.
Chinese Translation
安全性是自动驾驶的基本要求，但现有的端到端驾驶模型仍然缺乏明确的风险感知学习能力。现有的基于规则的风险模型提供了可解释的安全先验，但其绝对风险评分依赖于手工设计的函数、系数和阈值。基于学习的风险表示减少了部分手动设计，但其监督通常依赖于基于占用的标签或启发式成本值，这可能无法捕捉到自我条件规划的风险。在本文中，我们提出了DRiF，一个数据驱动的风险场框架，以实现更安全的端到端自动驾驶。DRiF通过静态地图分割、动态风险预测和车辆规划学习共享的鸟瞰视图（BEV）特征。对于动态风险学习，DRiF将基于规则的安全先验转换为成对的风险标签，并训练风险场以保持相对风险排序，而不是回归手工设计的绝对评分。在Bench2Drive上的实验表明，DRiF实现了具有竞争力的整体性能，在驾驶评分、成功率和与碰撞相关的指标上均有一致的改善。这些结果确立了相对风险监督作为将明确的安全结构与端到端规划连接的有效方式。数据和代码将公开提供。
cs.RO / 37 / 2609.10400
A traffic management system for large and heterogeneous vehicles in narrow industrial environments
窄小工业环境中大型异构车辆的交通管理系统
Abstract
The coordination of Automated Guided Vehicles (AGVs) in high-density industrial environments represents a critical challenge within Logistics 4.0, as traditional traffic management methods often lead to inefficiencies caused by negotiation-based priority assignment. To overcome the resulting limitations, this paper presents an innovative AGV traffic management system based on a Lifelong Multi-Agent Path Finding (L-MAPF) algorithm operating on roadmaps generated with Non-Uniform Rational B-Splines (NURBS) curves. The approach guarantees locally optimal coordination and ensures safe operation of large and heterogeneous AGVs. Building on this concept, the proposed framework integrates a modified version of the Bounded Horizon Conflict Based Search (CBS) technique within a Rolling Horizon Conflict Resolution strategy, utilizing an extended time horizon for each agent to enable effective conflict resolution in corridors identified by a topological map. In contrast to state-of-the-art methods for AGV fleet traffic management, the proposed solution is designed for real-world, non-standardized (i.e., non-grid-like) industrial settings characterized by narrow bidirectional corridors and high-traffic density, where AGVs of various sizes and capabilities operate simultaneously. Key contributions include an anytime conflict resolution strategy with adaptive time horizon regulation, an execution layer for safe and standard-compliant interaction with real AGVs, and an advanced mechanism for deadlock detection and resolution. Experimental results obtained in realistic industrial environments demonstrate higher throughput, with improvements of up to 11% over a conventional rule-based traffic management system, a state-of-the-art industrial method, and a priority-based L-MAPF variant, while maintaining continuous operation and improved efficiency.
Chinese Translation
在高密度工业环境中，自动导引车（AGVs）的协调是物流4.0中的一个关键挑战，因为传统的交通管理方法往往导致基于协商的优先级分配造成的低效。为克服这些限制，本文提出了一种基于终身多智能体路径规划（L-MAPF）算法的创新AGV交通管理系统，该算法在使用非均匀有理B样条（NURBS）曲线生成的路网图上运行。该方法保证了局部最优协调，并确保大型异构AGVs的安全运行。在此基础上，所提框架在滚动地平线冲突解决策略中集成了修改版的有界地平线冲突基础搜索（CBS）技术，为每个智能体利用扩展的时间地平线以实现有效的冲突解决，特别是在由拓扑图识别的走廊中。与AGV车队交通管理的最先进方法相比，所提解决方案旨在应对现实世界中非标准化（即非网格状）工业环境的挑战，这些环境的特点是狭窄的双向走廊和高交通密度，其中各种尺寸和能力的AGVs同时运行。主要贡献包括一种具有自适应时间地平线调节的随时冲突解决策略、一个安全且符合标准的与真实AGVs交互的执行层，以及一个先进的死锁检测与解决机制。在现实工业环境中获得的实验结果表明，所提方法的吞吐量更高，相较于传统的基于规则的交通管理系统、最先进的工业方法以及基于优先级的L-MAPF变体，提升幅度可达11%，同时保持连续运行和提高效率。
cs.RO / 38 / 2609.10405
Frequency-Conditioned Flow Matching for Vision-Language-Action Models
频率条件流匹配用于视觉-语言-动作模型
Abstract
Robot actions are temporally correlated trajectories whose frequency components encode motion at different scales with highly non-uniform energy distributions. Yet Flow Matching--based vision-language-action (VLA) models typically generate actions in temporal coordinates, without explicitly modeling or systematically leveraging this frequency heterogeneity. We introduce \emph{FreqFM}, a frequency-conditioned Flow Matching framework for VLA models. It raises action frequency from an implicit trajectory property to an explicit conditioning dimension that spans the entire generation pipeline. Concretely, in DCT frequency coordinates, FreqFM constructs a spectrum-matched source distribution, adaptively balances the objective across frequencies, and constrains per-frequency guidance residuals using the corresponding reference transport scales. FreqFM integrates into existing Flow Matching action experts without changing the VLA backbone. Across LIBERO, LIBERO-Plus, and VLA-Arena, FreqFM consistently improves performance, including a 9.3-point gain on LIBERO-Plus, and further demonstrates its effectiveness on six real-robot tasks.
Chinese Translation
机器人动作是时间上相关的轨迹，其频率成分在不同尺度上编码运动，并具有高度不均匀的能量分布。然而，基于流匹配（Flow Matching）的视觉-语言-动作（VLA）模型通常在时间坐标中生成动作，而没有明确建模或系统性利用这种频率异质性。我们提出了 extit{FreqFM}，一种用于VLA模型的频率条件流匹配框架。它将动作频率从隐式轨迹属性提升为一个显式的条件维度，贯穿整个生成流程。具体而言，在离散余弦变换（DCT）频率坐标中，FreqFM构建了一个频谱匹配的源分布，自适应地平衡各频率间的目标，并使用相应的参考传输尺度约束每个频率的引导残差。FreqFM可以无缝集成到现有的流匹配动作专家中，而无需更改VLA的主干网络。在LIBERO、LIBERO-Plus和VLA-Arena上，FreqFM始终提高性能，包括在LIBERO-Plus上获得9.3分的提升，并进一步证明其在六个真实机器人任务上的有效性。
cs.RO / 39 / 2609.10433
Multi-Agent Reinforcement Learning for Autonomous UAV Exploration in Wildfire Response
基于多智能体强化学习的无人机自主探索野火响应
Abstract
This study develops a deep reinforcement learning framework for training Unmanned Aerial Vehicle (UAV) agents to navigate and monitor simulated wildfire environments. Results show that agents learn increasingly stable and effective behaviors over time, as demonstrated by converging loss trends, improved reward signals, and more consistent navigation patterns such as fire-boundary tracking. Overall, these findings highlight the potential of deep reinforcement learning (DRL) based UAV systems for autonomous wildfire monitoring and suggest that environmental structure and reward design influence policy effectiveness.
Chinese Translation
本研究开发了一种深度强化学习框架，用于训练无人机（UAV）代理在模拟野火环境中进行导航和监测。结果表明，代理随着时间的推移学习到越来越稳定和有效的行为，这通过收敛的损失趋势、改善的奖励信号以及更一致的导航模式（如火界跟踪）得以证明。总体而言，这些发现突显了基于深度强化学习（DRL）的无人机系统在自主野火监测中的潜力，并表明环境结构和奖励设计对策略有效性具有影响。
cs.RO / 40 / 2609.10484
Coastal Environment Generation with HoloOcean
基于 HoloOcean 的海岸环境生成
Abstract
Marine robotic simulation provides a safe and inexpensive method of developing and testing algorithms for unmanned underwater vehicle (UUV) and unmanned surface vessel (USV) autonomy and perception before full field deployment. However, these simulations are often limited by the availability of simulated environments. Current marine robotics simulation suites offer manual ways to edit or create environments, but they require existing data or specialized knowledge of the environment system. To address these issues, we introduce a novel Unreal Engine 5 level generation pipeline that enables automatic creation of coastal environments for HoloOcean. Our pipeline relies on a user-provided overhead image of a coastal scene. The pipeline then uses the image to generate height map data, as well as automatically select assets and place them in the environment.
Chinese Translation
海洋机器人仿真提供了一种安全且经济的方法，用于在全面部署之前开发和测试无人水下航行器（UUV）和无人水面船（USV）的自主性和感知算法。然而，这些仿真通常受到可用仿真环境的限制。目前的海洋机器人仿真套件提供了手动编辑或创建环境的方法，但这些方法需要现有数据或对环境系统的专业知识。为了解决这些问题，我们提出了一种新颖的虚幻引擎 5 级别生成管道，能够自动创建 HoloOcean 的海岸环境。我们的管道依赖于用户提供的海岸场景的俯视图像。该管道随后利用图像生成高度图数据，并自动选择资产并将其放置在环境中。
cs.RO / 41 / 2609.10506
DUET-DINO: Simultaneous Cross-View World Modeling for Latent Planning in Robot Manipulation
DUET-DINO：用于机器人操作的潜在规划的同时跨视图世界建模
Abstract
Action-conditioned latent world models predict future visual representations, enabling zero-shot goal-conditioned robot planning and control. However, their predictions for fine-grained spatial and rotational actions are unreliable for full 7-DoF end-effector control. To address this gap, we introduce DUET-DINO, a simultaneous cross-view latent world model that jointly learns action-conditioned predictions from static side- and wrist-camera observations through cross-view conditioning. By exploiting complementary global scene and gripper-centric information, DUET-DINO enables latent planning over the full 7-DoF action space. Across spatially diverse reach, orientation-intensive angled-reach, and multi-goal grasp-and-lift tasks, DUET-DINO consistently outperforms single-view and independent dual-view baselines, achieving 92% success on reach, 72.5% on angled-reach, and 60.0% on lift tasks. DUET-DINO is trained from scratch on DROID and RoboArena datasets and generalizes robustly under visual distribution shifts. We further show that while V-JEPA 2 wrist-view predictions underestimate visual dynamics induced by fine-grained actions, DINOv3 predictions better capture action-conditioned scene changes, leading to stronger downstream planning. The code and model checkpoints will be open-sourced. Project page: https://utn-air.github.io/DUET-DINO
Chinese Translation
基于动作的潜在世界模型预测未来的视觉表示，从而实现零-shot目标条件的机器人规划和控制。然而，它们对于细粒度空间和旋转动作的预测在全7自由度末端执行器控制中是不可靠的。为了解决这一问题，我们提出了DUET-DINO，这是一种同时跨视图的潜在世界模型，通过跨视图条件共同学习来自静态侧视和腕部摄像头观察的基于动作的预测。通过利用互补的全局场景和夹持器中心信息，DUET-DINO能够在全7自由度动作空间上进行潜在规划。在空间多样的到达、方向密集的角度到达和多目标抓取与提升任务中，DUET-DINO始终优于单视图和独立双视图基线，在到达任务上取得92%的成功率，在角度到达任务上取得72.5%的成功率，在提升任务上取得60.0%的成功率。DUET-DINO在DROID和RoboArena数据集上从头训练，并在视觉分布变化下具有良好的泛化能力。我们进一步表明，尽管V-JEPA 2的腕部视图预测低估了细粒度动作引起的视觉动态，但DINOv3的预测更好地捕捉了基于动作的场景变化，从而导致更强的下游规划。代码和模型检查点将开源。项目页面：https://utn-air.github.io/DUET-DINO
cs.RO / 42 / 2609.10522
Show-Harness: Just a VLM Agent Can Play Robots
Show-Harness：仅凭 VLM 代理即可操控机器人
Abstract
Foundation vision-language models (VLMs) exhibit broad intelligence about the world, yet translating this intelligence into robot control remains challenging. We present Show-Harness, an Embodied Harness that enables VLMs to "play" robots through a compact semantic interface linking intent to action. Show-Harness exposes discrete semantic action units that VLMs can naturally reason over, while embodiment-specific interpreters deterministically ground them into local robot actions, keeping the VLM directly responsible for fine-grained physical decisions. Through the same interface, Show-Harness demonstrates the feasibility of (1) directly unlocking closed-source frontier VLMs for zero-shot robot control, and (2) adapting small-scale open-source VLMs for low-cost deployment with just a few GPU-hours of fine-tuning. We further develop GUMI (GUI Manipulation Interface), which extends the same semantic action space to GUI-based demonstration collection, allowing humans and agents to "play" robots across embodiments without specialized teleoperation hardware. Extensive experiments show that Show-Harness-equipped VLM agents generalize robustly across tasks, embodiments, and environments, outperforming representative agentic and VLA paradigms. These results suggest that the right interface can unlock substantial embodied capability from foundation VLMs, without requiring additional model capacity or costly embodiment-specific pretraining.
Chinese Translation
基础视觉-语言模型（VLMs）展现了对世界的广泛智能，但将这种智能转化为机器人控制仍然具有挑战性。我们提出了 Show-Harness，一种具身化的 Harness，使 VLMs 能够通过一个紧凑的语义接口将意图与行动连接，从而“操控”机器人。Show-Harness 公开了离散的语义行动单元，VLMs 可以自然地对其进行推理，而具身特定的解释器则将其确定性地映射为本地机器人动作，使 VLM 直接负责细致的物理决策。通过相同的接口，Show-Harness 展示了（1）直接解锁闭源前沿 VLMs 进行零-shot 机器人控制的可行性，以及（2）仅需少量 GPU 小时的微调即可将小规模开源 VLMs 适配为低成本部署。我们进一步开发了 GUMI（GUI Manipulation Interface），该接口将相同的语义行动空间扩展到基于 GUI 的演示收集，使人类和代理能够在不同的具身化中“操控”机器人，而无需专门的遥操作硬件。大量实验表明，配备 Show-Harness 的 VLM 代理在任务、具身化和环境之间具有良好的泛化能力，超越了代表性的代理和 VLA 范式。这些结果表明，合适的接口可以从基础 VLMs 中释放出显著的具身能力，而无需额外的模型容量或昂贵的具身特定预训练。
cs.CV / 1 / 2609.09184
Evidence-Order Calibration for Selective Visual Reasoning under Progressive Loss of Question-Critical Evidence
在逐步丧失问题关键证据下的选择性视觉推理的证据顺序校准
Abstract
Vision-language model (VLM) confidence may change in aggregate when visual evidence is degraded while remaining structurally inconsistent within individual examples. We study answer-level reliability along five-step, question-conditioned evidence-loss trajectories. Using a frozen Qwen2.5-VL-3B-Instruct model, we construct 176 accepted GQA-derived trajectories (880 masking conditions) by progressively masking scene-graph-localized question-critical regions. Native sequence confidence has an evidence monotonicity violation rate (EMVR) of 0.436, and 92.0% of trajectories contain at least one adjacent violation. A matched non-critical-region control shows that full critical masking reduces accuracy by 28.2 percentage points, compared with 0.6 points for equally sized non-critical masks; the paired difference is 27.6 points (95% CI [20.0, 34.7]). We train a lightweight post-hoc reliability head on frozen hidden states, sequence confidence, and entropy. Adding evidence-order supervision to binary cross-entropy (BCE) reduces masking EMVR from 0.330 to 0.303 (paired difference -0.027, 95% CI [-0.044, -0.010]). The same mask-trained objective reduces EMVR from 0.449 to 0.402 on held-out question IDs under unseen local Gaussian blur (difference -0.0468, 95% CI [-0.0739, -0.0199]). AUROC, Brier, and AURC differences between the two learned heads are statistically inconclusive, and native confidence remains stronger for selective-risk ranking. The results separate evidence-order consistency from conventional correctness discrimination rather than establishing generic confidence superiority.
Chinese Translation
视觉语言模型（VLM）的置信度在视觉证据退化时可能会整体变化，但在个别示例中仍然结构不一致。我们研究沿着五步、问题条件的证据丧失轨迹的答案级可靠性。使用冻结的 Qwen2.5-VL-3B-Instruct 模型，我们通过逐步遮蔽场景图定位的关键问题区域构建了 176 条接受的 GQA 派生轨迹（880 种遮蔽条件）。原始序列置信度的证据单调性违反率（EMVR）为 0.436，92.0% 的轨迹至少包含一个相邻的违反。与非关键区域对照组相比，完全遮蔽关键区域的准确率降低了 28.2 个百分点，而同样大小的非关键遮蔽仅降低了 0.6 个百分点；配对差异为 27.6 个百分点（95% 置信区间 [20.0, 34.7]）。我们在冻结的隐藏状态、序列置信度和熵上训练了一个轻量级的事后可靠性头。将证据顺序监督添加到二元交叉熵（BCE）中，将遮蔽 EMVR 从 0.330 降低到 0.303（配对差异 -0.027，95% 置信区间 [-0.044, -0.010]）。在未见的局部高斯模糊下，同样的遮蔽训练目标将 EMVR 从 0.449 降低到 0.402（差异 -0.0468，95% 置信区间 [-0.0739, -0.0199]）。两个学习头之间的 AUROC、Brier 和 AURC 差异在统计上没有结论性，原始置信度在选择性风险排序中仍然更强。结果将证据顺序一致性与传统的正确性区分分开，而不是建立通用的置信度优越性。
cs.CV / 2 / 2609.09185
Integrating Unimodal and Vision-Language Representations in Latent Space for Multi-Label Chest X-Ray Classification
在潜在空间中整合单模态和视觉-语言表示用于多标签胸部X光分类
Abstract
Multi-label chest X-ray classification has attracted considerable attention in recent years, with the effective use of visual representations and clinical semantic knowledge playing an important role. This study proposes a framework that combines unimodal representations from RAD-DINO with vision--language representations from BioViL-T for the classification of 14 labels in the MIMIC-CXR-JPG dataset. The RAD-DINO and BioViL-T embeddings and their combined representation are refined separately in latent space before being normalized and fused across the three branches. In addition to improving classification performance, the study aims to clarify the role of each embedding source and the degree to which they complement one another. Experiments show that RAD-DINO outperforms BioViL-T when used independently, whereas early fusion further improves the results, indicating that the two embedding sources contain complementary information. The best-performing model achieves a mean AUROC of 0.840 and an mAP of 0.467. Ablation analysis shows that hybrid fusion provides consistent and statistically significant improvements over early fusion when each embedding source is refined in latent space, suggesting that fusion effectiveness depends on the quality of the representation supplied by each branch. However, the study has only been evaluated internally on MIMIC-CXR-JPG; its generalizability to data from other healthcare institutions therefore remains to be validated. The source code is available at: https://anonymous.4open.science/r/mimic-report-C210/.
Chinese Translation
近年来，多标签胸部X光分类引起了广泛关注，视觉表示和临床语义知识的有效利用在其中发挥了重要作用。本研究提出了一个框架，结合了来自RAD-DINO的单模态表示和来自BioViL-T的视觉-语言表示，用于对MIMIC-CXR-JPG数据集中14个标签的分类。RAD-DINO和BioViL-T的嵌入及其组合表示在潜在空间中分别进行细化，然后在三个分支之间进行归一化和融合。除了提高分类性能外，本研究还旨在阐明每个嵌入源的作用及其互补程度。实验表明，RAD-DINO在独立使用时优于BioViL-T，而早期融合进一步改善了结果，表明这两个嵌入源包含互补信息。表现最佳的模型达到了0.840的平均AUROC和0.467的mAP。消融分析显示，当每个嵌入源在潜在空间中进行细化时，混合融合相较于早期融合提供了一致且具有统计显著性的改进，这表明融合的有效性依赖于每个分支提供的表示质量。然而，本研究仅在MIMIC-CXR-JPG上进行了内部评估，因此其对其他医疗机构数据的泛化能力仍需验证。源代码可在以下网址获取：https://anonymous.4open.science/r/mimic-report-C210/
cs.CV / 3 / 2609.09186
M2LG-DG: A Multi-modal Local-Global Domain Generalization Framework for Cross-site Major Depressive Disorder Classification
M2LG-DG：一种用于跨站点重性抑郁障碍分类的多模态局部-全局领域泛化框架
Abstract
Classification models based on resting-state functional magnetic resonance imaging (rs-fMRI) often show lower performance at imaging sites not included during model development, which can limit their use in clinical settings. Domain generalization (DG) addresses this issue by learning representations from source sites that remain effective for unseen target sites. However, existing DG approaches for psychiatric disorder classification commonly rely on a single imaging modality and may not fully account for site-specific acquisition effects on the learned representation space. Subjects scanned at the same site share scanner hardware, acquisition settings, and preprocessing characteristics, which can cause representations to reflect acquisition conditions rather than diagnostic information. In this work, we present M2LG-DG, a source-only multimodal local-global framework for cross-site major depressive disorder (MDD) classification. The framework employs a dual-stream rs-fMRI encoder, where the global pathway models inter-regional dependencies through self-attention and the local pathway performs graph-constrained aggregation over functional connectivity-derived brain graphs. Imaging and non-imaging representations are decomposed into shared and private components and integrated through bidirectional cross-attention with a learned modality gate. A cross-site supervised contrastive objective forms positive pairs from same-class subjects acquired at different source sites, encouraging the fused representation to preserve diagnostic information across acquisition domains. On four held-out REST-meta-MDD sites, M2LG-DG achieves an AUC of 69.48% and exceeds the closest comparison method by 2.18 percentage points. Experiments on the Autism Brain Imaging Data Exchange (ABIDE) dataset further support its applicability to other psychiatric neuroimaging classification tasks.
Chinese Translation
基于静息态功能磁共振成像（rs-fMRI）的分类模型在未包含于模型开发中的成像站点上通常表现出较低的性能，这限制了它们在临床环境中的应用。领域泛化（DG）通过学习来自源站点的表示来解决这一问题，使其在未见过的目标站点上仍然有效。然而，现有的精神疾病分类的DG方法通常依赖于单一成像模态，可能未能充分考虑特定站点的采集效应对学习表示空间的影响。在同一站点扫描的受试者共享扫描仪硬件、采集设置和预处理特征，这可能导致表示反映采集条件而非诊断信息。在本研究中，我们提出了M2LG-DG，一种仅基于源的多模态局部-全局框架，用于跨站点重性抑郁障碍（MDD）分类。该框架采用双流rs-fMRI编码器，其中全局路径通过自注意力建模区域间依赖关系，而局部路径在功能连接导出的脑图上执行图约束聚合。成像和非成像表示被分解为共享和私有组件，并通过带有学习模态门的双向交叉注意力进行集成。跨站点监督对比目标从在不同源站点采集的同类受试者中形成正对，鼓励融合表示在采集领域中保持诊断信息。在四个保留的REST-meta-MDD站点上，M2LG-DG达到了69.48%的AUC，并超过了最接近的比较方法2.18个百分点。在自闭症脑成像数据交换（ABIDE）数据集上的实验进一步支持其在其他精神神经影像分类任务中的适用性。
cs.CV / 4 / 2609.09187
AgenticGen: Reward-Guided Agentic Video Generation for Advertising
AgenticGen：基于奖励引导的广告视频生成
Abstract
Advertising video generation is not only a video synthesis task, but also a product-conditioned reasoning problem whose success is measured by online business metrics. Recent video foundation models can generate realistic clips from multimodal conditions, yet they do not optimize how a product should be transformed into an effective advertisement or how future generation should be improved from online business feedback. To close this loop, we propose AgenticGen, a reward-guided agentic framework that decomposes advertising video generation into two trainable reasoning stages, strategy selection and draft generation, thereby exposing optimization targets that online business feedback can supervise. AgenticGen learns a performance-based reward from accumulated online feedback and a complementary rubric-based reward aligned with human quality standards, then uses them to supervise policy optimization. DPO first moves the agentic policies toward online preferences, and GRPO further refines both stages with process and outcome rewards. Offline experiments validate the reward models and successive policy optimization. Online A/B experiments in the TikTok advertising system show that AgenticGen after DPO and GRPO improves CTR by 2.72%, CVR by 2.63%, and Advv by 9.61% over the SFT baseline.
Chinese Translation
广告视频生成不仅是一个视频合成任务，也是一个以产品为条件的推理问题，其成功通过在线商业指标来衡量。近期的视频基础模型能够从多模态条件生成逼真的片段，但它们并未优化产品如何转化为有效广告或如何根据在线商业反馈改进未来生成。为了解决这一问题，我们提出了AgenticGen，一个基于奖励引导的代理框架，将广告视频生成分解为两个可训练的推理阶段：策略选择和草稿生成，从而暴露出在线商业反馈可以监督的优化目标。AgenticGen从累积的在线反馈中学习基于性能的奖励，并与符合人类质量标准的补充评分奖励相结合，然后利用这些奖励来监督策略优化。DPO首先将代理策略朝向在线偏好移动，而GRPO进一步通过过程和结果奖励来细化两个阶段。离线实验验证了奖励模型和后续的策略优化。在TikTok广告系统中的在线A/B实验表明，经过DPO和GRPO的AgenticGen相比于SFT基线，CTR提高了2.72%，CVR提高了2.63%，Advv提高了9.61%。
cs.CV / 5 / 2609.09188
Lensless Gaze Is Not Private by Default: Auditing Identity Leakage Across Disclosure Surfaces
无镜头注视默认并不私密：跨披露表面的身份泄露审计
Abstract
Lensless near-eye sensing is often described as privacy-friendly because its coded measurements are visually unintelligible. Yet visual unintelligibility reflects human interpretation, not what a learned adversary can recover. We therefore treat identity privacy as a systems property of disclosure surfaces: representations crossing sensing, storage, computation, and output boundaries. We audit a simulated lensless gaze pipeline under a 36-subject known-gallery closed-set identification protocol with a fixed, known PSF; privacy from an unknown or varying optical key is outside our scope. Reported accuracies are empirical attack success rates under matched linear and MLP probes and do not upper-bound stronger adversaries. Simulated lensless measurements yield 96.7% top-1 identification versus 97.7% for matched original eye crops, while an MAE embedding retains 94.3%. Compression alone offers little protection: 8-D PCA and a matched 8-D bottleneck retain 93.2% and 91.8%, whereas separately trained 8-D GSPL bottlenecks yield 77.5% mean recovery across three seeds. A released 128-way gaze token lowers single-frame recovery to 38.1%, while its residual and continuous gaze output expose 62.1% and 72.6%, respectively. Under a source-frame-disjoint tiled protocol, token summaries reach 39.9% at T=25, showing that repeated-output risk depends on representation and aggregation. These rates reflect all subject-correlated information in the evaluated dataset, including acquisition and behavioral cues, rather than isolating intrinsic ocular biometrics. Ordinary least squares residualization against a six-dimensional crop geometry and intensity summary still leaves lensless recovery at 95.1%. Our results show that privacy claims for lensless sensing must be tested at disclosure boundaries rather than inferred from appearance.
Chinese Translation
无镜头近眼传感常被描述为隐私友好，因为其编码测量在视觉上不可理解。然而，视觉不可理解性反映的是人类的解读，而不是学习型对手能够恢复的内容。因此，我们将身份隐私视为披露表面的系统属性：跨越感知、存储、计算和输出边界的表示。我们在一个包含36名受试者的已知图库闭集识别协议下审计一个模拟的无镜头注视管道，使用固定的已知点扩散函数（PSF）；来自未知或变化的光学密钥的隐私超出了我们的研究范围。报告的准确率是基于匹配线性和多层感知器（MLP）探测器的经验攻击成功率，并不限制更强对手的能力。模拟的无镜头测量在顶级1识别中达到96.7%，而匹配的原始眼部裁剪为97.7%，同时均方误差（MAE）嵌入保持在94.3%。单独的压缩几乎没有提供保护：8维主成分分析（PCA）和匹配的8维瓶颈分别保持93.2%和91.8%，而单独训练的8维广义稀疏潜在（GSPL）瓶颈在三个种子下的平均恢复率为77.5%。发布的128维注视令牌将单帧恢复率降低到38.1%，而其残差和连续注视输出分别暴露62.1%和72.6%。在源帧不重叠的平铺协议下，令牌摘要在T=25时达到39.9%，显示重复输出风险依赖于表示和聚合。这些比率反映了评估数据集中所有与受试者相关的信息，包括获取和行为线索，而不是孤立的内在眼部生物特征。针对六维裁剪几何和强度摘要的普通最小二乘残差化仍然使无镜头恢复率达到95.1%。我们的结果表明，对于无镜头传感的隐私声明必须在披露边界进行测试，而不是仅仅从外观中推断。
cs.CV / 6 / 2609.09206
MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads
当信息分布在协同头中漂移时，MLLMs 会产生幻觉
Abstract
Multimodal Large Language Models (MLLMs) often struggle with hallucinations, thus hindering their reliable practical applications. Existing attention-based mitigation methods mainly rely on indirect signals (e.g., attention weights) that fail to accurately reflect the actual information shift underlying hallucination generation. In this paper, we propose HEAL, Head-lEvel information disentAnglement and caLibration for identifying and mitigating hallucinations. HEAL first employs causal noise intervention on multi-head outputs to filter out causally redundant heads. Subsequently, it disentangles information distribution within the remaining heads via the counterfactual Difference-in-Differences, categorizing heads into four types. Through analysis, we observe: hallucinations happen when information distribution drifts away from a healthy equilibrium in synergy heads, not strongly correlated with the quantity or strength of modality-specific heads. Motivated by this insight, HEAL injects dynamic information calibration factors into the value vectors of synergy heads, and actively regulates visual-language dependencies, steering the output distribution towards factual evidence. Extensive experiments demonstrate that HEAL effectively reduces hallucinations across multiple MLLMs, offering a simple and interpretable pathway to enhance model trustworthiness.
Chinese Translation
多模态大语言模型（MLLMs）常常面临幻觉问题，从而阻碍其可靠的实际应用。现有的基于注意力的缓解方法主要依赖于间接信号（例如，注意力权重），这些信号无法准确反映幻觉生成背后的实际信息变化。本文提出了 HEAL（头级信息解耦与校准），旨在识别和缓解幻觉。HEAL 首先对多头输出进行因果噪声干预，以过滤掉因果冗余的头。随后，通过反事实的差异中的差异（Difference-in-Differences）解耦剩余头中的信息分布，将头分类为四种类型。通过分析，我们观察到：当信息分布偏离协同头中的健康平衡时，幻觉会发生，而与模态特定头的数量或强度并没有强相关性。基于这一洞察，HEAL 将动态信息校准因子注入协同头的值向量中，并积极调节视觉-语言依赖关系，引导输出分布朝向事实证据。大量实验表明，HEAL 能有效减少多种 MLLMs 中的幻觉，提供了一条简单且可解释的途径来增强模型的可信度。
cs.CV / 7 / 2609.09300
Video-MOPD: Multi-Teacher On-Policy Distillation for Video Understanding
视频-MOPD：用于视频理解的多教师在线蒸馏
Abstract
Video understanding demands a convergence of complementary capabilities across perception, temporal understanding, and complex reasoning, which are difficult to jointly optimize within a single model. We introduce Video-MOPD-8B, an open-weight model dedicated to video understanding tasks. To fundamentally enhance its capabilities, we conduct targeted reinforcement learning (RL) optimization across three core domains: video temporal grounding (VTG), general video comprehension, and video STEM reasoning. We then unify their complementary capabilities via Multi-Teacher On-Policy Distillation (MOPD), which consolidates expert knowledge by supervising student-generated trajectories with routed teacher feedback. We further introduce Reliability-Aware Informative Sampling (RAIS), which selects examples with consistently reliable teacher supervision and large teacher-student performance gaps. Together, these components enable Video-MOPD-8B to achieve coordinated and comprehensive performance gains across diverse video understanding tasks. Extensive experiments on comprehensive benchmarks covering general video understanding, temporal grounding, video reasoning, and video STEM tasks demonstrate that Video-MOPD-8B achieves state-of-the-art performance among existing models at a comparable scale. The trained model weights are available at https://huggingface.co/LandH/Video-MOPD-8B.
Chinese Translation
视频理解需要在感知、时间理解和复杂推理等互补能力之间实现收敛，这在单一模型中难以共同优化。我们提出了Video-MOPD-8B，这是一个专门用于视频理解任务的开放权重模型。为了从根本上增强其能力，我们在视频时间定位（VTG）、一般视频理解和视频STEM推理三个核心领域进行针对性的强化学习（RL）优化。随后，我们通过多教师在线蒸馏（MOPD）统一它们的互补能力，该方法通过使用教师反馈来监督学生生成的轨迹，从而整合专家知识。我们进一步引入了可靠性意识信息采样（RAIS），该方法选择具有一致可靠教师监督和较大教师-学生性能差距的示例。所有这些组件使得Video-MOPD-8B能够在多样化的视频理解任务中实现协调和全面的性能提升。在涵盖一般视频理解、时间定位、视频推理和视频STEM任务的全面基准上进行的大量实验表明，Video-MOPD-8B在现有同类模型中实现了领先的性能，且规模相当。训练好的模型权重可在 https://huggingface.co/LandH/Video-MOPD-8B 获取。
cs.CV / 8 / 2609.09359
DensePol: Dense-Angle Polarization Dataset for Learning-Based Polarimetric Vision
DensePol：用于基于学习的偏振视觉的高密度角度偏振数据集
Abstract
Polarimetric vision is gaining increasing attention because it provides physical cues about scene shape, material, and reflection that are difficult to recover from RGB alone. Recent work has therefore explored predicting polarization directly from conventional RGB images; however, the fidelity of these methods strongly depends on the polarization supervision used for training. Most existing datasets rely on Division-of-Focal-Plane (DoFP) cameras with four spatially interleaved analyzer orientations, which provide limited angular redundancy and introduce interpolation and instantaneous-field-of-view errors. We introduce DensePol, a high-redundancy RGB--polarization dataset based on Division-of-Time (DoT) acquisition, capturing 180 full-resolution analyzer orientations at $1^\circ$ intervals. DensePol contains 2,018 paired RGB--polarization images with the angular measurements and fitting residuals retained. Dense angular sampling substantially improves polarization stability, reducing AoLP deviation from $13.36^\circ$ to $2.21^\circ$. We further introduce a deterministic diffusion-based RGB-to-polarization framework with cyclic AoLP representation and a local DoLP refiner. Experiments demonstrate improved polarization prediction and downstream surface-normal estimation. The dataset and code will be publicly available.
Chinese Translation
偏振视觉正受到越来越多的关注，因为它提供了关于场景形状、材料和反射的物理线索，这些信息仅通过 RGB 难以恢复。因此，最近的研究探索了如何直接从传统的 RGB 图像中预测偏振；然而，这些方法的保真度在很大程度上依赖于用于训练的偏振监督。现有的大多数数据集依赖于具有四个空间交错分析仪方向的焦平面分割（DoFP）相机，这提供了有限的角度冗余，并引入了插值和瞬时视场误差。我们提出了 DensePol，这是一个基于时间分割（DoT）采集的高冗余 RGB-偏振数据集，以 $1^ ext{°}$ 的间隔捕获 180 个全分辨率分析仪方向。DensePol 包含 2,018 对 RGB-偏振图像，保留了角度测量和拟合残差。密集的角度采样显著提高了偏振稳定性，将 AoLP 偏差从 $13.36^ ext{°}$ 降低到 $2.21^ ext{°}$。我们进一步引入了一种基于确定性扩散的 RGB 到偏振框架，采用循环 AoLP 表示和局部 DoLP 精细化器。实验表明偏振预测和后续表面法线估计的改善。数据集和代码将公开发布。
cs.CV / 9 / 2609.09368
The Living Library: Transforming Archival Collections into Conversational Knowledge Systems -- Lessons from the Theodore Roosevelt Presidential Library
活图书馆：将档案收藏转变为对话知识系统——来自西奥多·罗斯福总统图书馆的经验教训
Abstract
We present the Living Library, an end-to-end framework for transforming fragmented digital archives into governed, conversational, in-person exhibit experiences. Developed and deployed at the Theodore Roosevelt Presidential Library, the framework comprises four layers: digitization and corpus creation, AI-powered processing, retrieval and reasoning, and an optional embodied conversational interface. The first three layers aggregate a 300,000-record collection, apply OCR and structured metadata enrichment for expert curatorial review, and publish records to a hybrid dense/semantic index. Expert review is conducted through the Archivist App, a curator-facing interface that supports correction of AI-generated transcriptions and metadata. The governed corpus powers both a researcher-facing interface and Talk to TR, a continuously operating exhibit that embodies Theodore Roosevelt as a full-scale digital human within a museum environment. To support live, face-to-face interactions, Cross-Era Analogical Grounding reframes contemporary questions through documented historical parallels, allowing Roosevelt to address present-day topics without inventing facts. Dual-path retrieval and end-to-end streaming keep responses grounded and responsive. Layered watchdogs, visitor-session isolation, automated conversation management, and independently restartable services enable reliable unattended operation for hundreds of visitors. Avatar realism, spatial audio, lighting, staging, and conversational design are developed and evaluated as an integrated experience. Rather than report a controlled benchmark, we describe lessons from operating Talk to TR as a public exhibit and offer a transferable model for transforming archival collections into believable, in-person conversational experiences.
Chinese Translation
我们提出了活图书馆，这是一个将分散的数字档案转变为受管控的、对话式的面对面展览体验的端到端框架。该框架在西奥多·罗斯福总统图书馆开发和部署，包含四个层次：数字化和语料库创建、人工智能驱动的处理、检索与推理，以及一个可选的具身对话界面。前三个层次聚合了30万条记录，应用光学字符识别（OCR）和结构化元数据增强以供专家策展审查，并将记录发布到混合密集/语义索引中。专家审查通过档案管理员应用程序（Archivist App）进行，这是一个面向策展人的界面，支持对人工智能生成的转录和元数据进行修正。受管控的语料库为研究者界面和“与TR对话”（Talk to TR）提供支持，这是一个持续运行的展览，将西奥多·罗斯福以全尺度数字人类的形式呈现在博物馆环境中。为了支持实时的面对面互动，跨时代类比基础（Cross-Era Analogical Grounding）通过记录的历史平行重新构建当代问题，使罗斯福能够在不虚构事实的情况下讨论当今话题。双路径检索和端到端流媒体保持响应的准确性。分层监控、访客会话隔离、自动化对话管理和可独立重启的服务使数百名访客能够可靠地进行无人值守操作。化身的真实感、空间音频、照明、布景和对话设计作为一个综合体验进行开发和评估。我们并不报告一个受控基准，而是描述了运营“与TR对话”作为公共展览的经验教训，并提供了一个可转移的模型，以将档案收藏转变为可信的面对面对话体验。
cs.CV / 10 / 2609.09394
OmniPoint: Universal Monocular Metric Pointcloud from Any Camera
OmniPoint：来自任意相机的通用单目度量点云
Abstract
Recovering metric 3D geometry from monocular images is a fundamental computer vision task, yet current methods remain heavily fragmented by fixed camera model assumptions and inflexible input schemes. We present OmniPoint, a unified framework designed to generalize metric reconstruction across diverse imaging sensors, including pinhole, fisheye, and equirectangular projections, while accommodating varying geometric priors. To overcome projection rigidity, OmniPoint abandons conventional planar depth regression. It instead adopts a decoupled ray and distance representation alongside a decoupled training objective, explicitly separating the camera projection model from the scene structure. To address the severe scarcity of training data for alternative cameras, we introduce a bidirectional augmentation strategy that explicitly bridges labeled perspective data and unlabeled omnidirectional domains in 3D space. Furthermore, to seamlessly integrate optional inputs like camera intrinsics or sparse depth without destabilizing the network through feature distribution shifts, we propose a robust information injection mechanism. This mechanism utilizes learnable input state embeddings to resolve architectural ambiguity and applies vectorized Gaussian smoothing to densify irregular measurements. Extensive experiments demonstrate that OmniPoint achieves state-of-the-art zero-shot performance across multiple benchmarks, establishing a robust new standard for unified monocular 3D reconstruction.
Chinese Translation
从单目图像恢复度量三维几何结构是计算机视觉中的一项基本任务，但当前的方法仍然受到固定相机模型假设和不灵活输入方案的严重限制。我们提出了OmniPoint，这是一种统一框架，旨在跨越多种成像传感器（包括针孔、鱼眼和等距投影）进行度量重建，同时适应不同的几何先验。为了克服投影的刚性，OmniPoint摒弃了传统的平面深度回归方法，而是采用了解耦的光线和距离表示以及解耦的训练目标，明确将相机投影模型与场景结构分离。为了解决替代相机训练数据的严重稀缺问题，我们引入了一种双向增强策略，明确连接标记的透视数据和未标记的全向域在三维空间中的关系。此外，为了在不通过特征分布变化使网络不稳定的情况下无缝集成可选输入（如相机内参或稀疏深度），我们提出了一种稳健的信息注入机制。该机制利用可学习的输入状态嵌入来解决架构模糊性，并应用向量化的高斯平滑来密集化不规则测量。大量实验表明，OmniPoint在多个基准测试中实现了最先进的零样本性能，为统一单目三维重建建立了一个稳健的新标准。
cs.CV / 11 / 2609.09396
VANTAGE-Bench: Evaluating the Infrastructure AI Gap in Vision-Language Models
VANTAGE-Bench：评估视觉语言模型中的基础设施人工智能差距
Abstract
As Vision-Language Models (VLMs) advance toward physical deployment, the focus has remained on action-oriented Embodied AI evaluated on subject-centric consumer video. This overlooks a pervasive class of Physical AI: Infrastructure AI, which relies on fixed cameras for open-loop insights like safety monitoring and operational logging. We introduce VANTAGE-Bench, a benchmark measuring this "Infrastructure AI Gap." It spans three operational domains (Logistics, Transportation, and Smart Spaces), unifies image and video evaluation across semantic, spatial, temporal, and spatio-temporal capabilities, and moves beyond multiple-choice to eight task formulations including dense captioning and spatio-temporal grounding. It adds a single-pass trajectory protocol for Single Object Tracking and, to our knowledge, the first such evaluation on fixed-camera infrastructure video, scored against specialist trackers. Annotation spans three regimes over 3,346 media assets: 3,342 video-task annotations, 4,281 image-grounding annotations, and 27,404 detection boxes. Evaluating 17 models zero-shot, we find the shortfall relative to consumer-centric benchmarks is concentrated, not general. Event verification, referring expressions, and temporal localization fall roughly 9 to 24 points at every model scale, while video question answering stays within 5.3 points of VideoMME and 2D spatial pointing shows no shortfall against BLINK. The temporal pillar is weakest in absolute terms: no system exceeds 55.7 mIoU on temporal localization or 37.3 SODA_c on dense video captioning. On tracking, frontier models come within roughly 5 points of specialist trackers over short horizons but separate as the horizon extends. Open-weight models lead 2D object localization outright, so neither scale nor proprietary access explains the pattern. Data, evaluation harness, and leaderboard: https://vantage-bench.org/
Chinese Translation
随着视觉语言模型（VLMs）向实际部署的推进，研究重点仍然集中在以行动为导向的具身人工智能上，主要评估以主体为中心的消费者视频。这忽视了一类普遍存在的物理人工智能：基础设施人工智能，它依赖于固定摄像头进行开放式洞察，如安全监控和操作日志记录。我们引入了VANTAGE-Bench，这是一个衡量这一“基础设施人工智能差距”的基准。它涵盖三个操作领域（物流、交通和智能空间），统一了图像和视频在语义、空间、时间和时空能力方面的评估，并超越了多项选择，提出了包括密集字幕生成和时空定位在内的八种任务形式。它增加了单次轨迹协议用于单目标跟踪，并且据我们所知，这是在固定摄像头基础设施视频上首次进行的此类评估，得分与专业跟踪器进行比较。注释涵盖三个领域，共涉及3,346个媒体资产：3,342个视频任务注释、4,281个图像定位注释和27,404个检测框。在零样本评估17个模型时，我们发现相对于以消费者为中心的基准，短缺主要集中，而非普遍存在。事件验证、指称表达和时间定位在每个模型规模上大约下降9到24分，而视频问答则在VideoMME的基础上保持在5.3分以内，2D空间指向在与BLINK的比较中没有短缺。时间维度在绝对值上最弱：没有系统在时间定位上超过55.7 mIoU或在密集视频字幕生成上超过37.3 SODA_c。在跟踪方面，前沿模型在短时间范围内与专业跟踪器的差距约为5分，但随着时间范围的延长而分离。开放权重模型在2D物体定位中表现突出，因此规模或专有访问并不能解释这一模式。数据、评估工具和排行榜： https://vantage-bench.org/
cs.CV / 12 / 2609.09417
Vision-language models know more about agriculture than they show and rubric-grounded verifications close the gap
视觉-语言模型对农业的了解超出其表现，基于评分标准的验证缩小了差距
Abstract
Vision-language models (VLMs) show promise for agricultural classification, but zero-shot performance on disease, pest, damage, quality, and species identification remains poor, and it is unclear whether this reflects weak visual features or a failure to connect them to domain knowledge. We build a benchmark of 116 datasets, 834 classes, and 8,324 images spanning these tasks to isolate where the gap arises. Linear probing shows VLM vision encoders already encode agricultural features nearly as separable as a self-supervised DINOv3 baseline, ruling out weak visual representations as the primary bottleneck. Conditioning each model on an oracle reference description (an upper bound on its parametric knowledge) closes most of the gap left by an unaided lower bound, showing VLMs already know more about agriculture than they show. To close this gap without an oracle description at inference time, we structure test-time reasoning around a fixed, per-task diagnostic rubric: the model generates $K$ candidate responses and a Probabilistic Pivot Tournament (PPT) verifier, scored pairwise against the rubric, selects the best one. This nearly doubles judged F1 over the lower bound and matches or exceeds the upper bound on several tasks, notably pushing Gemma 4 E4B-it's disease F1 to 0.71, above its own upper bound of 0.60. However, the verifier's letter-scale confidence score has the opposite of its intended effect: filtering to its most confident predictions does not improve accuracy and correlates negatively with correctness across every model and pool size tested, so the score cannot serve as a measure of predictive uncertainty, and most of the observed gain likely comes from rubric-grounded generation rather than pairwise verification.
Chinese Translation
视觉-语言模型（VLMs）在农业分类中展现出良好的前景，但在疾病、害虫、损害、质量和物种识别等任务上的零样本表现仍然较差，目前尚不清楚这是否反映了视觉特征的弱点或未能将其与领域知识相连接。我们构建了一个包含116个数据集、834个类别和8,324张图像的基准，以隔离差距产生的原因。线性探测显示，VLM的视觉编码器已经编码了几乎与自监督的DINOv3基线相当可分的农业特征，排除了弱视觉表示作为主要瓶颈的可能性。通过将每个模型与一个神谕参考描述（其参数知识的上限）进行条件化，缩小了未辅助下限所留下的大部分差距，表明VLM对农业的了解已经超出其表现。为了在推理时不依赖神谕描述来缩小这一差距，我们围绕固定的每任务诊断评分标准构建了测试时推理：模型生成$K$个候选响应，并由概率枢轴锦标赛（Probabilistic Pivot Tournament, PPT）验证器根据评分标准进行成对评分，从中选择最佳响应。这几乎将评估的F1值在下限基础上翻倍，并在多个任务上匹配或超过上限，特别是将Gemma 4 E4B的疾病F1推高至0.71，超过其自身的上限0.60。然而，验证器的字母级置信分数却产生了与预期效果相反的结果：筛选出其最有信心的预测并未提高准确性，并且在每个模型和池规模的测试中与正确性呈负相关，因此该分数无法作为预测不确定性的衡量标准，而观察到的大部分增益可能来自基于评分标准的生成，而非成对验证。
cs.CV / 13 / 2609.09424
Longitudinal tracking of multiple sclerosis lesions in the spinal cord: A validation study
脊髓中多发性硬化病灶的纵向跟踪：一项验证研究
Abstract
Longitudinal characterization of multiple sclerosis (MS) lesions remains constrained by the lack of frameworks capable of establishing consistent instance-level correspondences across time. Conventional segmentation approaches produce semantic lesion masks at each visit and therefore fail to capture the complex instance temporal patterns associated with lesion appearance, disappearance, splitting, or merging. This study presents a comparative evaluation of five strategies for automated tracking of spinal cord MS lesions in longitudinal MRI data from a multi-site cohort. The investigated strategies rely either on deformable registration or on a spinal anatomical reference system, and encompass overlap-based matching, coordinate-based Hungarian algorithm, gradient-boosted classification, and Siamese model classification. Tracking accuracy is quantified using instance-level true positives, false positives, and false negatives, allowing to assess the presence of one-to-many and many-to-one associations. Results show best performance for the registration-based overlap method. This study provides the first systematic analysis of lesion-instance correspondence in the spinal cord and outlines the strengths and limitations of registration-based and registration-free paradigms for longitudinal MS assessment. The code is available at http://github.com/ivadomed/longitudinal-sc-ms-lesion-tracking .
Chinese Translation
多发性硬化（MS）病灶的纵向特征化受到缺乏能够在时间上建立一致实例级对应关系的框架的限制。传统的分割方法在每次访视中生成语义病灶掩膜，因此无法捕捉与病灶出现、消失、分裂或合并相关的复杂实例时间模式。本研究对来自多中心队列的纵向MRI数据中脊髓MS病灶的自动跟踪的五种策略进行了比较评估。所调查的策略依赖于可变形配准或脊髓解剖参考系统，并包括基于重叠的匹配、基于坐标的匈牙利算法、梯度提升分类和孪生模型分类。通过实例级真实阳性、假阳性和假阴性的量化来评估跟踪准确性，从而能够评估一对多和多对一关联的存在。结果表明，基于配准的重叠方法表现最佳。本研究提供了脊髓中病灶实例对应关系的首次系统分析，并概述了基于配准和无配准范式在纵向MS评估中的优缺点。代码可在 http://github.com/ivadomed/longitudinal-sc-ms-lesion-tracking 获取。
cs.CV / 14 / 2609.09462
Low-Rank Prompt Learning for Vision-Language Models with Fixed-Token Bases
基于固定标记基础的视觉-语言模型低秩提示学习
Abstract
Prompt learning adapts CLIP to downstream recognition by replacing hand-written templates with learned continuous context vectors, which in Context Optimization (CoOp) form a dense prompt matrix $\mathbf{P}\in\mathbb{R}^{m\times d}$ trained from only a few examples per class. We study whether this matrix is over-parameterized by factorizing it as $\mathbf{P}=\mathbf{B}\mathbf{A}$, which cuts the trainable prompt parameters from $md$ to $r(m+d)$, and to $rd$ once the token-side factor $\mathbf{B}$ is fixed. Across seven few-shot benchmarks and two CLIP backbones, low-rank prompts match or improve dense CoOp at far fewer parameters, with the clearest gains on low-shot base-to-new generalization. We then find that the token-side factor need not be learned at all: fixing $\mathbf{B}$ to a Gaussian, orthogonal, SVD-derived, or even random basis and training only the embedding-side factor $\mathbf{A}$ stays on par with the fully trainable factorization, and a source-trained $\mathbf{B}$ offers no advantage over a random one. A prompt-factor asymmetry and a local update-space dimension gap show why fixing $\mathbf{B}$ is far less restrictive than fixing $\mathbf{A}$, and a smoothness-only guarantee certifies that optimizing $\mathbf{A}$ over a fixed $\mathbf{B}$ converges. In the CLIP prompt setting, the embedding-side coefficients carry the adaptation while the token basis can simply be fixed.
Chinese Translation
提示学习通过用学习到的连续上下文向量替代手写模板，将 CLIP 适应于下游识别。在上下文优化（CoOp）中，这形成了一个稠密的提示矩阵 $ extbf{P} extbf{ extbackslash}in extbf{ extbackslash}mathbb{R}^{m imes d}$，该矩阵仅从每个类别的少量示例中训练。我们研究了这个矩阵是否过度参数化，通过将其分解为 $ extbf{P}= extbf{B} extbf{A}$，将可训练的提示参数从 $md$ 削减到 $r(m+d)$，并在固定标记侧因子 $ extbf{B}$ 后进一步减少到 $rd$。在七个少样本基准和两个 CLIP 主干网络上，低秩提示在参数数量远少的情况下与稠密的 CoOp 相匹配或改进，尤其在低样本基础到新类别的泛化上表现出明显的优势。我们发现，标记侧因子并不需要学习：将 $ extbf{B}$ 固定为高斯、正交、SVD 派生或甚至随机基底，仅训练嵌入侧因子 $ extbf{A}$ 的效果与完全可训练的分解相当，而源训练的 $ extbf{B}$ 相较于随机的 $ extbf{B}$ 并没有优势。提示因子的不对称性和局部更新空间维度差异表明，固定 $ extbf{B}$ 的限制远低于固定 $ extbf{A}$，而仅平滑性保证证明了在固定 $ extbf{B}$ 的情况下优化 $ extbf{A}$ 是收敛的。在 CLIP 提示设置中，嵌入侧系数承担了适应的任务，而标记基础可以简单地固定。
cs.CV / 15 / 2609.09477
LeCor: Learning to Be Corrected by Meta-Learned Test-Time Training for Interactive 3D Lung-Tumour Segmentation
LeCor：通过元学习的测试时训练学习被纠正以实现交互式3D肺肿瘤分割
Abstract
Delineating lung tumours on computed tomography (CT) takes a considerable share of the time spent on radiotherapy planning, and a contour proposed by a model can be refined interactively by the clinician. Promptable foundation models such as SAM 3 support this workflow by writing each correction into a session memory that conditions the remaining slices, while the model weights stay fixed. On 690 test cases from five public CT cohorts, fine-tuning SAM 3 on lung tumours raises the Dice obtained from a single point prompt from 0.298 to 0.757, and seven rounds of corrections raise it further to 0.765, but under memory conditioning alone the accuracy on slices the annotator has not touched stops improving after six rounds. We therefore treat each correction as a training signal and propose LeCor, which performs test-time training on a small set of case adapters that are reset for every case and meta-learned such that a single gradient step driven by a click improves the slices that were not clicked. On the 133 test cases that span at least eight slices, LeCor raises the Dice reached after seven correction rounds from 0.787 with the fine-tuned model to 0.827, reduces the number of cases that never reach a Dice of 0.80 from 47 to 27, and reaches in three correction rounds the accuracy that the fine-tuned model attains in seven.
Chinese Translation
在计算机断层扫描（CT）上勾画肺肿瘤占据了放射治疗规划中相当一部分时间，模型提出的轮廓可以由临床医生进行交互式细化。可提示的基础模型如SAM 3支持这一工作流程，通过将每次修正写入会话记忆中，以便对剩余切片进行条件化，同时模型权重保持不变。在来自五个公共CT队列的690个测试案例中，对肺肿瘤进行SAM 3的微调使得从单点提示获得的Dice系数从0.298提高到0.757，经过七轮修正进一步提高至0.765，但在仅依赖记忆条件下，标注者未触及的切片的准确性在六轮后停止改善。因此，我们将每次修正视为训练信号，提出LeCor，该方法在每个案例中重置的小型案例适配器上进行测试时训练，并进行元学习，使得由点击驱动的单次梯度步骤能够改善未被点击的切片。在涵盖至少八个切片的133个测试案例中，LeCor将经过七轮修正后达到的Dice系数从微调模型的0.787提高到0.827，将从未达到0.80的案例数量从47减少到27，并在三轮修正中达到微调模型在七轮中获得的准确性。
cs.CV / 16 / 2609.09482
Infra-Bench CLS: A Global, Open-Source Benchmark for Critical Infrastructure Classification with Earth Observation Foundation Models
Infra-Bench CLS：基于地球观测基础模型的全球开放关键基础设施分类基准
Abstract
Critical infrastructure location data is often incomplete and unevenly distributed globally, especially in developing regions. Earth observation foundation models are proposed as a new step in enabling us to more efficiently understand the natural and built environment, raising questions as to their effectiveness in performing challenging downstream tasks. Yet, foundation models remain largely untested for detecting and classifying the facility-scale critical infrastructure that underpins a range of important societal and economic functions. Subsequently, Infra-Bench CLS is introduced as a benchmark to test foundation models on 18,756 Sentinel-1 SAR and Sentinel-2 multispectral facility-scale critical infrastructure asset images covering seven continents and 13 infrastructure classes, with results reported for the 10 retained classes. Using linear probing and fine-tuning for two training dataset levels (1.0x and 0.3x), seven foundation models are evaluated (SatlasPretrain S2, SatlasPretrain S1, CROMA, Prithvi-EO-2.0, AlphaEarth Foundations, OlmoEarth v1.1-Base, and DINOv3 ViT-L/16). When comparing macro F1 scores to a ResNet-18 supervised baseline of 39.2 percent, the best foundation model achieved 57.9 percent, a 48 percent improvement. Top performing classes were airports (F1 85.3 percent), train stations (F1 82.1 percent), and data centers (F1 77.6 percent). By contrast, many of the power sector classes perform poorly (F1 27.5-46.2 percent). These findings suggest foundation models can enable superior critical infrastructure classification, but future work should evaluate performance on higher-resolution imagery, particularly for poorly performing sectors, such as power.
Chinese Translation
关键基础设施的位置信息通常存在不完整且全球分布不均的问题，尤其是在发展中地区。地球观测基础模型被提出作为更高效理解自然与建成环境的新途径，但其在执行复杂下游任务中的有效性仍存疑问。然而，基础模型在检测和分类支撑多项重要社会经济功能的设施级关键基础设施方面尚未得到充分测试。因此，本文引入Infra-Bench CLS作为一个基准，利用覆盖七大洲、包含13类基础设施的18,756幅Sentinel-1 SAR和Sentinel-2多光谱设施级关键基础设施资产图像，对基础模型进行测试，并报告了10个保留类别的结果。通过线性探测和微调两种训练数据集规模（1.0倍和0.3倍），评估了七个基础模型（SatlasPretrain S2、SatlasPretrain S1、CROMA、Prithvi-EO-2.0、AlphaEarth Foundations、OlmoEarth v1.1-Base及DINOv3 ViT-L/16）。与39.2%的ResNet-18监督基线宏平均F1分数相比，表现最佳的基础模型达到了57.9%，提升了48%。表现最好的类别为机场（F1 85.3%）、火车站（F1 82.1%）和数据中心（F1 77.6%）；相比之下，许多电力行业类别表现较差（F1 27.5%-46.2%）。这些结果表明基础模型能够实现更优的关键基础设施分类，但未来工作应针对分辨率更高的影像进行性能评估，尤其是针对表现不佳的行业如电力领域。
cs.CV / 17 / 2609.09491
Learning Global Camera Poses from Noisy View-Graphs for Structure from Motion
从噪声视图图中学习全局相机位姿以进行运动结构重建
Abstract
Camera pose estimation is a key step in 3D reconstruction and view-synthesis pipelines. We present a deep, global Structure-from-Motion framework based on learned view-graph aggregation. Our method employs a permutation-equivariant, edge-conditioned graph neural network that takes noisy pairwise relative poses as input and outputs globally consistent camera extrinsics. The network is trained without ground-truth supervision, relying solely on a relative-pose consistency objective. This is followed by 3D point triangulation and robust bundle adjustment. Our approach is efficient, scalable to more than a thousand images, and robust to graph density. We evaluate our method on MegaDepth, 1DSfM, Strecha, and BlendedMVS. These experiments demonstrate that our method achieves superior rotation and translation accuracy compared to deep track-centric methods while registering more images across many scenes, and competitive results compared to state-of-the-art classical pipelines, while being much faster.
Chinese Translation
相机位姿估计是三维重建和视图合成流程中的关键步骤。我们提出了一种基于学习视图图聚合的深度全局运动结构重建框架。我们的方法采用了一种置换等变的边条件图神经网络，该网络以噪声对相对位姿为输入，并输出全局一致的相机外参。该网络在没有真实值监督的情况下进行训练，仅依赖于相对位姿一致性目标。接下来进行三维点三角测量和稳健的束调整。我们的方法高效、可扩展至超过一千张图像，并且对图的密度具有鲁棒性。我们在MegaDepth、1DSfM、Strecha和BlendedMVS上评估了我们的方法。这些实验表明，我们的方法在旋转和位移精度上优于深度轨迹中心方法，同时在多个场景中注册更多图像，并且与最先进的经典流程相比，结果具有竞争力，同时速度更快。
cs.CV / 18 / 2609.09507
RoMa-$\Omega$: What Feed-Forward 3D Models Know About Image Matching
RoMa-$ ext{Ω}$：前馈3D模型对图像匹配的认知
Abstract
Learned image matching has experienced significant progress in recent years, culminating in robust and accurate matchers such as RoMa, whose robustness is often attributed to its use of frozen DINO features. In a parallel development, feed-forward reconstruction models, such as VGGT, have been trained on ever-growing datasets to accurately regress dense 3D point maps and camera poses. The distinction between matchers and feed-forward reconstruction models has become increasingly blurred with the introduction of matching losses in models such as MASt3R and VGGT-$\Omega$. This raises a natural question: what do feed-forward 3D models know about image matching? In this work, we answer this question by analyzing three scenarios: (i) zero-shot matching of patch features, (ii) direct matching of 3D point predictions, and (iii) training a full matcher on top of the learned representations. We find that, despite performing poorly in zero-shot matching, especially in later layers, feed-forward reconstruction models provide strong representations for linear probing and full matching pipelines. We further show that, even without any training, their raw predictions alone enable competitive matching, albeit only under moderate viewpoint changes and modality gaps. Based on these insights, we retrain RoMa v2 by replacing its DINO backbone with VGGT-$\Omega$. Our resulting model, \ours, outperforms state-of-the-art matchers on a wide range of benchmarks, e.g. +8.1 mAA compared to RoMa v2 on WxBS.
Chinese Translation
近年来，学习型图像匹配取得了显著进展，涌现出诸如RoMa等鲁棒且准确的匹配器，其鲁棒性通常归因于使用了冻结的DINO特征。在此过程中，前馈重建模型（如VGGT）在不断扩大的数据集上进行训练，以准确回归稠密的3D点图和相机姿态。随着在MASt3R和VGGT-$ ext{Ω}$等模型中引入匹配损失，匹配器与前馈重建模型之间的区别变得越来越模糊。这引发了一个自然的问题：前馈3D模型对图像匹配了解多少？在本研究中，我们通过分析三种场景来回答这个问题：（i）补丁特征的零样本匹配，（ii）3D点预测的直接匹配，以及（iii）在学习表示的基础上训练完整的匹配器。我们发现，尽管在零样本匹配中表现不佳，尤其是在后层，前馈重建模型为线性探测和完整匹配管道提供了强大的表示。我们进一步表明，即使没有任何训练，它们的原始预测也能实现具有竞争力的匹配，尽管仅在适度的视角变化和模态差异下有效。基于这些见解，我们通过用VGGT-$ ext{Ω}$替换RoMa v2的DINO主干进行再训练。我们得到的模型 ext{ours}在广泛的基准测试中超越了最先进的匹配器，例如在WxBS上相比RoMa v2提高了8.1 mAA。
cs.CV / 19 / 2609.09513
AnimalLift: Reconstructing Animatable 3D Animals from a Single Image by Learning Canonical Shape, Texture, and Fur Maps
AnimalLift：通过学习典型形状、纹理和毛发图，从单幅图像重建可动画的三维动物
Abstract
Reconstructing a fully animatable 3D animal from a single image remains challenging because animation-ready assets require not only plausible geometry, but also a unified topology, editable appearance, and fur representations compatible with deformation and simulation. Existing image-to-3D approaches often rely on implicit or loosely structured representations that are difficult to rig or edit, while parametric animal models support animation but cannot capture detailed texture and fur appearance. We present AnimalLift, a framework for reconstructing structured, animation-compatible 3D animal assets with explicit fur from a single image. Our method lifts an input image into a shared canonical space with a consistent topology and UV parameterization across the dataset, enabling joint prediction of canonical geometry, texture, and fur in a unified feed-forward architecture. A key component of our representation is a UV-aligned fur map that encodes strand geometry in a surface-aligned canonical domain, allowing explicit fur reconstruction compatible with mesh deformation and fur simulation. To train the model, we introduce a procedural data generation pipeline that provides large-scale supervision with aligned geometry, texture, and fur across diverse animal species and appearances. Experiments on synthetic and real-world datasets demonstrate strong reconstruction quality and generalization across animal categories. Beyond reconstruction, our structured representation directly supports downstream applications including animation, pose transfer, fur editing, and simulation-compatible rendering.
Chinese Translation
从单幅图像重建一个完全可动画的三维动物仍然具有挑战性，因为动画准备好的资产不仅需要合理的几何形状，还需要统一的拓扑结构、可编辑的外观以及与变形和模拟兼容的毛发表示。现有的图像到三维的方法通常依赖于隐式或松散结构的表示，这些表示难以进行绑定或编辑，而参数化动物模型支持动画但无法捕捉详细的纹理和毛发外观。我们提出了AnimalLift，一个从单幅图像重建结构化、兼容动画的三维动物资产的框架，具有明确的毛发表示。我们的方法将输入图像提升到一个共享的典型空间，该空间在整个数据集中具有一致的拓扑和UV参数化，从而能够在统一的前馈架构中联合预测典型几何形状、纹理和毛发。我们表示的一个关键组成部分是一个UV对齐的毛发图，它在表面对齐的典型域中编码了发丝几何形状，允许与网格变形和毛发模拟兼容的明确毛发重建。为了训练模型，我们引入了一种程序化数据生成管道，该管道提供了大规模的监督，涵盖了不同动物物种和外观的对齐几何形状、纹理和毛发。对合成和真实世界数据集的实验表明，重建质量强，且在动物类别之间具有良好的泛化能力。除了重建之外，我们的结构化表示还直接支持下游应用，包括动画、姿态转移、毛发编辑和兼容模拟的渲染。
cs.CV / 20 / 2609.09528
MotionBlind: Probing the Illusion of Motion Understanding in Video-LLMs
MotionBlind：探究视频大语言模型中运动理解的幻觉
Abstract
Video large language models (Video-LLMs) are increasingly used as the perceptual front end of world models, a role that assumes they can read motion: how fast something moves, which way it travels, how hard it is pushed. We show they cannot. A Video-LLM can watch two clips of the same person in the same room, name every object in both, and still fail to say which clip moves faster. We introduce MotionBlind, a contrastive benchmark of self-recorded video for physically grounded motion(speed, magnitude, and direction), the variables a world model must predict. Each instance is a pair of near-identical clips that differ only in motion. Each clip carries two complementary yes/no questions, giving four items per instance, and a model earns credit only if all four are correct. We report Instance Accuracy(IAcc), which has a 6.25% chance floor. Single-frame, appearance, and language-only shortcuts all collapse to it. MotionBlind complements the recent TimeBlind benchmark. We run a controlled study of six open and two frontier Video-LLMs, varying whether the video is present, whether frames are shown in the correct temporal order, and how frames are sampled (1 to 24 frames, four selection strategies). Open models sit near the 6.25% floor, and scale does not help. Removing the video drops every model to zero IAcc, and shuffling frames collapses IAcc to chance, so the task genuinely needs video in order. Neither more frames nor smarter frame selection closes the gap, because these change which frames are seen, not whether motion is read. Only Gemini3.1 Pro clears the benchmark overall, and even it fails on speed. A frontend that cannot tell two speeds of the same action apart is not yet a trustworthy source of supervision, reward, or evaluation for a world model.
Chinese Translation
视频大语言模型（Video-LLMs）越来越多地被用作世界模型的感知前端，这一角色假设它们能够理解运动：物体移动的速度、方向以及受到的推力大小。然而，我们的研究表明，它们并不能做到这一点。一个视频大语言模型可以观看同一房间中同一个人的两个片段，能够命名两个片段中的每一个物体，却仍然无法判断哪个片段移动得更快。我们引入了MotionBlind，这是一个对自录视频进行对比基准测试的工具，旨在评估物理基础的运动（速度、大小和方向），这些都是世界模型必须预测的变量。每个实例是一对几乎相同的片段，仅在运动上有所不同。每个片段包含两个互补的是/否问题，每个实例总共提供四个问题，模型只有在所有四个问题都正确时才能获得积分。我们报告了实例准确率（Instance Accuracy, IAcc），其最低概率为6.25%。单帧、外观和仅语言的捷径均会降至这一水平。MotionBlind补充了最近的TimeBlind基准。我们对六个开放模型和两个前沿模型进行了控制研究，改变了视频是否存在、帧是否按正确的时间顺序显示以及帧的采样方式（1到24帧，四种选择策略）。开放模型的表现接近6.25%的底线，规模并没有带来帮助。去掉视频后，每个模型的IAcc降至零，而打乱帧的顺序则将IAcc降至随机水平，因此该任务确实需要视频的顺序。增加帧数或更智能的帧选择并未缩小差距，因为这些变化影响的是看到的帧，而不是运动是否被理解。只有Gemini3.1 Pro在整体上通过了基准测试，但它在速度方面仍然失败。一个无法区分同一动作的两种速度的前端，尚不能作为世界模型的可靠监督、奖励或评估来源。
cs.CV / 21 / 2609.09606
RouteBridge: Reliability-Routed Bidirectional Distillation Between Neural Radiance Fields and 3D Gaussian Splatting
RouteBridge：神经辐射场与三维高斯点云之间的可靠性路由双向蒸馏
Abstract
Neural radiance fields (NeRFs) and 3D Gaussian Splatting (3DGS) encode a scene with complementary inductive biases, but existing cross-representation distillation typically fixes one representation as teacher for the entire scene. A globally fixed teacher can propagate local reconstruction errors. We present RouteBridge, a bidirectional framework that selects the teaching direction for each ray. Its reliability estimator combines photometric residuals with representation-specific geometric evidence and routes supervision from NeRF to 3DGS, from 3DGS to NeRF, or abstains. A renderer-independent interface transfers color, opacity, and normalized depth without shared features or point correspondence. On mip-NeRF 360, the NeRF and 3DGS exports reach 28.56 and 28.77 dB, respectively. The 3DGS export improves over 3DGS by 1.56 dB and over NeRF-GS by 0.45 dB while reducing LPIPS to 0.207. On static three-view DTU, RouteBridge obtains 21.12 dB. Ablations show that both adaptive routing and geometric ray targets contribute to the improvement.
Chinese Translation
神经辐射场（NeRFs）和三维高斯点云（3D Gaussian Splatting, 3DGS）以互补的归纳偏差编码场景，但现有的跨表示蒸馏通常将一个表示固定为整个场景的教师。全局固定的教师可能会传播局部重建误差。我们提出了RouteBridge，一个双向框架，针对每个光线选择教学方向。其可靠性估计器结合了光度残差与特定于表示的几何证据，并将监督从NeRF路由到3DGS，从3DGS路由到NeRF，或选择不进行路由。在渲染器无关的接口中，颜色、不透明度和归一化深度可以在没有共享特征或点对应的情况下进行传输。在mip-NeRF 360上，NeRF和3DGS的导出分别达到28.56和28.77 dB。3DGS的导出比3DGS提高了1.56 dB，比NeRF-GS提高了0.45 dB，同时将LPIPS降低至0.207。在静态三视图DTU上，RouteBridge获得了21.12 dB。消融实验表明，自适应路由和几何光线目标均对改进有贡献。
cs.CV / 22 / 2609.09610
Marker-free eye-gaze estimation using a single image and depth from defocus
基于单幅图像和模糊深度的无标记眼动估计
Abstract
This paper presents a marker-free eye-gaze estimation approach using a single 2D camera, such as an integrated laptop webcam. The gaze-related features are estimated from iris localization and head pose estimated by using depth from defocus. A variational Bayesian multinomial logistic regression framework is used as mapping from the estimated features to the position of regard, based on an 8-dimensional feature vector of head-pose and iris-displacement parameters. No external marker is needed. Experiments were conducted by estimating the gaze of people watching a computer screen at different distances and compared against five existing methods. The obtained scores demonstrate the effectiveness of the proposed approach.
Chinese Translation
本文提出了一种基于单个二维摄像头（例如集成的笔记本电脑摄像头）的无标记眼动估计方法。通过使用模糊深度估计虹膜定位和头部姿态，从而估计与注视相关的特征。采用变分贝叶斯多项式逻辑回归框架，将估计的特征映射到注视位置，基于一个包含头部姿态和虹膜位移参数的8维特征向量。该方法无需外部标记。通过估计在不同距离观看计算机屏幕的人的注视，进行了实验，并与五种现有方法进行了比较。获得的分数证明了所提方法的有效性。
cs.CV / 23 / 2609.09626
Hyperbolic Geometry for Open-World Object Detection in Remote Sensing Imagery
用于遥感影像开放世界目标检测的双曲几何
Abstract
Open-world object detection (OWOD) extends closed-set detection by requiring models to identify unknown objects and incrementally learn them once annotations become available. In remote sensing imagery, object categories often exhibit latent hierarchical relationships that may be inadequately represented in the Euclidean spaces commonly adopted by existing methods, limiting unknown-object recall and incremental-learning performance. To address this issue, we investigate hyperbolic geometry for OWOD in remote sensing imagery and propose HyRS-OWOD. To improve unknown object recall, we design a two-step unknown-object discovery mechanism: a Decoupled Objectness Learning (DOL) module that disentangles foreground perception from semantic information to separate foreground proposals from background regions, followed by a Hyperbolic Uncertainty Learning (HUL) component that leverages the radius of hyperbolic embeddings as an uncertainty-aware cue for known-unknown discrimination. For incremental learning, we develop a Hyperbolic Metric Learning (HML) strategy that enhances inter-class separability, facilitating the incorporation of novel categories while mitigating catastrophic forgetting. Experiments on three remote sensing benchmarks demonstrate consistent improvements in unknown recall and incremental learning over state-of-the-art OWOD methods.
Chinese Translation
开放世界目标检测（OWOD）通过要求模型识别未知目标并在注释可用后逐步学习这些目标，从而扩展了封闭集检测。在遥感影像中，目标类别通常表现出潜在的层次关系，而这些关系在现有方法常用的欧几里得空间中可能表现不足，限制了未知目标的召回率和增量学习性能。为了解决这一问题，我们研究了在遥感影像中应用双曲几何的OWOD，并提出了HyRS-OWOD。为了提高未知目标的召回率，我们设计了一个两步未知目标发现机制：一个解耦目标性学习（DOL）模块，它将前景感知与语义信息解耦，以将前景提议与背景区域分离，随后是一个双曲不确定性学习（HUL）组件，它利用双曲嵌入的半径作为已知-未知区分的感知线索。对于增量学习，我们开发了一种双曲度量学习（HML）策略，以增强类间可分离性，促进新类别的纳入，同时减轻灾难性遗忘。在三个遥感基准测试上的实验表明，与最先进的OWOD方法相比，未知召回率和增量学习均有一致的提升。
cs.CV / 24 / 2609.09634
LightMedSeg-ISLES: Stroke Lesion Segmentation with 81x Fewer Parameters than nnU-Net
LightMedSeg-ISLES：与 nnU-Net 相比参数减少 81 倍的中风病灶分割
Abstract
Large networks and ensembles often lead medical image segmentation challenges, but their storage and inference demands complicate deployment. We present LightMedSeg-ISLES, a 1.26-million-parameter pipeline for T1-weighted stroke lesion segmentation in ISLES'26. On a 146-case held-out cohort, flip test-time augmentation produces 0.618 mean Dice and 0.599 lesion-wise F1. A 102.35-million-parameter nnU-Net ResEnc-L produces 0.634 Dice and 0.544 lesion-wise F1 after size filtering. LightMedSeg therefore retains 97.5\% of nnU-Net's Dice with 81.4$\times$ fewer parameters while improving lesion-wise F1 by 0.055. Its four-pass TTA operating point requires 4.7$\times$ fewer FLOPs per standardized patch than nnU-Net. It also slightly exceeds filtered UNETR++ and nnFormer. Longer training and stronger augmentation add 0.0358 Dice without increasing capacity, establishing a strong single-checkpoint alternative to much larger models.
Chinese Translation
大型网络和集成模型通常在医学图像分割挑战中占据主导地位，但它们的存储和推理需求使得部署变得复杂。我们提出了 LightMedSeg-ISLES，这是一个拥有 126 万参数的管道，用于 ISLES'26 中 T1 加权中风病灶的分割。在一个包含 146 个病例的保留队列中，翻转测试时增强（flip test-time augmentation）产生了 0.618 的平均 Dice 系数和 0.599 的病灶级 F1 值。经过大小过滤的 1.0235 亿参数的 nnU-Net ResEnc-L 产生了 0.634 的 Dice 系数和 0.544 的病灶级 F1 值。因此，LightMedSeg 在参数减少 81.4 倍的情况下，保留了 nnU-Net 97.5\% 的 Dice 值，同时病灶级 F1 值提高了 0.055。其四次通过的测试时增强（TTA）操作点每个标准化补丁所需的 FLOPs 比 nnU-Net 少 4.7 倍。它的性能也略微超过了过滤后的 UNETR++ 和 nnFormer。更长的训练时间和更强的增强使得 Dice 值增加了 0.0358，而没有增加模型容量，确立了作为更大模型的强有力的单检查点替代方案。
cs.CV / 25 / 2609.09670
Recovering Biomechanical Signals from Missing Keypoints Using Temporal Interpolation in Monocular Gait Analysis
利用时间插值从缺失关键点恢复生物力学信号的单目步态分析
Abstract
Monocular pose estimation enables low-cost gait analysis but is sensitive to missing keypoints caused by occlusion, detection errors, or efficiency-driven model reduction. While prior work on recovering missing joints focuses on complex learned models, the effectiveness of simple temporal methods remains underexplored. We evaluate knee-angle estimation under a missing-ankle-keypoint condition and test a first-order temporal interpolation scheme as a recovery mechanism. Across 527 frames of monocular walking video (428 with valid baseline detections), removing the ankle keypoint increased mean angular error to 23.4{\deg} +/- 46.7{\deg} and collapsed signal variance to near zero. Temporal interpolation reduced error to 1.1{\deg} +/- 6.7{\deg} and restored variance and smoothness to within a few percent of baseline. These results indicate that gait signals possess sufficient temporal redundancy for a simple, computationally trivial interpolation scheme to recover a critical missing joint, without resorting to learned reconstruction models. The findings support low-complexity, real-time-compatible designs for gait analysis in resource-constrained or occlusion-prone monocular settings.
Chinese Translation
单目姿态估计能够实现低成本的步态分析，但对因遮挡、检测错误或效率驱动的模型简化而导致的缺失关键点非常敏感。尽管之前的研究集中于复杂的学习模型来恢复缺失的关节，但简单的时间方法的有效性仍未得到充分探索。我们在缺失踝关节关键点的情况下评估膝关节角度估计，并测试了一种一阶时间插值方案作为恢复机制。在527帧单目行走视频中（其中428帧具有有效的基线检测），去除踝关节关键点使得平均角度误差增加至23.4{ extdegree} +/- 46.7{ extdegree}，信号方差几乎降至零。时间插值将误差减少至1.1{ extdegree} +/- 6.7{ extdegree}，并将方差和平滑度恢复至基线的几个百分点内。这些结果表明，步态信号具有足够的时间冗余，使得简单、计算上无关紧要的插值方案能够恢复一个关键的缺失关节，而无需依赖学习重建模型。这些发现支持在资源受限或易受遮挡的单目环境中进行步态分析的低复杂度、实时兼容设计。
cs.CV / 26 / 2609.09705
Cross-Species Animal Re-Identification with Semantic Consistency Learning
跨物种动物重识别与语义一致性学习
Abstract
Generalizable animal Re-Identification (ReID) aims to recognize individual animals across species with diverse morphologies and ecological contexts. Unlike person ReID, where different domains share similar body structures, animal species often exhibit drastically different anatomical structures and visual patterns, making it difficult to establish shared visual correspondences. As a result, representations learned across species tend to form fragmented embedding spaces, which severely limits cross-species generalization. To address this challenge, we propose Semantic Consistency Learning (SCL), a framework designed to learn representations that remain stable across appearance variations while preserving semantic structures shared across species. SCL consists of two complementary components. Foreground-Background Decoupled Spectral Normalization (FDSNorm) stabilizes feature statistics by suppressing environment-induced style variations in a region-aware manner, while Cross-species Neighborhood Modeling (CNM) captures transferable relational structures across species through dynamic feature neighborhoods. Extensive experiments on 11 public animal ReID datasets demonstrate that SCL consistently outperforms state-of-the-art methods under multiple cross-species evaluation protocols and generalizes effectively to previously unseen species and ecological domains. Code is available at https://github.com/Kemalau/ECCV-26-SCL.
Chinese Translation
可泛化的动物重识别（ReID）旨在识别不同物种中具有多样形态和生态背景的个体动物。与人类重识别不同，不同领域的人体结构相似，而动物物种往往表现出截然不同的解剖结构和视觉模式，这使得建立共享的视觉对应关系变得困难。因此，跨物种学习的表示往往形成碎片化的嵌入空间，这严重限制了跨物种的泛化能力。为了解决这一挑战，我们提出了语义一致性学习（SCL），这是一个旨在学习在外观变化下保持稳定的表示，同时保留跨物种共享的语义结构的框架。SCL由两个互补组件组成。前景-背景解耦谱归一化（FDSNorm）通过以区域感知的方式抑制环境引起的风格变化来稳定特征统计，而跨物种邻域建模（CNM）则通过动态特征邻域捕捉跨物种的可转移关系结构。在11个公共动物重识别数据集上的大量实验表明，SCL在多种跨物种评估协议下始终优于最先进的方法，并有效地泛化到以前未见过的物种和生态领域。代码可在 https://github.com/Kemalau/ECCV-26-SCL 获取。
cs.CV / 27 / 2609.09711
VFNet: Multi-View Spatio-Temporal Model for Void Fraction Estimation in Gas-Liquid Two-Phase Flow
VFNet：用于气液两相流中空隙率估计的多视角时空模型
Abstract
Void fraction, which quantifies the proportion of the fluid flow volume occupied by the gas phase, is a key parameter in the characterization of gas-liquid two-phase flow. Existing estimation methods either rely on flow assumptions that do not generalize across different fluids or on intrusive sensing that disturbs the flow behavior. We propose VFNet, a dual-branch spatio-temporal neural network for void-fraction prediction from synchronized multi-view videos of two-phase flow. A local branch extracts features from confined spatial regions and fuses the synchronized dual views, while a spatio-temporal branch captures the global evolution of the flow across space and time to refine a coarse geometric estimate. Trained on simulated computational fluid dynamics (CFD) data with known ground-truth void fractions and evaluated against both learning-based and traditional baselines, VFNet achieves the best performance across a broad range of metrics and also improves downstream flow-pattern classification on real two-phase flow data.
Chinese Translation
空隙率量化了气相占据流体流量体积的比例，是气液两相流特征化中的一个关键参数。现有的估计方法要么依赖于不适用于不同流体的流动假设，要么依赖于干扰流动行为的侵入式传感。我们提出了VFNet，一种双支路时空神经网络，用于从同步的多视角两相流视频中预测空隙率。局部支路从受限的空间区域提取特征并融合同步的双视图，而时空支路则捕捉流动在空间和时间上的全局演变，以细化粗略的几何估计。VFNet在已知真实空隙率的模拟计算流体动力学（CFD）数据上进行训练，并与基于学习的方法和传统基线进行评估，显示出在广泛的指标上实现了最佳性能，并且在真实的两相流数据上改善了下游流动模式分类。
cs.CV / 28 / 2609.09736
IAE-VTG: Interaction-Aligned Action-Entity Video Temporal Grounding
IAE-VTG：交互对齐的动作-实体视频时序定位
Abstract
Video Temporal Grounding (VTG) localizes the video segment that matches a natural-language query. Many queries describe an action performed by a particular entity. Existing methods often encode the query as a whole or use general video-text interactions, without explicitly checking whether the action and entity occur together. They may therefore select a segment that contains both concepts but not the event described by the query. We propose Interaction Aligned Action-Entity Video Temporal Grounding (IAE-VTG), which models this rela?tionship at both the representation and training assignment levels. First, the Fine-grained Disentangled Interaction Module (FDIM) separates action and entity related query information and aligns it with complementary motion and appearance features. It then combines token-level interactions to build representations that capture the relationship between the action and entity. Second, Interaction-Sensitive Assignment (ISA) adds this interaction evidence to bipartite matching, so training targets are selected using both temporal overlap and semantic compatibility. This reduces supervision from temporally plausible but semantically incorrect proposals. Experiments on QVHighlights, Charades?STA, and TACoS show that IAE-VTG consistently improves strong baselines and achieves competitive or state-of-the-art performance on standard grounding metrics. Additional analyses show that the method is especially effective when similar actions or entities appear at multiple times and produces more reliable assignments for complex events.
Chinese Translation
视频时序定位（VTG）旨在定位与自然语言查询相匹配的视频片段。许多查询描述了特定实体执行的动作。现有方法通常将查询整体编码或使用一般的视频-文本交互，而没有明确检查动作和实体是否同时出现。因此，它们可能选择一个包含这两个概念的片段，但并不符合查询所描述的事件。我们提出了交互对齐的动作-实体视频时序定位（IAE-VTG），该方法在表示和训练分配层面建模这种关系。首先，细粒度解耦交互模块（FDIM）分离与动作和实体相关的查询信息，并将其与互补的运动和外观特征对齐。然后，它结合令牌级交互构建捕捉动作与实体之间关系的表示。其次，交互敏感分配（ISA）将这种交互证据添加到二分匹配中，因此训练目标是通过时间重叠和语义兼容性共同选择的。这减少了来自时间上合理但语义上不正确的提案的监督。在QVHighlights、Charades-STA和TACoS上的实验表明，IAE-VTG始终改善强基线，并在标准定位指标上实现了具有竞争力或领先的性能。额外分析表明，该方法在相似动作或实体多次出现时尤其有效，并为复杂事件生成更可靠的分配。
cs.CV / 29 / 2609.09737
Distilling Image Prototypes for Guided Test-Time Adaptation
提取图像原型以指导测试时适应
Abstract
Test-Time Adaptation (TTA) enhances the robustness of models against distribution shifts but faces two critical challenges: error accumulation from noisy pseudo-labels and catastrophic forgetting of source knowledge. Uncertainty-based approaches designed to mitigate error accumulation often yield overconfident or computationally expensive estimates, while strategies intended to prevent forgetting via prototype replay rely on static representations that easily become misaligned as the model adapts. To address these issues, this paper proposes a novel framework, Distilling Image Prototype for Guided Test-Time Adaptation (DIPTTA). The core of the proposed approach is the introduction of a Distill Image Prototype (DIP), a compact set of synthetic images that serves as a dynamic and regenerative anchor of source knowledge. This prototype enables a dynamic feature replay mechanism that continuously generates feature prototypes aligned with the current state of the model, thus effectively preventing catastrophic forgetting. Furthermore, the DIP anchors a source-calibrated uncertainty estimation method, which provides a less biased measure of sample reliability by leveraging stable source knowledge, thereby robustly suppressing error accumulation. Extensive experiments on multiple benchmarks demonstrate that DIPTTA significantly outperforms state-of-the-art methods, particularly under severe domain shifts. The source code is available at https://github.com/LiwenWang919/DIPTTA.
Chinese Translation
测试时适应（TTA）增强了模型对分布变化的鲁棒性，但面临两个关键挑战：来自噪声伪标签的错误累积和源知识的灾难性遗忘。旨在减轻错误累积的不确定性基础方法往往会产生过于自信或计算开销大的估计，而旨在通过原型重放防止遗忘的策略则依赖于静态表示，这些表示在模型适应时容易失去对齐。为了解决这些问题，本文提出了一种新颖的框架——提取图像原型以指导测试时适应（DIPTTA）。该方法的核心是引入了提取图像原型（DIP），这是一个紧凑的合成图像集，作为源知识的动态和再生锚点。该原型使得动态特征重放机制成为可能，能够持续生成与模型当前状态对齐的特征原型，从而有效防止灾难性遗忘。此外，DIP还锚定了一种源校准的不确定性估计方法，通过利用稳定的源知识提供更少偏见的样本可靠性度量，从而稳健地抑制错误累积。在多个基准上的广泛实验表明，DIPTTA在严重领域转移的情况下显著优于最先进的方法。源代码可在 https://github.com/LiwenWang919/DIPTTA 获取。
cs.CV / 30 / 2609.09762
MethaneFuse: Learning from Multi-Sensor Satellite Observations for Methane Plume Detection
MethaneFuse：基于多传感器卫星观测学习甲烷羽流检测
Abstract
Methane plume detection from satellite imagery is constrained by incomplete observations: public satellites provide complementary spatial, spectral, and atmospheric evidence, but real plume cases rarely contain fully paired multi-sensor measurements because of revisit schedules, cloud coverage, acquisition quality, and the transient nature of emissions. Most learning-based detectors rely on single-sensor inputs, especially Sentinel-2 (S2), leaving many reported plume cases unusable. We construct MethaneUnion, a temporal multi-sensor dataset built from Carbon Mapper plume reports and matched S2, Landsat 8/9 (L8/9), EMIT, and Sentinel-5P (S5P) observations. Built on MethaneUnion, MethaneFuse learns from heterogeneous satellite observations under partial sensor availability without requiring complete four-sensor measurements. MethaneUnion expands usable coverage from 3,211 valid S2-matched plume cases to 8,981 reported plume cases with multi-sensor observations. At the representative 480 m setting, MethaneFuse achieves 84.87 F1 and 93.62 AUROC, improving over the strongest baseline by 5.65 F1 and 8.30 AUROC points while reducing false positives by 8.19 points. Sensor-availability experiments show that MethaneFuse improves detection when S2 is available and transfers plume knowledge to L8/9, EMIT, and S5P when S2 is unavailable. These results demonstrate the value of learning from incomplete heterogeneous sensor observations for practical methane plume detection.
Chinese Translation
从卫星图像中检测甲烷羽流受到不完整观测的限制：公共卫星提供互补的空间、光谱和大气证据，但由于重访计划、云覆盖、获取质量和排放的瞬态特性，真实的羽流案例很少包含完全配对的多传感器测量。大多数基于学习的检测器依赖于单传感器输入，特别是 Sentinel-2 (S2)，这使得许多报告的羽流案例无法使用。我们构建了 MethaneUnion，这是一个基于 Carbon Mapper 羽流报告和匹配的 S2、Landsat 8/9 (L8/9)、EMIT 和 Sentinel-5P (S5P) 观测的时间多传感器数据集。在 MethaneUnion 的基础上，MethaneFuse 在部分传感器可用的情况下，从异构卫星观测中学习，而无需完整的四传感器测量。MethaneUnion 将可用覆盖范围从 3,211 个有效的 S2 匹配羽流案例扩展到 8,981 个报告的多传感器观测羽流案例。在代表性的 480 米设置下，MethaneFuse 达到了 84.87 的 F1 和 93.62 的 AUROC，相较于最强基线提高了 5.65 的 F1 和 8.30 的 AUROC，同时将假阳性减少了 8.19 个点。传感器可用性实验表明，当 S2 可用时，MethaneFuse 提高了检测能力，并在 S2 不可用时将羽流知识转移到 L8/9、EMIT 和 S5P。这些结果展示了从不完整的异构传感器观测中学习对于实际甲烷羽流检测的价值。
cs.CV / 31 / 2609.09790
LogiScope-VQA: Benchmarking Vision-Language Models for Logistics Hazard Identification in Industrial Scenarios
LogiScope-VQA：工业场景中物流危险识别的视觉-语言模型基准测试
Abstract
Large Multimodal Models (LMMs) large-scale deployment in industrial warehouse settings specifically necessitates that models exhibit human-expert-level hazard-oriented perception, understanding, and reasoning capabilities. However, the scarcity of real industrial data, tightly coupled to commercial terms, significantly hampers further advancement. To bridge this gap, we curate LogiScope-VQA to investigate the practical applicability of mainstream LMMs in real-world logistics operations. LogiScope-VQA comprises 2,476 images and 2,918 videos primarily sourced from real-world logistics parks, along with 10,274 VQAs meticulously curated and validated by human annotators. Grounded in 18 core objects and 20 risk types, we devise 39 subtasks aligned with three principal themes: industrial element perception, warehouse knowledge understanding, and potential risk reasoning. Furthermore, we incorporate dynamic thinking-budget configurations and dual-dimensional risk bias analyses to elucidate the properties of LMMs. Extensive experiments unveil that even powerful proprietary models, including GPT-5.5, Gemini-3.1-Pro, and Claude-Opus-4.7, exhibit a significant gap relative to human performance. The unique challenge of jointly integrating perception, understanding, and reasoning for hazard identification poses substantial headroom for further improvement on LogiScope-VQA. We additionally reveal the pervasive security bias issue that impedes LLMs' practical deployment in real-world settings. The industrial dataset is publicly available under the CC BY-NC-SA 4.0 license.
Chinese Translation
大型多模态模型（LMMs）在工业仓库环境中的大规模部署，特别要求模型具备人类专家级别的危险导向感知、理解和推理能力。然而，真实工业数据的稀缺性，因其与商业条款紧密相关，显著阻碍了进一步的发展。为了解决这一问题，我们策划了LogiScope-VQA，以探讨主流LMMs在现实物流操作中的实际适用性。LogiScope-VQA包含2,476张图像和2,918段视频，主要来源于真实的物流园区，以及10,274个经过人类注释者精心策划和验证的视觉问答（VQA）。基于18个核心对象和20种风险类型，我们设计了39个子任务，围绕三个主要主题展开：工业元素感知、仓库知识理解和潜在风险推理。此外，我们还引入了动态思维预算配置和双维风险偏差分析，以阐明LMMs的特性。广泛的实验表明，即使是强大的专有模型，包括GPT-5.5、Gemini-3.1-Pro和Claude-Opus-4.7，相对于人类表现仍存在显著差距。联合整合感知、理解和推理以进行危险识别的独特挑战，为LogiScope-VQA的进一步改进提供了巨大的空间。我们还揭示了普遍存在的安全偏差问题，这阻碍了LLMs在现实环境中的实际部署。该工业数据集已在CC BY-NC-SA 4.0许可下公开发布。
cs.CV / 32 / 2609.09826
Freezing of Gait Prediction Under Spatial Occlusion: An IMU-Supervised Cross-Modal Distillation Approach
空间遮挡下步态冻结预测：一种IMU监督的跨模态蒸馏方法
Abstract
Parkinson's disease is a progressive neurodegenerative disorder characterised by gradual deterioration of movement control. Automated freezing-of-gait (FOG) detection supports the objective assessment of gait-related motor impairment. Two common approaches are used for FOG prediction: (i) analysing video recordings of the patient's movements and (ii) analysing data collected using inertial measurement unit (IMU) wearable sensors attached to the patient's lower limbs. Video-based approaches may suffer detection errors during continuous turning-in-place tasks because the lower limbs undergo substantial geometric self-occlusion, degrading pose-estimation accuracy. IMU-based approaches are generally less affected by visual occlusion; however, they are difficult to deploy outside clinical or laboratory settings, as the sensors must be attached securely and remain in place throughout the assessment. Motivated by this, we propose a cross-modal subspace distillation framework to mitigate the limitations of unimodal FOG detection by combining IMU accuracy with video-based practicality. We extract invariant latent topologies from a pre-trained kinematic oracle to structurally supervise a non-encoded visual architecture during training. To resolve periods of severe spatial occlusion, a dual-stream visual model probabilistically fuses skeletal graph nodes and continuous spatial pixels, dynamically shifting reliance to uninterrupted pixel boundaries as joint tracking confidence drops. Evaluated against a public, multi-modal sequence dataset of Parkinson's individuals executing continuous $360^\circ$ turns, empirical results demonstrate that applying sensory boundary topologies strictly mitigates tracking evaluation entropy. Our constrained optimisation confirms that highly precise FOG prediction bounds can be achieved over zero-wearable inference environments.
Chinese Translation
帕金森病是一种进行性神经退行性疾病，其特征是运动控制逐渐恶化。自动步态冻结（FOG）检测支持对与步态相关的运动障碍的客观评估。FOG预测通常采用两种方法：（i）分析患者运动的视频录制；（ii）分析附着在患者下肢的惯性测量单元（IMU）可穿戴传感器收集的数据。基于视频的方法在进行持续原地转身任务时可能会出现检测错误，因为下肢经历了显著的几何自遮挡，从而降低了姿态估计的准确性。基于IMU的方法通常不易受到视觉遮挡的影响；然而，它们在临床或实验室环境之外的部署较为困难，因为传感器必须牢固固定并在整个评估过程中保持原位。基于此，我们提出了一种跨模态子空间蒸馏框架，通过结合IMU的准确性与基于视频的方法的实用性，来减轻单模态FOG检测的局限性。我们从预训练的运动学预言机中提取不变的潜在拓扑结构，在训练过程中对未编码的视觉架构进行结构性监督。为了解决严重空间遮挡期间的问题，双流视觉模型以概率方式融合骨骼图节点和连续空间像素，动态地在关节跟踪信心下降时将依赖关系转移到不间断的像素边界。通过对执行连续$360^ extcirc$转身的帕金森患者的公共多模态序列数据集进行评估，实证结果表明，应用传感器边界拓扑严格减轻了跟踪评估熵。我们的约束优化确认，在零可穿戴推理环境中，可以实现高精度的FOG预测界限。
cs.CV / 33 / 2609.09827
Layerwise Tunable Lifting Scheme for the Convolutional Neural Network
可调层次提升方案用于卷积神经网络
Abstract
This work introduces a family of tunable lifting schemes for biorthogonal wavelet filter banks. We propose three lifting strategies: low-pass tuning (LS-LayLatt-LP), high-pass tuning (LS-LayLatt-HP), and a sequential lifting scheme that jointly adapts low- and high-frequency branches (LS-LayLatt-Sequential). All proposed designs are formulated using a lattice-based lifting structure, which guarantees invertibility and stability for arbitrary parameter values within the lifting functions. We evaluated the proposed methods by integrating them into a ResNet-18 backbone for image classification on the Describable Textures Dataset (DTD), as well as for anomaly detection on hazelnut images from the MVTec-AD dataset and private KRC102S dataset. Experimental results demonstrate consistent performance improvements across all evaluated tasks.
Chinese Translation
本研究介绍了一系列可调的提升方案，用于双正交小波滤波器组。我们提出了三种提升策略：低通调节（LS-LayLatt-LP）、高通调节（LS-LayLatt-HP）以及一种联合适应低频和高频分支的顺序提升方案（LS-LayLatt-Sequential）。所有提出的设计均采用基于格子的提升结构进行构建，这保证了在提升函数内任意参数值下的可逆性和稳定性。我们通过将这些方法集成到ResNet-18骨干网络中，评估了其在描述性纹理数据集（DTD）上的图像分类性能，以及在MVTec-AD数据集和私有KRC102S数据集上的榛子图像异常检测性能。实验结果表明，在所有评估任务中，性能均有一致的提升。
cs.CV / 34 / 2609.09832
SkNeXt enables topology-guided neuronal reconstruction from petabyte-scale microscopy data
SkNeXt实现了基于拓扑引导的神经元重建，适用于PB级显微镜数据
Abstract
Recent advances in high-resolution fluorescence and electron microscopy have enabled nanoscale imaging across increasingly large brain volumes, but the resulting terabyte- to petabyte-scale datasets make complete neuronal reconstruction prohibitively expensive in computation, data movement, and manual proofreading. Here, we present SkNeXt, a topology-first framework for scalable neuronal reconstruction from large volumetric microscopy datasets. Instead of densely processing entire image volumes, SkNeXt first converts neuronal morphology into compact SWC skeletons that preserve long-range connectivity. Proofreading is therefore focused on sparse neuronal trees, allowing branch, continuity, and connectivity errors to be corrected before high-resolution reconstruction. The corrected skeletons then serve as persistent structural priors for recovering detailed morphology while preserving neuronal identity and topology. Crucially, SkNeXt also uses neuronal skeletons as spatial indices for selective data access, retrieving high-resolution image regions only along reconstructed trajectories and bypassing most background and signal-free volumes. This substantially reduces I/O and computational overhead, allowing reconstruction cost to scale with neuronal morphology rather than total dataset size. Using SkNeXt, we reconstructed neurons from a petabyte-scale super-resolution fluorescence dataset of the mouse brain on a single GPU within one week, without requiring exhaustive dense inference across the complete imaging volume.
Chinese Translation
近年来，高分辨率荧光和电子显微镜的进展使得在越来越大脑体积中进行纳米级成像成为可能，但由此产生的TB到PB级数据集使得完整的神经元重建在计算、数据传输和人工校对方面变得极为昂贵。在此，我们提出了SkNeXt，一个基于拓扑优先的框架，用于从大体积显微镜数据集中进行可扩展的神经元重建。SkNeXt并不是对整个图像体积进行密集处理，而是首先将神经元形态转换为紧凑的SWC骨架，以保留长程连接。因此，校对工作集中在稀疏的神经元树上，允许在高分辨率重建之前纠正分支、连续性和连接性错误。经过修正的骨架随后作为持久的结构先验，用于恢复详细的形态，同时保留神经元的身份和拓扑。至关重要的是，SkNeXt还将神经元骨架用作空间索引，以实现选择性的数据访问，仅沿重建的轨迹检索高分辨率图像区域，从而绕过大部分背景和无信号体积。这大大减少了I/O和计算开销，使重建成本与神经元形态的复杂性而非总数据集大小成比例。使用SkNeXt，我们在一周内在单个GPU上从小鼠大脑的PB级超分辨率荧光数据集中重建了神经元，而无需在整个成像体积上进行全面的密集推断。
cs.CV / 35 / 2609.09863
Pretraining and Distillation Matter More Than Architecture Family for Label-Free Single-Cell Classification
预训练和蒸馏比架构家族对无标签单细胞分类更为重要
Abstract
Choosing a deep learning architecture for label-free single-cell classification remains an open question, with microscopy benchmarks reporting conflicting conclusions about CNNs versus transformers. We present a controlled benchmark on LIVECell phase-contrast microscopy data using source-image-disjoint train/validation/test splits to prevent parent-image leakage and matched optimisation, augmentation, and evaluation protocols across EfficientNet, Vision Transformer (ViT), and EVA-02 models. This allows the effects of architecture, pretraining, fine-tuning, tokenisation, and distillation to be disentangled. We find that the previously reported CNN advantage is largely explained by pretraining rather than architecture: the smallest pretrained model outperforms the strongest model trained from scratch despite far fewer parameters. Pretraining improves macro-F1 by 3-4 points, while the gap between the best pretrained CNN and transformer is below 0.5 points. Architectural choices nevertheless matter: ViT-S/8 outperforms ViT-S/16 and matches the four-times-larger ViT-B/16 at a quarter of the parameters, showing that finer tokenisation benefits small cell crops. Conversely, layer-wise learning-rate decay, central to the EVA-02 fine-tuning recipe, degrades performance, highlighting that transfer heuristics from natural-image recognition may not generalise to microscopy. Finally, knowledge distillation substantially improves the deployment frontier: compact EfficientNet-B0 students distilled from teacher councils outperform every individually trained backbone, including the EfficientNet-B5 and EVA-02 teachers. Overall, our results show that rigorous control of pretraining and evaluation is essential for interpreting biomedical architecture benchmarks, while distillation may be a more effective route to practical single-cell classification than architecture choice alone.
Chinese Translation
选择用于无标签单细胞分类的深度学习架构仍然是一个未解的问题，显微镜基准测试对卷积神经网络（CNN）与变换器（transformers）之间的比较报告了相互矛盾的结论。我们在LIVECell相位对比显微镜数据上进行了受控基准测试，使用源图像不重叠的训练/验证/测试分割，以防止父图像泄漏，并在EfficientNet、视觉变换器（Vision Transformer, ViT）和EVA-02模型之间匹配优化、增强和评估协议。这使得架构、预训练、微调、标记化和蒸馏的影响得以分离。我们发现，之前报告的CNN优势主要是由预训练解释的，而非架构：最小的预训练模型在参数数量远少于从头训练的最强模型的情况下表现更好。预训练使宏观F1分数提高了3-4个点，而最佳预训练CNN与变换器之间的差距低于0.5个点。然而，架构选择仍然重要：ViT-S/8的表现优于ViT-S/16，并且在参数数量仅为ViT-B/16四分之一的情况下与其相匹配，显示出更细的标记化对小细胞裁剪有益。相反，逐层学习率衰减（EVA-02微调方案的核心）会降低性能，突显出来自自然图像识别的迁移启发式可能无法推广到显微镜图像。最后，知识蒸馏显著改善了部署边界：从教师模型蒸馏出的紧凑型EfficientNet-B0学生模型的表现超越了每个单独训练的主干，包括EfficientNet-B5和EVA-02教师。总体而言，我们的结果表明，严格控制预训练和评估对于解释生物医学架构基准测试至关重要，而蒸馏可能比单纯选择架构更有效地实现实际的单细胞分类。
cs.CV / 36 / 2609.09876
From Pixels to Hierarchical Sequences: Quadtree Mask Encoding for Vision-Language Binary Change Detection
从像素到层次序列：用于视觉-语言二元变化检测的四叉树掩码编码
Abstract
Dense change detection in remote sensing requires vision-language models (VLMs) to compare bi-temporal images and generate accurate pixel-level masks. Existing VLMs are largely confined to change captioning outputs, and the few that produce pixel-level masks still rely on external decoders or flat text-as-mask serialization, which are less effective for small and fragmented changes. We introduce QUAKE-CD, a framework that recasts dense change prediction as syntax-verifiable structured generation. QUAKE-CD represents binary change masks as grammar-constrained quadtree token sequences, making the masks compact, syntactically checkable, and deterministically decodable within an autoregressive generation space. We further construct QUAKE-CoT, which pairs these sequences with chain-of-thought traces grounded in visual evidence, and jointly optimizes textual reasoning and spatial dense prediction through a progressive curriculum followed by grammar-gated dual-reward RL. On QUAKE-CoT, QUAKE-CD achieves 78.31% accumulated F1, outperforming decoder-based and flat text-as-mask VLMs while producing more faithful bi-temporal reasoning.
Chinese Translation
遥感中的密集变化检测需要视觉-语言模型（VLMs）比较双时相图像并生成准确的像素级掩码。现有的VLMs大多局限于变化描述输出，而少数能够生成像素级掩码的模型仍依赖于外部解码器或平面文本作为掩码的序列化，这对于小而零散的变化效果较差。我们提出了QUAKE-CD，一个将密集变化预测重新表述为语法可验证的结构化生成的框架。QUAKE-CD将二元变化掩码表示为语法约束的四叉树令牌序列，使掩码紧凑、语法可检查，并在自回归生成空间内可确定解码。我们进一步构建了QUAKE-CoT，将这些序列与基于视觉证据的思维链迹相结合，通过逐步课程和语法门控双重奖励强化学习共同优化文本推理和空间密集预测。在QUAKE-CoT上，QUAKE-CD实现了78.31%的累积F1，超越了基于解码器和平面文本作为掩码的VLMs，同时产生了更为真实的双时相推理。
cs.CV / 37 / 2609.09881
CLFTv2: Efficient Camera-LiDAR Fusion for Semantic Segmentation via Hierarchical Feature Pyramids
CLFTv2：通过层次特征金字塔实现高效的相机-激光雷达融合用于语义分割
Abstract
Semantic segmentation for autonomous driving requires reliable detection of vulnerable road users (VRUs) despite heavy class imbalance. We introduce CLFTv2, a hierarchical camera-LiDAR fusion framework replacing global ViT attention with a Swin-based multi-scale encoder and a lightweight FPN-style residual decoder. Operating in the 2D perspective domain, CLFTv2 integrates multi-scale geometric cues through shifted-window attention and per-scale residual fusion, avoiding the computational overhead of query-matching decoders. Across three driving datasets, CLFTv2 consistently improves VRU recall. On ZOD, CLFTv2-Large achieves 53.5\% mIoU, improving pedestrian IoU from 35.5\% to 44.9\% over the prior CLFT model. On Waymo, CLFTv2 reaches 61.7\% mIoU. Additionally, a modality-isolation study suggests ViT's global receptive field yields stronger fusion gains only under dense LiDAR returns. Compared to a Swin-based Mask2Former adaptation, CLFTv2 requires 1.4$\times$ fewer GFLOPs and delivers 2.2$\times$ higher throughput, while achieving comparable overall accuracy. These results demonstrate that hierarchical local-attention fusion offers an efficient, scalable alternative to global-attention and query-based decoders for real-time on-vehicle perception in intelligent transportation systems. Source code is publicly available.
Chinese Translation
自动驾驶的语义分割需要可靠地检测脆弱道路用户（VRUs），尽管存在严重的类别不平衡。我们提出了CLFTv2，一个层次化的相机-激光雷达融合框架，用Swin基础的多尺度编码器替代了全局ViT注意力，并采用轻量级的FPN风格残差解码器。CLFTv2在二维视角域内运行，通过移位窗口注意力和每尺度残差融合集成多尺度几何线索，避免了查询匹配解码器的计算开销。在三个驾驶数据集上，CLFTv2始终提高了VRU的召回率。在ZOD数据集上，CLFTv2-Large达到了53.5%的mIoU，相比于之前的CLFT模型，行人IoU从35.5%提高到44.9%。在Waymo数据集上，CLFTv2达到了61.7%的mIoU。此外，模态隔离研究表明，ViT的全局感受野仅在密集的激光雷达返回下产生更强的融合增益。与基于Swin的Mask2Former适配相比，CLFTv2所需的GFLOPs减少了1.4倍，并且提供了2.2倍的更高吞吐量，同时实现了相当的整体准确性。这些结果表明，层次化局部注意力融合为智能交通系统中的实时车载感知提供了一种高效、可扩展的替代方案，取代了全局注意力和基于查询的解码器。源代码已公开。
cs.CV / 38 / 2609.09884
Albedo Estimation via Latent Bridge Matching
通过潜在桥匹配进行反照率估计
Abstract
Recent advances in Intrinsic Image Decomposition (IID) have increasingly relied on generative models. However, progress remains limited by three key challenges: (a) insufficient physical consistency, (b) high computational cost at inference time, and (c) limited generalization capabilities. In this work, we show that latent bridge matching (LBM) effectively addresses these limitations for albedo estimation. We introduce a novel LBM-based architecture that enforces physical consistency through a pixel reconstruction loss, benefits from the inherent efficiency of LBM low-cost inference, and improves generalization across diverse datasets by incorporating a shading conditioning. In this extended version, we additionally show that conditioning the shading estimator itself on the predicted albedo further improves reconstruction fidelity, and we benchmark our best model against stateof-the-art IID methods across five real and synthetic datasets.
Chinese Translation
近年来，内在图像分解（Intrinsic Image Decomposition, IID）的进展越来越依赖于生成模型。然而，进展仍然受到三个关键挑战的限制：（a）物理一致性不足，（b）推理时的高计算成本，以及（c）有限的泛化能力。在本研究中，我们展示了潜在桥匹配（Latent Bridge Matching, LBM）有效地解决了反照率估计的这些局限性。我们引入了一种新颖的基于LBM的架构，通过像素重建损失强制物理一致性，利用LBM低成本推理的内在效率，并通过引入阴影条件改善在多样数据集上的泛化能力。在这个扩展版本中，我们还展示了将阴影估计器本身基于预测的反照率进行条件化进一步提高了重建的保真度，并且我们在五个真实和合成数据集上对我们最好的模型进行了与最先进的IID方法的基准测试。
cs.CV / 39 / 2609.09890
StreetDiff: Multi-view Street Scenes Generation via Cross-view Consistent Multi-view Stable Diffusion with Structure Prompts
StreetDiff：通过结构提示的跨视角一致性多视角稳定扩散生成街景
Abstract
Multi-view diffusion models have shown strong performance in scenes with strong geometric priors and sparse semantics, such as indoor rooms or simple outdoor environments (e.g., fields, courtyards). However, they often fail to maintain cross-view consistency under camera rotation, especially in structurally complex urban environments. Without explicit modeling of spherical correspondence across views, existing approaches tend to produce object duplication, structural distortion, and layout inconsistency. To address this limitation, we propose StreetDiff, a multi-view diffusion framework that explicitly enforces cross-view alignment during denoising. StreetDiff introduces a Panorama--Perspective Synergy design to decouple global layout reasoning from local detail synthesis, and incorporates a Panorama Alignment Module (PAM) that establishes spherical-projection-based attention constraints across views. By injecting structured alignment constraints without modifying the diffusion backbone, our framework achieves robust cross-view coherence in challenging urban street scene generation tasks. In addition, we construct Street360, a large-scale HDR multi-view urban panorama dataset. Extensive experiments demonstrate that StreetDiff significantly improves structural consistency and visual fidelity compared to prior multi-view diffusion generation methods.
Chinese Translation
多视角扩散模型在具有强几何先验和稀疏语义的场景中表现出色，例如室内房间或简单的户外环境（如田野、庭院）。然而，它们在相机旋转下往往无法保持跨视角一致性，尤其是在结构复杂的城市环境中。如果不明确建模视角间的球面对应关系，现有方法往往会导致物体重复、结构扭曲和布局不一致。为了解决这一限制，我们提出了StreetDiff，一种在去噪过程中明确强制跨视角对齐的多视角扩散框架。StreetDiff引入了一种全景-透视协同设计，将全局布局推理与局部细节合成解耦，并结合了全景对齐模块（Panorama Alignment Module, PAM），该模块在视角间建立基于球面投影的注意力约束。通过注入结构化对齐约束而不修改扩散主干，我们的框架在具有挑战性的城市街景生成任务中实现了稳健的跨视角一致性。此外，我们构建了Street360，一个大规模HDR多视角城市全景数据集。大量实验表明，与之前的多视角扩散生成方法相比，StreetDiff显著提高了结构一致性和视觉真实感。
cs.CV / 40 / 2609.09895
Can We Trust Video Hallucination Detectors? VidHalLoc for Evaluating the Evaluators
我们能信任视频幻觉检测器吗？VidHalLoc用于评估评估者
Abstract
Video-language models and video agents can produce hallucinations that conflict with spatiotemporal evidence. Existing benchmarks mainly evaluate model hallucinations, and heterogeneous mechanisms make detector reliability difficult to compare. We introduce VidHalLoc, a benchmark that evaluates hallucination detection methods under a unified diagnostic evaluation protocol using 2,000 adversarial hallucination samples across Video Question Answering and Video Captioning tasks, spanning Ontology and Dynamic hallucination categories. To construct VidHalLoc efficiently, we introduce VideoHALO, a Harness Engineering-informed multi-agent workflow that decomposes data construction into four executable stages supported by a memory system and a communication protocol. Evaluation of fifteen methods reveals that the four dedicated detectors peak at an Overall accuracy of only 34.63%, indicating limited reliability across video hallucination types [Dataset Repository: https://huggingface.co/datasets/wesfggfd/VidHalLoc].
Chinese Translation
视频语言模型和视频代理可能会产生与时空证据相矛盾的幻觉。现有基准主要评估模型的幻觉，而异构机制使得检测器的可靠性难以比较。我们引入了VidHalLoc，这是一个在统一诊断评估协议下评估幻觉检测方法的基准，使用了2,000个对抗性幻觉样本，涵盖视频问答和视频字幕任务，涉及本体（Ontology）和动态（Dynamic）幻觉类别。为了高效构建VidHalLoc，我们引入了VideoHALO，这是一种基于Harness Engineering的多代理工作流程，将数据构建分解为四个可执行阶段，并由内存系统和通信协议支持。对十五种方法的评估表明，四个专用检测器的整体准确率仅达到34.63%，这表明在不同类型的视频幻觉中可靠性有限。[数据集库：https://huggingface.co/datasets/wesfggfd/VidHalLoc]
cs.CV / 41 / 2609.09909
Interpreting Object-Dependent Concept Brittleness in Text-to-Image Diffusion Models
文本到图像扩散模型中对象依赖概念脆弱性的解释
Abstract
Although text-to-image diffusion models generally exhibit strong prompt-following ability, we identify a persistent and previously underexplored failure pattern in which a small subset of prompts differing only in the object consistently fails to realize the same target concept under identical generation settings. We term this phenomenon object-dependent concept brittleness. Such cases suggest systematic internal blind spots rather than random sampling noise. In this paper, we present an interpretability-oriented framework to audit and minimally correct these failures. Our key idea is to analyze denoising trajectories in a step-wise sparse autoencoder (SAE) space, where abstract style and attribute concepts become more separable than in the raw denoising representation. This sparse space enables us to compare successful and failed generations, identify concept dimensions whose evidence is missing, weakened, or temporally delayed, and construct class-level concept prototypes from reliable class-consistent samples. Based on this audit process, we introduce a lightweight inference-time correction strategy that interpolates denoising features toward the corresponding prototype in SAE space. Rather than serving as a task-specific retraining method, this intervention acts as a validation of the diagnosed concept deficiency. We evaluate the proposed framework on style and attribute failure cases across multiple diffusion backbones, with significant improvements in concept consistency, text fidelity, and repair success. Further analyses show that deeper denoising representations provide clearer concept structure, while early-stage intervention offers the strongest correction leverage. Code is available at https://github.com/Metecade/Object-Dependent-Concept-Brittleness.
Chinese Translation
尽管文本到图像扩散模型通常表现出强大的提示跟随能力，但我们识别出一种持久且之前未被充分探索的失败模式，即在相同生成设置下，仅在对象上有所不同的小部分提示始终无法实现相同的目标概念。我们将这种现象称为对象依赖概念脆弱性。这种情况表明存在系统性的内部盲点，而非随机采样噪声。本文提出了一种以可解释性为导向的框架，以审计和最小化这些失败。我们的关键思想是分析在逐步稀疏自编码器（SAE）空间中的去噪轨迹，在该空间中，抽象风格和属性概念比在原始去噪表示中更具可分离性。这个稀疏空间使我们能够比较成功和失败的生成，识别证据缺失、减弱或时间延迟的概念维度，并从可靠的类一致样本中构建类级概念原型。基于这一审计过程，我们引入了一种轻量级推理时纠正策略，该策略将去噪特征插值到SAE空间中相应的原型。该干预并不是作为特定任务的再训练方法，而是对诊断出的概念缺陷的验证。我们在多个扩散基础模型上评估了所提出的框架，针对风格和属性失败案例，显著提高了概念一致性、文本忠实度和修复成功率。进一步分析表明，深层去噪表示提供了更清晰的概念结构，而早期干预则提供了最强的纠正杠杆。代码可在 https://github.com/Metecade/Object-Dependent-Concept-Brittleness 获取。
cs.CV / 42 / 2609.09924
Multimodal Emotion Recognition in Conversations via Class-Wise Adaptive Modality Fusion and Affective Geometry
通过类别自适应模态融合和情感几何实现对话中的多模态情感识别
Abstract
Emotion Recognition in Conversations (ERC) requires integrating heterogeneous textual, audio, and visual cues while accounting for conversational context and emotional dynamics. We extend the Self-Distillation Transformer architecture for ERC with appearance+geometry visual representations, class-wise adaptive modality fusion, and a valence-arousal prior for affective transitions. On the MELD and IEMOCAP datasets, geometry-enhanced visual representations improve weighted F1 by 0.27 and 4.36 points over appearance-only features, respectively, while class-wise adaptive fusion provides further gains of 0.17 and 0.25 points over the original softmax gate. The valence-arousal prior yields targeted improvements of 0.30 and 0.74 accuracy points on emotionally shifted utterances while preserving performance on stable turns. These results indicate that structured facial cues, emotion-dependent modality weighting, and affective geometry provide complementary benefits for multimodal ERC.
Chinese Translation
对话中的情感识别（ERC）需要整合异质的文本、音频和视觉线索，同时考虑对话上下文和情感动态。我们扩展了用于ERC的自蒸馏变换器架构，结合了外观+几何视觉表示、类别自适应模态融合以及用于情感转变的效价-唤醒先验。在MELD和IEMOCAP数据集上，增强几何的视觉表示相比仅使用外观特征分别提高了加权F1分数0.27和4.36点，而类别自适应融合在原始softmax门的基础上又进一步提高了0.17和0.25点。效价-唤醒先验在情感转变的发言上实现了0.30和0.74的准确率提升，同时保持了在稳定发言上的表现。这些结果表明，结构化的面部线索、情感依赖的模态加权和情感几何为多模态ERC提供了互补的优势。
cs.CV / 43 / 2609.09949
Vague2Detect: Handling Ambiguous Prompts in Knowledge-Based Open-World Detection
Vague2Detect：处理基于知识的开放世界检测中的模糊提示
Abstract
Real-world detectors must often interpret functional or ambiguous prompts, yet conventional models such as YOLO remain restricted to fixed class lists. Even open-vocabulary models like YOLO-World frequently misalign vague language with the intended objects. Building on our prior work Commonsense-Guided Open-World Object Detection Using LLMs and Visual-Semantic Matching, we address YOLO-World's limitations in grounding task-driven queries. We propose Vague2Detect, a hybrid pipeline in which a fine-tuned Sentence-BERT retrieves candidates from a structured household Knowledge Base (KB), and YOLO-World verifies their presence in the image. For prompts outside the KB, a large language model (GPT-3.5-turbo) generates candidate descriptions, dynamically expanding the KB to cover novel concepts. On a benchmark of household scenes using custom images and an Open Images V7 subset, YOLO-World alone achieves only 32% Vague Prompt Success Rate (VPSR), the ability to map ambiguous queries to correct detections. In contrast, Vague2Detect improves performance to 61% VPSR with high precision, and up to 85% when augmented with GPT fallback.
Chinese Translation
现实世界的检测器往往需要解释功能性或模糊的提示，然而传统模型如YOLO仍然限制于固定的类别列表。即使是像YOLO-World这样的开放词汇模型，常常也会将模糊语言与预期对象错误对齐。在我们之前的工作《基于常识的开放世界物体检测：使用大型语言模型和视觉-语义匹配》的基础上，我们解决了YOLO-World在基础任务驱动查询中的局限性。我们提出了Vague2Detect，这是一种混合管道，其中经过微调的Sentence-BERT从结构化的家庭知识库（Knowledge Base, KB）中检索候选项，而YOLO-World则验证它们在图像中的存在。对于知识库之外的提示，一个大型语言模型（GPT-3.5-turbo）生成候选描述，动态扩展知识库以涵盖新概念。在使用自定义图像和Open Images V7子集的家庭场景基准测试中，仅使用YOLO-World的模糊提示成功率（Vague Prompt Success Rate, VPSR）仅为32%，即将模糊查询映射到正确检测的能力。相比之下，Vague2Detect的性能提高至61%的VPSR，并且具有高精度，当与GPT回退结合时，VPSR可高达85%。
cs.CV / 44 / 2609.09973
Putting Captions to the Test: Evaluating Video Caption Quality through Multiple-Choice Question Answering
对字幕进行测试：通过多项选择题回答评估视频字幕质量
Abstract
Evaluating video captioning remains a critical challenge for Visual Large Language Models (VLLMs). Existing metrics primarily rely on matching generated text against ground-truth references. This paradigm suffers from the ``one-to-many'' nature of video description, where high-quality captions are often penalized for lexical mismatches or valid shifts in visual focus. Furthermore, such assessments are typically one-dimensional, failing to provide a fine-grained analysis of caption quality. To address this, we redefine caption quality through the lens of information fidelity: A caption must maximize the coverage of salient visual information while ensuring strict factuality. We introduce CapQuiz, a novel reference-free benchmark that assesses captions based on their utility in answering human-verified, fine-grained, multiple-choice questions derived from the video. CapQuiz features a hierarchical taxonomy of 10 question types (spanning Descriptive and Inferential categories) across 24 diverse video domains. Extensive experiments demonstrate that CapQuiz correlates significantly better with human judgments than existing metrics and offers interpretable insights into model performance.
Chinese Translation
评估视频字幕仍然是视觉大型语言模型（VLLMs）面临的一项关键挑战。现有的评估指标主要依赖于将生成的文本与真实参考进行匹配。这一范式受到视频描述的“多对一”特性的影响，高质量的字幕往往因词汇不匹配或视觉焦点的合理变化而受到惩罚。此外，这种评估通常是一维的，未能提供对字幕质量的细致分析。为了解决这个问题，我们通过信息保真度的视角重新定义字幕质量：字幕必须最大化显著视觉信息的覆盖，同时确保严格的事实准确性。我们引入了CapQuiz，这是一个新颖的无参考基准，基于字幕在回答从视频中提取的人类验证的细粒度多项选择题中的实用性进行评估。CapQuiz包含10种问题类型的分层分类（涵盖描述性和推理性类别），横跨24个不同的视频领域。大量实验表明，CapQuiz与人类判断的相关性显著优于现有指标，并提供了对模型性能的可解释性洞察。
cs.CV / 45 / 2609.10001
From Few-Shot Segmentation to Clinician-in-the-Loop Medical Image Analysis
从少样本分割到临床医生参与的医学图像分析
Abstract
Few-shot medical image segmentation (FSMIS) seeks to delineate unseen structures from a small support set, but its standard formulation fixes task-defining evidence before inference. This assumption is fragile when query cases exhibit acquisition shift, atypical pathology, ambiguous boundaries, or poor image quality. Prototype learning, cross-domain matching, interactive segmentation, uncertainty estimation, test-time adaptation, and promptable foundation models address parts of this problem, yet have not been jointly evaluated under a common model of expert attention and clinical risk. This Perspective reframes FSMIS as a sequential clinician-model decision problem with a static support budget $K$ and a distinct interaction budget $B$. At each step, a system accepts the current segmentation, requests feedback, or defers to full expert review. Queries vary in location and modality and are selected by response-conditioned net expected value of information; clinician-provided feedback informs bounded adaptation only after prespecified provenance, consistency, and safety gates. The framework separates distributional atypicality from predicted clinical failure and treats clinician responses as informative but fallible observations. We synthesize the transition from few-shot and cross-domain segmentation to interactive and selective adaptation, delineate the integration gap, and define four research directions with falsifiable hypotheses. Evaluation spans external-domain calibration, quality-effort trade-offs, reader studies, and prospective workflow assessment. The central claim is not that interaction alone resolves domain shift, but that scarce expert attention should be allocated only when it is expected to reduce clinically relevant risk.
Chinese Translation
少样本医学图像分割（FSMIS）旨在从小规模支持集划分未见结构，但其标准形式在推理前固定了任务定义证据。当查询案例表现出获取偏移、非典型病理、模糊边界或图像质量差时，这一假设显得脆弱。原型学习、跨领域匹配、交互式分割、不确定性估计、测试时适应和可提示基础模型解决了这一问题的部分内容，但尚未在专家注意力和临床风险的共同模型下进行联合评估。本文将FSMIS重新框定为一个具有静态支持预算$K$和独特交互预算$B$的顺序临床医生-模型决策问题。在每一步，系统接受当前的分割结果、请求反馈或推迟至全面专家审查。查询在位置和模态上各不相同，并通过响应条件的净期望信息价值进行选择；临床医生提供的反馈仅在预先指定的来源、一致性和安全性门槛后，才会用于有限适应。该框架将分布异常性与预测的临床失败分开，并将临床医生的反应视为信息丰富但可能出错的观察。我们综合了从少样本和跨领域分割到交互式和选择性适应的过渡，阐明了整合差距，并定义了四个具有可证伪假设的研究方向。评估范围包括外部领域校准、质量-努力权衡、读者研究和前瞻性工作流程评估。核心论点并非仅仅通过交互解决领域偏移，而是稀缺的专家注意力应仅在预期能降低临床相关风险时分配。
cs.CV / 46 / 2609.10002
What Makes Adversarial Examples Transfer Across Deepfake Detectors?
是什么导致对抗样本在深伪检测器之间转移？
Abstract
Deepfake detectors remain vulnerable to transfer-based black-box attacks, in which adversarial examples are generated on a source surrogate model and transferred to a target model, unknown to the attacker. Yet how source--target compatibility shapes attack success remains poorly understood. Prior studies evaluate limited detector pools and rarely disentangle architectural from training factors. We conduct a controlled evaluation of adversarial transferability across 60 detectors spanning six backbones, two pretraining regimes, and five training-data configurations, using two attack procedures: AutoAttack (AA) and the Carlini--Wagner attack with Expectation over Transformation (CW--EOT). Matched comparisons reveal significantly higher transfer when source and target share an exact backbone, architecture family, pretraining regime, or training data. This compatibility structure is attack-dependent: exact backbone compatibility has the largest effect under AA, whereas shared pretraining and training data have the largest effects under CW--EOT. When transfer is averaged across non-target sources, mean attack success rate (ASR) is $7.21\%$ under AA and $19.52\%$ under CW--EOT. By contrast, a multi-source oracle combining both attacks attains a \(64.48\%\) mean ASR after excluding exact backbone and training-data matches, showing that source averaging can substantially understate target vulnerability. We release 240,000 adversarially perturbed images, complete pairwise transfer results, detector configurations, and evaluation code. These findings establish source--target compatibility and source-model selection as central dimensions of credible transfer-based black-box robustness evaluation.
Chinese Translation
深伪检测器仍然容易受到基于转移的黑箱攻击，其中对抗样本在源代理模型上生成，并转移到攻击者未知的目标模型。然而，源-目标兼容性如何影响攻击成功率仍然不甚了解。以往研究评估的检测器池有限，且很少将架构因素与训练因素区分开来。我们对跨越六种主干网络、两种预训练方案和五种训练数据配置的60个检测器进行了对抗转移性的控制评估，使用了两种攻击程序：AutoAttack (AA) 和 Carlini-Wagner 攻击与期望变换 (CW-EOT)。匹配比较显示，当源和目标共享相同的主干网络、架构家族、预训练方案或训练数据时，转移效果显著提高。这种兼容性结构依赖于攻击类型：在AA下，相同主干网络的兼容性影响最大，而在CW-EOT下，共享的预训练和训练数据影响最大。当对非目标源的转移进行平均时，AA下的平均攻击成功率 (ASR) 为7.21%，CW-EOT下为19.52%。相比之下，结合两种攻击的多源oracle在排除相同主干网络和训练数据匹配后达到了64.48%的平均ASR，显示源平均可能显著低估目标的脆弱性。我们发布了240,000张对抗性扰动图像、完整的成对转移结果、检测器配置和评估代码。这些发现确立了源-目标兼容性和源模型选择作为可信的基于转移的黑箱鲁棒性评估的核心维度。
cs.CV / 47 / 2609.10008
Beyond Similarity: Foundation Models as an Efficient Backbone for Training-Free Composed Video Retrieval
超越相似性：基础模型作为无训练组合视频检索的高效骨干
Abstract
Composed video retrieval (CoVR) searches a gallery for the target video that realizes a natural-language modification of a source clip. However, at gallery scale, this creates a fundamental tension: compact embeddings enable efficient, reusable search but can miss the transient actions, state changes, and subtle constraints that demand fine-grained video reasoning, whereas applying large multimodal models uniformly sacrifices scalability. To address these limitations, we propose that frozen foundation models should instead occupy complementary roles, with inference depth adapted to query difficulty. Based on this premise, we introduce \methodname{}, a framework for training-free \methodexpansion{}. Specifically, a composed-query embedding first searches reusable video-only gallery representations; uncertain queries undergo bounded reranking and candidate expansion; ambiguous edits trigger target-description generation; and only close leading candidates reach multimodal verification. To support these roles, frame selection, spatial resolution, and time cues are adapted to each stage. Across complete target-gallery evaluations, our method reaches state-of-the-art performance among training-free approaches, with 89.55 and 93.43 R@1 on Dense-WebVid-CoVR and CoVR-R, respectively (with more than +35\% and +25\% absolute margins to the closest counterpart). These results show that adaptively orchestrating foundation-model capabilities can combine scalable retrieval with fine-grained reasoning without task-specific training. The source code and all relevant guidelines are available on https://github.com/demidovd98/CoVRAGE.
Chinese Translation
组合视频检索（CoVR）在图库中搜索实现源视频片段自然语言修改的目标视频。然而，在图库规模下，这产生了根本性的矛盾：紧凑的嵌入能够实现高效、可重用的搜索，但可能会遗漏瞬态动作、状态变化和需要细粒度视频推理的微妙约束，而均匀应用大型多模态模型则牺牲了可扩展性。为了解决这些限制，我们提出冷冻基础模型应占据互补角色，推理深度根据查询难度进行调整。基于这一前提，我们引入了 extit{methodname}，一个无训练的 extit{methodexpansion}框架。具体而言，组合查询嵌入首先搜索可重用的视频图库表示；不确定的查询经过有限的重新排序和候选扩展；模糊的编辑触发目标描述生成；只有接近的主要候选才会进行多模态验证。为了支持这些角色，帧选择、空间分辨率和时间线索在每个阶段进行调整。在完整的目标图库评估中，我们的方法在无训练方法中达到了最先进的性能，在Dense-WebVid-CoVR和CoVR-R上分别达到了89.55和93.43的R@1（与最近的对手相比，绝对优势超过+35 ext{%}和+25 ext{%}）。这些结果表明，自适应协调基础模型能力可以将可扩展检索与细粒度推理相结合，而无需特定任务的训练。源代码和所有相关指南可在https://github.com/demidovd98/CoVRAGE获取。
cs.CV / 48 / 2609.10018
Elastoformer: Enabling Dynamic Adaptivity via Elastic Model Transformation
弹性变换器：通过弹性模型转换实现动态适应性
Abstract
EdgeAI systems are increasingly employing computer vision applications to enable intelligent, on-device decision-making in real-time. However, these deployments face highly dynamic operational conditions, with fluctuating constraints on latency, power availability, and memory resources. Deep Neural Networks (DNN), which follow fixed computational execution flows, lack the flexibility to adapt to such variability, resulting in inefficient and suboptimal performance in edge scenarios. This underscores the need for architectures that are not only efficient but also dynamically scalable at runtime. In this paper, we propose Elastoformer: A framework that transforms conventional neural networks (NN) into Elastic NN capable of real-time elastic inference. Unlike the conventional bag-of-models approach, which requires maintaining multiple independent models for different operating conditions, Elastoformer offers a single, modular solution that dynamically switches between multiple modes of operation at runtime, adapting efficiently to the changing computational budgets of edge devices without the overhead of managing separate models. Experiments reveal that our framework achieves up to 85% reduction in computation FLOPs, 50% reduction in latency and 76% reduction in memory overhead, while showcasing the architecture agnostic nature of the framework across both Vision Transformers and CNNs. Our code is available at https://github.com/sudaksh14/Elastoformer.
Chinese Translation
边缘人工智能（EdgeAI）系统越来越多地采用计算机视觉应用，以实现实时的智能设备内决策。然而，这些部署面临高度动态的操作条件，延迟、功耗和内存资源的限制时常波动。深度神经网络（DNN）遵循固定的计算执行流程，缺乏适应这种变化的灵活性，导致在边缘场景中表现出低效和次优的性能。这突显了需要一种不仅高效而且在运行时能够动态扩展的架构。在本文中，我们提出了弹性变换器（Elastoformer）：一个将传统神经网络（NN）转变为能够实时弹性推理的弹性神经网络的框架。与传统的模型包（bag-of-models）方法不同，后者需要为不同的操作条件维护多个独立模型，弹性变换器提供了一个单一的模块化解决方案，能够在运行时动态切换多种操作模式，有效适应边缘设备不断变化的计算预算，而无需管理多个独立模型的开销。实验表明，我们的框架在计算浮点运算（FLOPs）上减少了多达85%，延迟减少了50%，内存开销减少了76%，同时展示了该框架在视觉变换器（Vision Transformers）和卷积神经网络（CNNs）中的架构无关性。我们的代码可在 https://github.com/sudaksh14/Elastoformer 获取。
cs.CV / 49 / 2609.10095
LinearMask-GS: Stable-Mask Importance Pruning for Compact 3D Gaussian Splatting
LinearMask-GS：紧凑型3D高斯点云的稳定掩码重要性剪枝
Abstract
3D Gaussian Splatting (3DGS) enables real-time novel view synthesis but produces millions of primitives through adaptive densification, leading to significant storage overhead. Learned-mask pruning methods such as LP-3DGS address this by assigning each Gaussian a learnable mask to identify and prune redundant primitives. However, we identify a limitation of this paradigm: the steep slope of the Gumbel-Sigmoid activation drives mask values to the extremes within the short mask-training window, before the importance ranking has stabilized, producing a sharply bimodal distribution from which that ranking can no longer be reliably recovered. We propose LinearMask-GS, which replaces Gumbel-Sigmoid with a linear increment activation that keeps mask values in a mid-confidence regime throughout mask training, producing a stable, unimodal mask distribution whose ranking tracks importance. On Mip-NeRF 360, our method achieves 3.6x and 1.6x Gaussian reductions over 3DGS and LP-3DGS, respectively, while maintaining or improving rendering quality. For outdoor scenes, it yields a 1.6x reduction (from 2.18M to 1.36M) with notable gains in PSNR (+0.38 dB), SSIM (+0.025), and LPIPS (-0.029).
Chinese Translation
3D高斯点云（3D Gaussian Splatting, 3DGS）能够实现实时新视图合成，但通过自适应稠密化产生数百万个原始图元，导致显著的存储开销。学习掩码剪枝方法如LP-3DGS通过为每个高斯分配一个可学习的掩码来识别和剪除冗余原始图元，从而解决这一问题。然而，我们发现这一范式的一个局限性：Gumbel-Sigmoid激活函数的陡峭斜率使得掩码值在短暂的掩码训练窗口内迅速达到极值，而在重要性排序尚未稳定之前，导致产生一个尖锐的双峰分布，从中无法可靠地恢复该排序。我们提出了LinearMask-GS，它用线性增量激活函数替代Gumbel-Sigmoid，保持掩码值在整个掩码训练过程中处于中等置信度范围，从而产生一个稳定的单峰掩码分布，其排序跟踪重要性。在Mip-NeRF 360上，我们的方法分别实现了对3DGS和LP-3DGS的高斯数量减少3.6倍和1.6倍，同时保持或提高了渲染质量。对于户外场景，它实现了1.6倍的减少（从2.18M降至1.36M），并在PSNR（+0.38 dB）、SSIM（+0.025）和LPIPS（-0.029）方面取得了显著提升。
cs.CV / 50 / 2609.10125
SA-Profile: Automated Sulcus Angle Profiling from Super-Resolution MRI
SA-Profile：基于超分辨率MRI的自动化沟角轮廓分析
Abstract
Trochlear dysplasia (TD) is an abnormality of the femoral trochlea associated with anterior knee pain and patellar instability. The sulcus angle (SA) is used to assess trochlear morphology, but it is typically measured on a single axial MR slice with no clear guidance on which to select, making it sensitive to slice selection and landmark placement. We propose an automatic framework for continuous SA profiling from super-resolved MR volumes. Clinically acquired axial, coronal, and sagittal MR scans are combined using implicit neural representations to reconstruct a high-resolution volume. SA measurements are computed across the trochlear region using two landmark detection U-Net models. The approach was evaluated on the public fastMRI dataset and a small in-house cohort of patients with TD. Compared with conventional manual single-slice SA measurements, the proposed automated method yielded a mean absolute error of 11.6$^\circ$ while providing continuous characterization of trochlear morphology. Population-level analysis demonstrated distinct mean SA profiles between the public cohort and the in-house TD cohort, highlighting the potential of profile-based assessment to characterize TD. By reducing reliance on a single manually selected axial slice, the proposed framework extends conventional SA assessment to a continuous profile-based description of trochlear morphology without additional imaging, while remaining conceptually linked to current clinical assessment. Further validation is required. The code is available: https://github.com/wehrlimi/SA_Profile.
Chinese Translation
髌骨沟发育不良（TD）是与前膝疼痛和髌骨不稳定性相关的股骨髌骨沟异常。沟角（SA）用于评估髌骨沟形态，但通常是在单个轴向MRI切片上进行测量，且没有明确的选择指导，这使得其对切片选择和标志物放置非常敏感。我们提出了一种从超分辨率MRI体积中自动化连续SA轮廓分析的框架。通过隐式神经表示结合临床获取的轴向、冠状和矢状MRI扫描，重建高分辨率体积。使用两个标志物检测U-Net模型计算髌骨沟区域的SA测量。该方法在公共的fastMRI数据集和一小组TD患者的内部队列上进行了评估。与传统的手动单切片SA测量相比，所提自动化方法的平均绝对误差为11.6$^ ext{°}$，同时提供了髌骨沟形态的连续特征描述。群体水平分析显示公共队列与内部TD队列之间存在明显的平均SA轮廓差异，突显了基于轮廓的评估在表征TD方面的潜力。通过减少对单个手动选择的轴向切片的依赖，所提框架将传统SA评估扩展为对髌骨沟形态的连续轮廓描述，而无需额外成像，同时在概念上与当前临床评估保持关联。需要进一步验证。代码可在此获取：https://github.com/wehrlimi/SA_Profile。
cs.CV / 51 / 2609.10139
TransGaze-Object: Transformer Based Driver Gaze Object Prediction Framework in Real Driving
TransGaze-Object：基于变换器的真实驾驶中驾驶员注视物体预测框架
Abstract
Driver gaze provides information regarding driver visual attention and situational awareness to the surrounding traffic. Existing driver gaze estimation studies represent gaze in terms of gaze zone or gaze vector/point-of-gaze (PoG). However, object-level gaze information provides a more semantically meaningful representation of visual attention by identifying attended objects, such as vehicles, pedestrians, or traffic signals. In this study, we propose an end-to-end driver gaze object prediction framework, TransGaze-Object, Transformer-based Gaze Object prediction model. The proposed framework first extracts facial features, including face and iris-weighted eye features, along with trafficobject spatial features. A transformer based cross-attention mechanism is then used to compute similarity scores and attention weights for predicting the drivers gaze object. To train this model, we propose a benchmark driver gaze dataset, Urban Driving-Face Scene Gaze (UD-FSG), comprising synchronized driver-face and traffic-scene images, scene objects bounding boxes, and gaze labels in terms of 2D gaze coordinate and gaze object. The TransGaze-Object model achieves an overall accuracy of 60% for gaze-object prediction, compared to 51% accuracy obtained from associating the estimated Point-of-Gaze to traffic objects. The error analysis reveals that TransGaze-Object reduces confusion between traffic objects (predicted) and the background (ground-truth), achieving an error rate of 11.68%, a 49.7% relative reduction compared with 23.21% error obtained from PoG-based gaze-object association. Overall, the results demonstrate the effectiveness of directly predicting gaze objects from driver-face and traffic-scene information, rather than estimating an intermediate Point-of-Gaze and subsequently associating it with traffic objects.
Chinese Translation
驾驶员的注视提供了关于驾驶员视觉注意力和对周围交通的情境意识的信息。现有的驾驶员注视估计研究通常通过注视区域或注视向量/注视点（Point-of-Gaze, PoG）来表示注视。然而，基于物体的注视信息通过识别关注的物体（如车辆、行人或交通信号）提供了更具语义意义的视觉注意力表示。在本研究中，我们提出了一个端到端的驾驶员注视物体预测框架TransGaze-Object，基于变换器的注视物体预测模型。该框架首先提取面部特征，包括面部和虹膜加权的眼部特征，以及交通物体的空间特征。然后，采用基于变换器的交叉注意机制来计算相似性分数和注意权重，以预测驾驶员的注视物体。为了训练该模型，我们提出了一个基准驾驶员注视数据集，城市驾驶-面部场景注视（Urban Driving-Face Scene Gaze, UD-FSG），该数据集包含同步的驾驶员面部和交通场景图像、场景物体的边界框以及以2D注视坐标和注视物体形式标注的注视标签。与通过将估计的注视点与交通物体关联得到的51%准确率相比，TransGaze-Object模型在注视物体预测中的总体准确率达到60%。误差分析表明，TransGaze-Object减少了预测的交通物体与背景（真实值）之间的混淆，错误率为11.68%，相比于基于PoG的注视物体关联所获得的23.21%错误率减少了49.7%。总体而言，结果表明直接从驾驶员面部和交通场景信息中预测注视物体的有效性，而不是先估计中间的注视点再将其与交通物体关联。
cs.CV / 52 / 2609.10156
ScopeMamba-YOLO: Widening the Perceptual Scope Inward and Outward for Small Object Detection in Remote Sensing Imagery
ScopeMamba-YOLO：扩大感知范围以实现遥感图像中的小物体检测
Abstract
Small object detection in unmanned aerial vehicle (UAV) and remote sensing imagery requires preserving high-resolution detail while modeling long-range context. Adding a stride-4 detection level and removing the stride-32 stage benefits tiny targets but weakens peripheral spatial support, whereas directly inserting selective scanning into the main feature path can interfere with weak local cues. We propose ScopeMamba-YOLO, built around an off-path, zero-gated selective-scanning principle that decouples contextual modeling from the convolutional stream. The principle is instantiated by a Cascaded Global-Context Module (CGCM) in the backbone and a Selective-Scan PAN (SS-PAN) in the neck. An Adaptive Multi-scale Strip (AMS) Block reduces the cost of high-resolution feature extraction, while a Scale-Adaptive DFL (SA-DFL) head reallocates distributional support and regression capacity across scales with only 0.008M additional parameters. Controlled experiments show that matched main-path selective scanning reduces mAP50 by 0.98 pp, whereas off-path CGCM improves the final configuration by 0.67 pp over the three-seed no-CGCM mean; operator controls indicate that this gain is not explained by auxiliary branch capacity alone. ERF analysis further shows that the complete context pathway increases the peripheral energy ratio from 0.008 to 0.090 at stride 8. On VisDrone-2019, ScopeMamba-S achieves 50.8% mAP50 with 3.57M parameters, exceeding YOLOv8s by 10.8 pp while using 32% of its parameters; ScopeMamba-M reaches 52.6% mAP50 with 6.48M parameters. Consistent improvements are also observed on AI-TOD, especially for very-tiny and tiny objects.
Chinese Translation
在无人机（UAV）和遥感图像中，小物体检测需要在建模长距离上下文的同时保留高分辨率细节。增加一个步幅为4的检测层并移除步幅为32的阶段有利于微小目标，但削弱了周边空间支持，而直接将选择性扫描插入主特征路径可能会干扰微弱的局部线索。我们提出了ScopeMamba-YOLO，基于一种离路径的零门控选择性扫描原则，该原则将上下文建模与卷积流解耦。该原则通过主干中的级联全局上下文模块（Cascaded Global-Context Module, CGCM）和颈部中的选择性扫描PAN（Selective-Scan PAN, SS-PAN）得以实现。自适应多尺度条带（Adaptive Multi-scale Strip, AMS）模块减少了高分辨率特征提取的成本，而尺度自适应深度特征学习（Scale-Adaptive DFL, SA-DFL）头部在各尺度间重新分配分布支持和回归能力，仅增加0.008M的参数。控制实验表明，匹配主路径的选择性扫描使mAP50降低了0.98个百分点，而离路径的CGCM在三种种子无CGCM均值上提高了最终配置0.67个百分点；操作员控制表明，这一增益并非仅由辅助分支能力解释。ERF分析进一步表明，完整的上下文路径在步幅为8时将周边能量比从0.008提高到0.090。在VisDrone-2019上，ScopeMamba-S以3.57M参数实现了50.8%的mAP50，比YOLOv8s高出10.8个百分点，同时仅使用其32%的参数；ScopeMamba-M以6.48M参数达到了52.6%的mAP50。在AI-TOD上也观察到了持续的改进，尤其是对于非常小和小物体。
cs.CV / 53 / 2609.10187
3rd Place Solution to Human Motion Challenges in Real-World and Clinical Settings (MoCha) @ECCV2026: Language-Aligned Motion Representations for Domain-Generalizable UPDRS-Gait Severity Estimation
人类运动挑战的第三名解决方案（MoCha）@ECCV2026：用于领域通用的UPDRS-步态严重性估计的语言对齐运动表示
Abstract
In this work, we introduce language-aligned motion representations for domain-generalizable UPDRS-Gait severity estimation, aiming to learn semantically structured motion features that generalize across heterogeneous clinical domains. We first learn motion representations using a Bi-GRU backbone that captures the temporal dynamics of SMPL sequences. Prior to model training, motion captions are generated offline using Qwen2.5-7B-Instruct. The backbone is then trained with both classification and text-alignment objectives to learn discriminative and semantically structured motion representations while accounting for the class imbalance present in the training data. We subsequently adapt the learned backbone independently to each source domain so that the model can capture domain-specific motion characteristics. The resulting source-specific models are then merged at the parameter level to consolidate complementary knowledge across source domains into a single domain-generalized model. To further mitigate class imbalance, we perform GPT-5.5-based pseudo labeling, and our final merged models for each site do not use any class-prior correction during inference. The resulting model is evaluated under the unseen-site setting of the MoCha Challenge, using Macro F1 as the primary evaluation metric. Our method achieves a macro-F1 of 0.57 on the hidden test set with only 637K active parameters at inference, ranking 3rd among 58 leaderboard entries in the MoCha 2026 Challenge. The challenge attracted 1,669 submissions from 112 participants and offered monetary prizes sponsored by Machine Medicine Technologies.
Chinese Translation
在本研究中，我们引入了用于领域通用的UPDRS-步态严重性估计的语言对齐运动表示，旨在学习在异质临床领域中具有语义结构的运动特征。我们首先使用Bi-GRU骨干网络学习运动表示，该网络捕捉SMPL序列的时间动态。在模型训练之前，使用Qwen2.5-7B-Instruct离线生成运动描述。然后，骨干网络在分类和文本对齐目标下进行训练，以学习具有区分性和语义结构的运动表示，同时考虑训练数据中存在的类别不平衡。随后，我们独立地将学习到的骨干网络适应于每个源领域，以便模型能够捕捉领域特定的运动特征。然后，将结果的源特定模型在参数级别合并，以将源领域之间的互补知识整合到一个单一的领域通用模型中。为了进一步减轻类别不平衡，我们执行基于GPT-5.5的伪标注，并且我们为每个站点最终合并的模型在推理过程中不使用任何类别先验校正。所得到的模型在MoCha挑战的未见站点设置下进行评估，使用宏F1作为主要评估指标。我们的方法在隐藏测试集上达到了0.57的宏F1，仅在推理时使用637K的活跃参数，在MoCha 2026挑战中排名58个排行榜条目的第3名。该挑战吸引了来自112名参与者的1,669份提交，并提供了由机器医学技术赞助的奖金。
cs.CV / 54 / 2609.10224
UOT-Gap: A Variational Principle for the Modality Gap in Vision-Language Models via Unbalanced Optimal Transport
UOT-Gap：通过不平衡最优传输的视觉-语言模型中的模态差距变分原理
Abstract
Vision-language models such as CLIP embed images and text in a shared space, where modality-specific distributions often remain separated. Existing accounts connect this modality gap to initialization, contrastive dynamics, and information imbalance, while its distributional and pairwise contributions to retrieval remain unresolved. We introduce UOT-Gap, a training-free variational diagnostic that models frozen image and text embeddings with unbalanced entropic optimal transport (UOT). The UOT optimum separates transport, coupling complexity, and marginal mass variation; a complementary pair-aware residual compares observed image-caption pairs with the UOT soft matching. On Flickr8K and COCO-1K with frozen CLIP, OpenCLIP, and SigLIP encoders, caption degradation reduces Flickr8K Recall@1 from 0.559 to 0.003. Across six dataset-model conditions, the pair-aware residual tracks retrieval degradation with mean absolute Spearman 0.973, compared with 0.392 for the mean gap. The association remains stable across five random COCO-1K subsets at $0.954\pm0.026$, with a minimum of 0.943. UOT barycentric updates reduce the transport objective while degrading retrieval, distinguishing geometric objective descent from task improvement. These results establish UOT-Gap as a diagnostic for caption quality, modality alignment, and retrieval robustness.
Chinese Translation
视觉-语言模型如 CLIP 将图像和文本嵌入到一个共享空间中，其中模态特定的分布往往保持分离。现有研究将这种模态差距与初始化、对比动态和信息不平衡联系起来，而其在检索中的分布和成对贡献仍未得到解决。我们引入了 UOT-Gap，一种无训练的变分诊断工具，利用不平衡熵最优传输（UOT）对冻结的图像和文本嵌入进行建模。UOT 最优解将传输、耦合复杂性和边际质量变化分开；一个互补的成对残差比较观察到的图像-标题对与 UOT 软匹配。在使用冻结的 CLIP、OpenCLIP 和 SigLIP 编码器的 Flickr8K 和 COCO-1K 数据集上，标题降级使得 Flickr8K 的 Recall@1 从 0.559 降至 0.003。在六个数据集-模型条件下，成对残差跟踪检索降级的平均绝对斯皮尔曼相关系数为 0.973，而平均差距为 0.392。该关联在五个随机 COCO-1K 子集上保持稳定，相关系数为 $0.954 ext{±}0.026$，最小值为 0.943。UOT 重心更新在降低传输目标的同时降低检索性能，区分了几何目标下降与任务改进。这些结果确立了 UOT-Gap 作为标题质量、模态对齐和检索鲁棒性的诊断工具。
cs.CV / 55 / 2609.10261
When Fusion Fails: Corruption-Aware Rebalanced Fusion for Multi-Modal Medical Image Segmentation
当融合失败时：面向腐败感知的重平衡融合用于多模态医学图像分割
Abstract
Multi-modal medical image segmentation leverages complementary diagnostic information, yet fusion can underperform single-modality baselines when spatially aligned inputs differ in quality. Here, "corruption" primarily denotes resolution-induced degradation rather than misalignment or complete modality absence, while synthetic noise is evaluated only as an auxiliary setting. We identify a critical optimization-inference inconsistency: degraded modalities can receive weak training updates yet substantially affect predictions, indicating active interference with fusion. We attribute this failure to resampling-induced feature corruption and optimization bias, where noisy features propagate through skip connections and encourage unreliable modality selection. We therefore propose CoReFuse-Med, a Corruption-aware Rebalanced Fusion framework that suppresses corruption during feature transmission and rebalances modality contributions during high-level fusion. Experiments on EPVS, BraTS, and WMH, including multiple Z-axis slice-retention ratios and an auxiliary noise test, demonstrate improved accuracy and robustness under modality-quality discrepancies. Our code is available at https://github.com/lrever/CoReFuse.
Chinese Translation
多模态医学图像分割利用互补的诊断信息，然而当空间对齐的输入在质量上存在差异时，融合可能表现不如单模态基线。在这里，“腐败”主要指的是由分辨率引起的退化，而不是错位或完全缺失的模态，同时合成噪声仅作为辅助设置进行评估。我们识别出一个关键的优化-推理不一致性：退化的模态可能会接收微弱的训练更新，但却显著影响预测，表明其对融合的主动干扰。我们将这一失败归因于重采样引起的特征腐败和优化偏差，其中噪声特征通过跳跃连接传播并鼓励不可靠的模态选择。因此，我们提出了CoReFuse-Med，一个腐败感知重平衡融合框架，该框架在特征传输过程中抑制腐败，并在高层融合过程中重平衡模态贡献。在EPVS、BraTS和WMH上的实验，包括多个Z轴切片保留比率和辅助噪声测试，证明了在模态质量差异下的准确性和鲁棒性得到了改善。我们的代码可在https://github.com/lrever/CoReFuse获取。
cs.CV / 56 / 2609.10278
FreqFLD: Towards All-in-One Facial Landmark Detection via Frequency Modulation
FreqFLD：通过频率调制实现一体化人脸关键点检测
Abstract
Recent progress in deep learning has significantly advanced facial landmark detection. However, most existing methods process features in a spatial-domain manner under a dataset-specific training paradigm, which overlooks the fact that facial landmark detection is inherently geometry-driven and sensitive to frequency variations, thereby limiting cross-dataset generalization under complex scenarios and hindering the development of a facial landmark detection model. To address this issue, we propose \textbf{FreqFLD}, a \textbf{freq}uency-modulated framework towards All-in-One \textbf{f}acial \textbf{l}andmark \textbf{d}etection. Specifically, FreqFLD introduces a Frequency Modulation Module (FreqMoM) to explicitly induce the frequency prior by decoupling and modulating low- and high-frequency components, which is then injected into subsequent feature modeling to enable balanced modeling of global facial structure and local landmark details. Furthermore, FreqFLD employs a Frequency-Modulated Mixture-of-Experts (FreqMoE), with expert selection adaptively conditioned on frequency-modulated priors, enabling flexible modeling of heterogeneous facial landmark patterns under diverse and challenging scenarios. To regularize frequency-consistent modeling under the All-in-One paradigm, we further introduce a Frequency-Consistent Routing (FreqCR) loss, which constrains the routing and assignment of frequency-aware experts to promote balanced expert utilization across diverse facial scenarios, thereby enabling stable expert specialization and achieving robust facial landmark detection. Extensive experiments demonstrate that the proposed FreqFLD achieves comparable performance on popular datasets. The code is available at: https://github.com/jkj1059657014/FreqFLD.
Chinese Translation
近年来，深度学习的进展显著推动了人脸关键点检测的发展。然而，大多数现有方法在数据集特定的训练范式下以空间域的方式处理特征，这忽视了人脸关键点检测本质上是由几何驱动且对频率变化敏感的事实，从而限制了在复杂场景下的跨数据集泛化，并阻碍了人脸关键点检测模型的发展。为了解决这一问题，我们提出了 extbf{FreqFLD}，一个面向一体化人脸关键点检测的 extbf{频}率调制框架。具体而言，FreqFLD引入了频率调制模块（Frequency Modulation Module，FreqMoM），通过解耦和调制低频和高频成分，明确引入频率先验，然后将其注入后续特征建模中，以实现全球人脸结构与局部关键点细节的平衡建模。此外，FreqFLD采用了频率调制专家混合模型（Frequency-Modulated Mixture-of-Experts，FreqMoE），其专家选择根据频率调制先验自适应条件，使得在多样且具有挑战性的场景下灵活建模异质的人脸关键点模式。为了在一体化范式下规范频率一致的建模，我们进一步引入了频率一致路由（Frequency-Consistent Routing，FreqCR）损失，约束频率感知专家的路由和分配，以促进在多样人脸场景下的专家平衡利用，从而实现稳定的专家专业化并达到鲁棒的人脸关键点检测。大量实验表明，所提出的FreqFLD在流行数据集上达到了可比的性能。代码可在以下链接获取：https://github.com/jkj1059657014/FreqFLD。
cs.CV / 57 / 2609.10292
Isotropic Embedding Perturbations for Robust Vision Language Encoders
用于鲁棒视觉语言编码器的各向同性嵌入扰动
Abstract
Data augmentation is fundamental to training modern deep vision and multimodal models. While individual methods, such as RandAug, CutMix, Mixup, RandErase, and DropPath, offer strong regularization effects, their combined use has saturated in performance due to overlapping functionalities, and aggressive pixel-level manipulations may disrupt delicate cross-modal alignment. This saturation motivates the search for a new augmentation axis within the embedding space rather than the input space. We introduce Aether, a simple plug-in method that applies diffusion-style random perturbations in the embedding space via controlled alpha-mixing, specifically designed to provide isotropic regularization that remains semantically consistent. Inspired by feature-space perturbations in language models and image degradation in generative pretraining, Aether induces mild yet effective perturbations that smooth the representations without compromising the fine-grained structural information required for strong vision-language encoders. Across diverse architectures and across multiple recognition tasks, Aether delivers consistent gains over the advanced recipe combining CutMix, Mixup, DropPath, and RandAug---a level of improvement rarely observed with modern augmentation alternatives. Notably, Aether demonstrates superior effectiveness in multi-modal alignment, succeeding where traditional pixel-space augmentations fail by providing a stable, isotropic regularization signal that respects the integrity of the high-dimensional feature space.
Chinese Translation
数据增强是训练现代深度视觉和多模态模型的基础。虽然单独的方法，如 RandAug、CutMix、Mixup、RandErase 和 DropPath，提供了强大的正则化效果，但由于功能重叠，它们的组合使用在性能上已达到饱和，且激进的像素级操作可能会破坏细腻的跨模态对齐。这种饱和促使我们在嵌入空间中寻找新的增强方向，而非输入空间。我们引入了 Aether，这是一种简单的插件方法，通过受控的 alpha 混合在嵌入空间中应用扩散风格的随机扰动，专门设计用于提供各向同性的正则化，同时保持语义一致性。Aether 的灵感来源于语言模型中的特征空间扰动和生成预训练中的图像退化，能够引入温和而有效的扰动，平滑表示而不损害强视觉语言编码器所需的细粒度结构信息。在多种架构和多个识别任务中，Aether 在结合 CutMix、Mixup、DropPath 和 RandAug 的先进方案上提供了一致的性能提升——这种改进在现代增强替代方案中很少见。值得注意的是，Aether 在多模态对齐方面表现出卓越的有效性，成功地在传统的像素空间增强失败的地方提供了稳定的各向同性正则化信号，尊重高维特征空间的完整性。
cs.CV / 58 / 2609.10297
TRACE: Trajectory-robust Admission with Evidence Ordering for Efficient GUI Agents
TRACE：基于证据排序的轨迹鲁棒性接纳方法用于高效的GUI代理
Abstract
GUI agents accumulate high-resolution screenshots as the trajectory unfolds, increasing inference latency and memory usage. Training-free visual token pruning can reduce this cost, but cache reuse introduces a fundamental constraint. Once tokens are discarded, the corresponding visual evidence cannot be recovered without re-encoding. Pruning therefore becomes an \textit{irreversible admission decision} that must remain useful for unknown future targets while preserving coverage of operable regions under tight budgets. To address these challenges, we propose \textbf{\method{}}, a training-free framework for \emph{\textbf{T}rajectory-\textbf{r}obust \textbf{A}dmission and \textbf{C}overage-aware \textbf{E}vidence ordering}. Specifically, we combine a query-independent layout-derived interaction prior with instruction relevance and feature novelty to rank visual evidence according to both potential future utility and diversity. Then, we reserve part of the budget for native visual tokens distributed across the screen, repairing missing spatial coverage without breaking the ordering. Together, these mechanisms produce a nested token order, allowing retained visual evidence to shrink monotonically across budgets while remaining reusable throughout the trajectory. Finally, our monotone KV contraction incrementally contracts retired frames into compact session state, avoiding repeated visual encoding or pruning. Extensive experiments across six GUI benchmarks and diverse models verify the effectiveness of our proposed \method{} under tight budgets. The source code will be released.
Chinese Translation
GUI代理在轨迹展开过程中积累高分辨率的屏幕截图，这增加了推理延迟和内存使用。无训练的视觉令牌修剪可以降低这一成本，但缓存重用引入了一个基本限制。一旦令牌被丢弃，相应的视觉证据在没有重新编码的情况下无法恢复。因此，修剪成为一种 extit{不可逆的接纳决策}，必须在保持可操作区域覆盖的同时，对未知的未来目标保持有效。为了解决这些挑战，我们提出了 extbf{ extit{TRACE}}，一个无训练的框架，旨在实现 extbf{T}rajectory- extbf{r}obust extbf{A}dmission和 extbf{C}overage-aware extbf{E}vidence ordering。具体而言，我们结合了与查询无关的布局派生交互先验、指令相关性和特征新颖性，根据潜在的未来效用和多样性对视觉证据进行排序。然后，我们为分布在屏幕上的本地视觉令牌保留部分预算，修复缺失的空间覆盖而不打破排序。这些机制共同产生了一个嵌套的令牌顺序，使得保留的视觉证据在预算范围内单调缩小，同时在整个轨迹中保持可重用性。最后，我们的单调KV收缩逐步将已退休帧收缩为紧凑的会话状态，避免重复的视觉编码或修剪。在六个GUI基准和多种模型上的广泛实验验证了我们提出的 extit{TRACE}在紧张预算下的有效性。源代码将会发布。
cs.CV / 59 / 2609.10303
SynThermFace: Amplifying Limited Paired Data for Visible-Thermal Face Recognition via Synthetic Data Generation
SynThermFace：通过合成数据生成放大有限配对数据以实现可见-热成像人脸识别
Abstract
Face recognition (FR) is a widely used modality for biometric authentication, but conventional models rely on visible-spectrum imagery and degrade when high-quality RGB images cannot be captured. Cross-spectral face recognition addresses this limitation by matching visible images with other modalities such as thermal imagery, enabling more reliable performance in low-light, nighttime, and unconstrained conditions. However, progress is limited by the scarcity of paired visible-thermal data, which is difficult and costly to collect at scale. We propose SynThermFace, a framework that amplifies limited real visible-thermal supervision into larger paired adaptation datasets for cross-spectral face recognition. A diffusion model is first adapted using a limited set of paired visible--thermal images and then used to generate large-scale paired visible--synthetic thermal data from existing real or synthetic visible face datasets. The generated pairs are used to adapt a pretrained visible-spectrum face recognition model into a CFR model. Unlike synthesis-based approaches that require image translation at test time, the proposed method shifts generation to the training stage and performs inference with a single forward pass through the adapted recognition model. Under the same MCXFace real-pair protocol, PACT improves over the evaluated CFR adaptation baselines, isolating the effect of the proposed adaptation objective. Training PACT on larger generated paired datasets provides additional improvements over both the unadapted model and the real-pair PACT configuration. Cross-database evaluation on the Tufts dataset provides evidence that the learned representation transfers to an unseen database. The source code and trained models will be made publicly available.
Chinese Translation
人脸识别（FR）是一种广泛使用的生物识别认证方式，但传统模型依赖于可见光谱图像，当无法捕获高质量的RGB图像时，性能会下降。跨光谱人脸识别通过将可见图像与热成像等其他模态匹配，解决了这一限制，从而在低光照、夜间和不受约束的条件下实现更可靠的性能。然而，进展受到有限的配对可见-热成像数据的稀缺性限制，这种数据的收集既困难又昂贵。我们提出了SynThermFace，一个框架，旨在将有限的真实可见-热成像监督放大为更大规模的配对适应数据集，以用于跨光谱人脸识别。首先，使用有限的配对可见-热成像图像对扩散模型进行适配，然后利用该模型从现有的真实或合成可见人脸数据集中生成大规模的配对可见-合成热成像数据。生成的配对数据用于将预训练的可见光谱人脸识别模型适配为CFR模型。与需要在测试时进行图像转换的合成方法不同，所提方法将生成过程转移到训练阶段，并通过适配后的识别模型进行单次前向推理。在相同的MCXFace真实配对协议下，PACT在评估的CFR适配基线之上有所提升，隔离了所提适配目标的影响。在更大规模的生成配对数据集上训练PACT，相较于未适配模型和真实配对PACT配置，提供了额外的改进。对Tufts数据集的跨数据库评估提供了证据，表明所学习的表示能够迁移到未见数据库。源代码和训练模型将公开发布。
cs.CV / 60 / 2609.10317
Decoupled Self-Forcing Distillation for Streaming Talking Head Generation
解耦自强蒸馏用于流媒体对话头生成
Abstract
Streaming talking-head generation produces each frame as its driving audio arrives, yet fidelity and efficiency have so far pulled in opposite directions: end-to-end methods condition a video diffusion model on audio directly and achieve high quality but only at large scale, while cheaper two-stage methods generate an intermediate motion representation and trail in fidelity. We argue the cost of the former lies in the target of fusion: the video latent is dominated by identity, appearance and background, none of which audio bears on, so coupling audio to every pixel blurs detail and wastes capacity. We instead fuse conditions in a low-dimensional identity-disentangled motion space, routing audio and motion captions by their temporal granularity, and generate motion latents with a small causal autoregressive transformer that a pretrained diffusion renderer turns into video. Conditions thus control video transitively, and high fidelity no longer requires a large backbone. Streaming this decomposition needs both models to be causal, and the exposure-bias problem could be solved by self-forcing given a bidirectional teacher. But there is no such teacher in motion space. Our decoupled self-forcing distillation resolves both models under one frozen teacher: conditioned on motion, it distills the renderer into a block-causal student; unconditionally, it scores rendered rollouts against real videos, supervising motion by the video it produces. This lifts the fidelity ceiling from the motion generator onto the stronger renderer. The two models run as parallel causal streams, reaching 15.4 FPS at 1.3 s latency with no quality degradation.
Chinese Translation
流媒体对话头生成在驱动音频到达时生成每一帧，但迄今为止，保真度和效率却朝着相反的方向发展：端到端方法直接将视频扩散模型与音频进行条件化，虽然能够实现高质量，但仅在大规模下才能实现，而更便宜的两阶段方法生成中间运动表示，保真度却落后。我们认为前者的成本在于融合的目标：视频潜在特征主要由身份、外观和背景主导，而这些都与音频无关，因此将音频与每个像素耦合会模糊细节并浪费容量。我们则在一个低维的身份解耦运动空间中融合条件，根据音频和运动字幕的时间粒度进行路由，并通过一个小型因果自回归变换器生成运动潜在特征，该变换器由预训练的扩散渲染器转化为视频。因此，条件以传递方式控制视频，高保真度不再需要大型主干。流媒体这种解构需要两个模型都是因果的，而曝光偏差问题可以通过自强来解决，前提是有一个双向教师。但在运动空间中并没有这样的教师。我们的解耦自强蒸馏在一个冻结的教师下解决了两个模型：在运动条件下，它将渲染器蒸馏为一个块因果学生；在无条件下，它对渲染的输出与真实视频进行评分，通过其生成的视频监督运动。这将运动生成器的保真度上限提升到更强的渲染器上。这两个模型作为并行因果流运行，在1.3秒延迟下达到15.4帧每秒的速度且没有质量下降。
cs.CV / 61 / 2609.10322
Geometry Without Coordinates: LiDAR Diffusion as a 3D Feature Bridge
无坐标几何：LiDAR扩散作为3D特征桥梁
Abstract
Transferring the rich priors of large 2D foundation models to sparse 3D LiDAR remains challenging, as training native 3D foundation models at comparable scale is limited by data and annotation scarcity. We introduce a LiDAR-conditioned diffusion model trained on pseudo-labels from off-the-shelf 2D foundation models. The model supports multiple output modalities, including depth, semantic segmentation and instance prediction, selectable via a textual task prompt. Because the model is conditioned on LiDAR, both its outputs and its intermediate UNet features can be projected back onto the input point cloud, enabling analysis of a 3D representation learned entirely under 2D supervision. We study this representation directly in point-cloud space, explicitly excluding raw spatial coordinates to isolate feature content from projection geometry. Linear probes recover up to ~23% Mean Intersection over Union (MIoU) on 3D semantic classes, compared to ~3.5% for a matched Gaussian-noise control, indicating substantial non-trivial structure. Pairwise cosine similarity across modality-specific feature streams reveals a layered organization. Early encoder layers remain weakly aligned across modalities while individually decodable, intermediate layers converge toward a shared representation, and decoder layers re-specialize toward task-specific outputs. These findings indicate that LiDAR-conditioned diffusion models can induce structured 3D representations from 2D supervision alone, with a modality-dependent manifold that locally unifies near a shared bottleneck. This positions diffusion as a viable mechanism for transferring large-scale 2D priors into sparse 3D domains.
Chinese Translation
将大型2D基础模型的丰富先验知识转移到稀疏的3D LiDAR上仍然具有挑战性，因为在可比规模上训练原生3D基础模型受到数据和标注稀缺的限制。我们提出了一种基于LiDAR条件的扩散模型，该模型在现成的2D基础模型的伪标签上进行训练。该模型支持多种输出模式，包括深度、语义分割和实例预测，可以通过文本任务提示进行选择。由于模型是基于LiDAR条件的，因此其输出和中间的UNet特征都可以投影回输入点云，从而实现对完全在2D监督下学习的3D表示的分析。我们直接在点云空间中研究这种表示，明确排除原始空间坐标，以将特征内容与投影几何分离。线性探测器在3D语义类别上恢复了约23%的平均交并比（Mean Intersection over Union, MIoU），而匹配的高斯噪声对照组仅为约3.5%，这表明存在显著的非平凡结构。跨模态特征流的成对余弦相似性揭示了分层组织。早期编码器层在各模态间保持弱对齐，同时可以单独解码，中间层趋向于共享表示，而解码器层则重新专门化为任务特定的输出。这些发现表明，基于LiDAR条件的扩散模型可以仅通过2D监督诱导结构化的3D表示，并且具有依赖模态的流形，在共享瓶颈附近局部统一。这使得扩散成为将大规模2D先验知识转移到稀疏3D领域的可行机制。
cs.CV / 62 / 2609.10333
Learning to Adapt and Calibrate: Score Distribution Alignment for Few-Shot Uncertainty Prediction in Medical VLMs
学习适应与校准：少样本不确定性预测中的得分分布对齐在医学视觉-语言模型中的应用
Abstract
Uncertainty estimation for medical vision--language models (VLMs) using conformal prediction has gained increasing attention due to its distribution-free coverage guarantees. However, standard conformal prediction relies on exchangeability between calibration and test data and typically requires a sufficiently large calibration set to obtain reliable coverage. These assumptions are difficult to satisfy in few-shot transfer settings, where only a small labeled support set is available to adapt a pretrained VLM to a new medical task, while an unlabeled query set is used for evaluation. Supervised fine-tuning on the support set changes the model parameters and consequently shifts the nonconformity score distribution, breaking exchangeability between calibration and query samples and leading to unreliable coverage under distribution shift. Existing transductive conformal adaptation methods often preserve validity by avoiding supervised updates. While this helps maintain conformal assumptions, it underutilizes the scarce labeled support data and limits task adaptation, which is the primary objective in few-shot learning. In this setting, conformal prediction should serve as an uncertainty estimation layer that supports the adapted model, rather than preventing adaptation itself. To this end, we propose AlignCP, a framework that reconciles supervised few-shot adaptation with conformal uncertainty estimation under non-exchangeability. AlignCP learns a reweighted calibration distribution that reduces the score-level discrepancy between the labeled support set and the unlabeled query set. By aligning the one-dimensional nonconformity score distributions, AlignCP aims to close the coverage gap induced by adaptation without requiring query labels.
Chinese Translation
使用保形预测进行医学视觉-语言模型（VLMs）不确定性估计因其无分布覆盖保证而受到越来越多的关注。然而，标准的保形预测依赖于校准数据与测试数据之间的可交换性，并通常需要足够大的校准集以获得可靠的覆盖。这些假设在少样本迁移设置中难以满足，因为在这种情况下，仅有一个小的标记支持集可用于将预训练的VLM适应于新的医学任务，而无标记的查询集则用于评估。在支持集上的监督微调会改变模型参数，从而导致非一致性得分分布的变化，打破了校准样本与查询样本之间的可交换性，并在分布变化下导致不可靠的覆盖。现有的传导保形适应方法通常通过避免监督更新来保持有效性。虽然这有助于维持保形假设，但却未能充分利用稀缺的标记支持数据，并限制了任务适应性，这是少样本学习的主要目标。在这种情况下，保形预测应作为支持适应模型的不确定性估计层，而不是阻止适应。为此，我们提出了AlignCP，一个在非可交换性下调和监督少样本适应与保形不确定性估计的框架。AlignCP学习一个重新加权的校准分布，以减少标记支持集与无标记查询集之间的得分级差异。通过对齐一维非一致性得分分布，AlignCP旨在缩小适应引起的覆盖差距，而无需查询标签。
cs.CV / 63 / 2609.10334
Dimensionality Reduction for Hyperspectral Image Classification
高光谱图像分类的降维方法
Abstract
This paper addresses the issue of supervised classification in the context of hyperspectral satellite images. It deals with two fundamental aspects: dimensionality reduction of data and the selection of appropriate supervised classification techniques. Firstly, we delve into dimensionality reduction, a critical step in simplifying the management of hyperspectral data. The reduction aims to decrease complexity in terms of memory and computing time. We examine two commonly used methods: Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). Subsequently, we explore the selection of the most suitable supervised classification algorithms for hyperspectral images. We compare the performance of three methods: K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Random Forest (RF) using real hyperspectral data. The results highlight that the combination of PCA and RF yields the highest overall accuracy and Kappa coefficient.
Chinese Translation
本文探讨了高光谱卫星图像背景下的监督分类问题，涉及两个基本方面：数据的降维和适当监督分类技术的选择。首先，我们深入研究降维，这是简化高光谱数据管理的关键步骤。降维旨在减少内存和计算时间方面的复杂性。我们考察了两种常用方法：主成分分析（PCA）和线性判别分析（LDA）。随后，我们探讨了最适合高光谱图像的监督分类算法的选择。我们使用真实的高光谱数据比较了三种方法的性能：K-近邻（KNN）、支持向量机（SVM）和随机森林（RF）。结果表明，PCA与RF的组合在整体准确率和Kappa系数上表现最佳。
cs.CV / 64 / 2609.10346
Beyond One-Size-Fits-All: Sample-Adaptive Strategy Routing for Vision Token Pruning in MLLMs
超越一刀切：针对视觉令牌剪枝的样本自适应策略路由在多模态大语言模型中的应用
Abstract
Multimodal large language models (MLLMs) process hundreds or thousands of visual tokens per image, incurring prohibitive inference costs. While existing vision token pruning methods mitigate this overhead, they implicitly assume that a single fixed pruning strategy can be applied uniformly across all inputs. Our analysis further reveals that ranking pruning methods by average benchmark accuracy conceals substantial sample-wise complementarity: although the average-best strategy excels overall, alternative strategies prove superior on a significant fraction of individual samples. To harness this diversity, we propose VIP-Router, a lightweight VIsion Pruning Router that adaptively selects the pruning strategy predicted to be best suited to each input at a specified pruning level. Conditioned on low-cost visual and textual features, VIP-Router identifies the most suitable candidate strategy while retaining full-token inference as an option when pruning is predicted to be unfavorable. Evaluated on a curated suite of pruning-sensitive visual perception benchmarks, VTC-Bench Group A, VIP-Router consistently outperforms the best fixed strategy baseline across all reduction ratios, achieving a 26.9% relative improvement in average accuracy, and a 22.0% relative increase in average utility after accounting for realized token cost. Crucially, VIP-Router operates in a plug-and-play manner without modifying underlying pruning algorithms or model weights, introducing trainable parameters equivalent to merely 0.017\% of the backbone. Furthermore, VIP-Router proves effective across various MLLM backbones and yields consistent gains on unseen benchmarks, highlighting the potential of sample adaptive routing for visual token pruning.
Chinese Translation
多模态大语言模型（MLLMs）每幅图像处理数百或数千个视觉令牌，导致推理成本高昂。尽管现有的视觉令牌剪枝方法减轻了这一负担，但它们隐含地假设可以对所有输入统一应用单一固定的剪枝策略。我们的分析进一步揭示，通过平均基准准确率对剪枝方法进行排名掩盖了样本间的显著互补性：尽管平均最佳策略在整体上表现出色，但替代策略在相当一部分个体样本上表现更佳。为了利用这种多样性，我们提出了VIP-Router，一种轻量级的视觉剪枝路由器，它自适应地选择预测最适合每个输入的剪枝策略，并在指定的剪枝水平下进行操作。在低成本的视觉和文本特征的条件下，VIP-Router识别出最合适的候选策略，同时在预测剪枝不利时保留全令牌推理作为选项。在经过精心挑选的对剪枝敏感的视觉感知基准测试集VTC-Bench Group A上评估，VIP-Router在所有减缩比下始终优于最佳固定策略基线，平均准确率相对提高了26.9%，在考虑实现的令牌成本后，平均效用相对增加了22.0%。重要的是，VIP-Router以即插即用的方式运行，无需修改底层剪枝算法或模型权重，引入的可训练参数仅占主干的0.017%。此外，VIP-Router在各种MLLM主干上均表现有效，并在未见基准上实现了一致的增益，突显了样本自适应路由在视觉令牌剪枝中的潜力。
cs.CV / 65 / 2609.10355
Why Is Video Still So Expensive? A Survey of Inference-Efficiency Mechanisms in Video and Audiovisual LLMs
视频为何仍然如此昂贵？视频和视听大语言模型中的推理效率机制调查
Abstract
Video understanding has rapidly evolved toward video large language models (VideoLLMs): systems that couple video representations with pretrained large language models and condition generation on a textual prompt. Their strong performance on captioning, question answering, retrieval and temporal grounding comes at a computation and memory cost that grows with frame count and context length, limiting deployment in real-time, mobile and resource-constrained settings. This survey covers inference-efficiency mechanisms for visual and audiovisual VideoLLMs that report concrete reductions in parameter count, FLOPs per input, latency, memory, or visual and audio token count. We analyze bottlenecks across frame sampling, modality encoding, connector-level token reduction, and LLM prefilling and decoding. We organize methods by the pipeline stage at which they act, covering VideoLLMs developed since late 2022 together with earlier frame-sampling and vision-encoder mechanisms that remain components of current pipelines. We assemble literature-reported accuracy--cost comparisons under shared host models and input protocols wherever available, distinguish them from heterogeneous cross-paper evidence, and identify gaps in audiovisual efficiency and standardized evaluation. We maintain a repository at https://github.com/momentslab/awesome-efficient-videollm.
Chinese Translation
视频理解迅速发展为视频大语言模型（VideoLLMs）：这些系统将视频表示与预训练的大语言模型结合，并基于文本提示进行生成。它们在字幕生成、问答、检索和时间定位等任务上的强大表现伴随着计算和内存成本的增加，这些成本随着帧数和上下文长度的增长而上升，从而限制了在实时、移动和资源受限环境中的部署。本调查涵盖了视觉和视听VideoLLMs的推理效率机制，这些机制在参数数量、每输入的浮点运算（FLOPs）、延迟、内存或视觉和音频令牌数量上报告了具体的减少。我们分析了帧采样、模态编码、连接器级令牌减少以及大语言模型的预填充和解码等方面的瓶颈。我们根据方法作用的管道阶段对其进行组织，涵盖了自2022年底以来开发的VideoLLMs，以及仍然是当前管道组件的早期帧采样和视觉编码机制。我们汇总了在共享主机模型和输入协议下文献报告的准确性与成本比较，区分它们与异质跨论文证据，并识别视听效率和标准化评估中的空白。我们在 https://github.com/momentslab/awesome-efficient-videollm 维护一个资源库。
cs.CV / 66 / 2609.10356
Spot-the-shift: Evaluating Grounded Image Difference Captioning of Long-term Changes
发现变化：评估长期变化的基于图像差异的描述
Abstract
Long-term change understanding from images of the same place revisited over time is a challenging task with applications in map maintenance and urban infrastructure monitoring. Prior work addresses it either through pixel-level prediction or difference captioning, neither of which is sufficient to reliably measure how well models detect and describe such changes. We introduce SPOT-THE-SHIFT, a human-verified benchmark for grounded image difference captioning of long-term changes in real-world driving scenes. Our benchmark provides natural language captions and spatial masks for structural changes across each image pair. We further propose an evaluation protocol that reliably assesses models' captioning ability, validated through human studies. Benchmarking state-of-the-art MLLMs, we find that models struggle with the fine-grained multi-image spatial capability required for this task. Finally, we develop a synthetic data generation pipeline that improves an off-the-shelf MLLM without sacrificing general capabilities.
Chinese Translation
从同一地点在不同时间拍摄的图像中理解长期变化是一项具有挑战性的任务，应用于地图维护和城市基础设施监测。之前的研究要么通过像素级预测，要么通过差异描述来解决这一问题，但这两者都不足以可靠地衡量模型检测和描述这些变化的能力。我们引入了SPOT-THE-SHIFT，这是一个经过人工验证的基准，用于评估真实驾驶场景中长期变化的基于图像差异的描述。我们的基准为每对图像提供自然语言描述和结构变化的空间掩码。我们进一步提出了一种评估协议，可靠地评估模型的描述能力，并通过人类研究进行了验证。在对最先进的多模态语言模型（MLLMs）进行基准测试时，我们发现模型在执行此任务所需的细粒度多图像空间能力方面存在困难。最后，我们开发了一种合成数据生成管道，能够在不牺牲通用能力的情况下改善现成的多模态语言模型。
cs.CV / 67 / 2609.10363
SceneHI: High-Resolution 3D-Consistent Scene Texturing with Controllable Illumination
SceneHI：具有可控光照的高分辨率3D一致场景纹理生成
Abstract
SceneHI is a framework that lifts high-resolution, illumination-aware priors from 2D diffusion models to perform 3D texture synthesis. It is the first to demonstrate that high-resolution textures, previously limited to 2D synthesis, can be generated directly on 3D objects without model fine-tuning or optimization. Designed for complex, multi-object environments, SceneHI uniquely combines 3D-consistency, high-resolution fidelity, and physically plausible baked shadows within a single generative pipeline. To enforce strict geometric coherence, we introduce an exact analytical pixel-to-texel mapping that aligns diffusion trajectories across multiple viewpoints. We utilize High-Resolution Latent Textures (HRLTs) as a persistent canvas for gradually denoised textures, while camera views perform the denoising steps in latent pixel space. This ensures a shared base texture that can be subsequently refined to high resolution without compromising multi-view consistency. Finally, a light-aware generative pass embeds realistic geometry-consistent shadows directly into the atlases, bridging the gap to production workflows. SceneHI achieves high visual fidelity while reducing generation time by 80% compared to existing scene-level methods.
Chinese Translation
SceneHI是一个框架，它将高分辨率、考虑光照的先验知识从2D扩散模型提升到3D纹理合成。它首次展示了高分辨率纹理，之前仅限于2D合成，可以直接在3D物体上生成，而无需对模型进行微调或优化。SceneHI专为复杂的多物体环境设计，独特地将3D一致性、高分辨率保真度和物理上合理的烘焙阴影结合在一个生成管道中。为了确保严格的几何一致性，我们引入了一种精确的像素到纹理单元（texel）映射，能够在多个视角之间对齐扩散轨迹。我们利用高分辨率潜在纹理（High-Resolution Latent Textures, HRLTs）作为逐步去噪纹理的持久画布，同时相机视角在潜在像素空间中执行去噪步骤。这确保了一个共享的基础纹理，随后可以在不妨碍多视图一致性的情况下细化到高分辨率。最后，一个考虑光照的生成过程将真实的几何一致阴影直接嵌入到图集（atlases）中，弥合了与生产工作流程之间的差距。与现有场景级方法相比，SceneHI在实现高视觉保真度的同时将生成时间减少了80%。
cs.CV / 68 / 2609.10371
Beyond Weak Labels: Prompt-Guided Local Refinement for Weakly Supervised Water Segmentation in High-Resolution Multispectral Imagery
超越弱标签：基于提示的局部精细化用于高分辨率多光谱图像中的弱监督水体分割
Abstract
High-resolution water mapping supports environmental monitoring and related applications, but accurate pixel-level labels are difficult and costly to produce. Official hydrographic vectors provide scalable weak supervision, but they contain artifacts like boundary noise, temporal mismatch, and omissions of small water structures. We propose a two-stage framework for weakly supervised water segmentation in high resolution multispectral imagery. Stage 1 learns initial masks from rasterized vector pseudo-labels, and Stage 2 converts these masks into structured component-wise prompts for localized refinement. On a manually corrected validation set, refinement improves SegFormer-B0 from 0.9509 to 0.9535 IoU and U-Net from 0.9408 to 0.9486 IoU, with corresponding F1 gains from 0.9749 to 0.9762 and 0.9695 to 0.9736. It leads to sharper shorelines, reduced boundary spillover, and better thin-structure delineation. The results indicate that prompt-guided refinement can improve pseudo-label-based water segmentation by targeting local errors that are poorly captured by global training supervision.
Chinese Translation
高分辨率水体映射支持环境监测及相关应用，但准确的像素级标签难以获得且成本高昂。官方水文矢量提供了可扩展的弱监督，但它们包含边界噪声、时间不匹配和小水体结构遗漏等伪影。我们提出了一种用于高分辨率多光谱图像的弱监督水体分割的两阶段框架。第一阶段从栅格化的矢量伪标签中学习初始掩膜，第二阶段将这些掩膜转换为结构化的组件级提示，以进行局部精细化。在一个经过人工修正的验证集上，精细化使SegFormer-B0的IoU从0.9509提高到0.9535，U-Net的IoU从0.9408提高到0.9486，相应的F1得分从0.9749提高到0.9762，0.9695提高到0.9736。这导致了更清晰的海岸线、减少的边界溢出和更好的细结构描绘。结果表明，基于提示的精细化可以通过针对全球训练监督捕捉不佳的局部错误来改善基于伪标签的水体分割。
cs.CV / 69 / 2609.10372
PACE: Perceived-Latency-Aware Cascading Service Routing and Filler Control for QoE-Efficient Retrieval-Augmented Dialogue Serving
PACE：感知延迟意识的级联服务路由和填充控制，用于提升用户体验的检索增强对话服务
Abstract
We present the PACE, a framework for retrieval-augmented dialogue serving that formalizes Perceived Time-to-First-Response (PTFR) as a QoE objective and minimizes it under quality/cost constraints. Unlike prior work on cascaded routing, semantic caching, or adaptive retrieval, PACE jointly controls which answer source composes the response and what fills the waiting window. Deployed on a humanoid-robot sales service, it combines three mechanisms: a load-adaptive cascading router, a joint path-filler controller, and volatility-aware cache admission. On 75k CarQA requests, the cascade halves pure-LLM PTFR at P95 (0.29 vs 0.53s at c16). The adaptive controller reaches 0.41s P95, outperforming RAG by 2.4 times at high load with equal quality. The filler controller cuts calls by 94% with zero conflict. Volatility-aware admission reduces stale answers from 86% to 0%. A gating rule ensures the controller never worse than the baseline, with exposure bounded by one hold period. This is the first quantification of filler-answer conflict risk in deployed services.
Chinese Translation
我们提出了PACE，一个用于检索增强对话服务的框架，该框架将感知首次响应时间（PTFR）形式化为用户体验（QoE）目标，并在质量/成本约束下最小化该目标。与之前关于级联路由、语义缓存或自适应检索的研究不同，PACE共同控制响应的答案来源和填充等待窗口的内容。在一个类人机器人销售服务中部署，PACE结合了三种机制：负载自适应的级联路由器、联合路径填充控制器和波动感知缓存接纳。在75,000个CarQA请求中，级联将纯LLM的PTFR在P95时减半（0.29秒对比0.53秒，c16）。自适应控制器在高负载下以相同质量达到0.41秒的P95，性能比RAG高出2.4倍。填充控制器将调用减少了94%，且没有冲突。波动感知接纳将过时答案的比例从86%降低到0%。一个门控规则确保控制器的表现不低于基线，且暴露时间限制在一个保持周期内。这是对已部署服务中填充答案冲突风险的首次量化。
cs.CV / 70 / 2609.10376
Shape-guided Gaussian Splatting for Sparse-View X-ray 3D Reconstruction
形状引导的高斯喷溅用于稀视角 X 射线 3D 重建
Abstract
Sparse-view X-ray 3D reconstruction is essential for reducing radiation exposure, but recovering a density field from a handful of X-ray projections is severely ill-posed. Recently, 3D Gaussian Splatting has achieved state-of-the-art performance in sparse-view reconstruction by representing the volume using explicit, optimized primitives, but it requires dozens of projected views. With fewer views, reconstruction quality degrades severely since the explicit primitives are optimized freely without any anatomical information. Anatomical structures, in contrast, share similar geometry and density across a population. Their variations are bounded within a limited range that statistical shape models can capture. This paper proposes a shape-guided Gaussian splatting framework for sparse-view X-ray 3D reconstructions. Our contribution lies in driving Gaussian positions toward anatomically valid configurations, alongside atlas-based density regularization. Our method ensures anatomically consistent reconstruction and improves PSNR by 2.83 dB over a state-of-the-art Gaussian splatting baseline with as few as 5 views. Code Available: https://github.com/polyshape-lab/ShapeGuidedGaussian
Chinese Translation
稀视角 X 射线 3D 重建对于减少辐射暴露至关重要，但从少量 X 射线投影中恢复密度场是一个严重的病态问题。最近，3D 高斯喷溅通过使用显式优化的原语表示体积，在稀视角重建中取得了最先进的性能，但它需要数十个投影视角。在视角较少的情况下，重建质量严重下降，因为显式原语在没有任何解剖信息的情况下自由优化。相比之下，解剖结构在一个群体中共享相似的几何形状和密度。它们的变化被限制在一个有限的范围内，统计形状模型可以捕捉到这一点。本文提出了一种形状引导的高斯喷溅框架，用于稀视角 X 射线 3D 重建。我们的贡献在于将高斯位置引导到解剖学上有效的配置，同时进行基于图谱的密度正则化。我们的方法确保了解剖学一致的重建，并在仅使用 5 个视角的情况下，相比于最先进的高斯喷溅基线提高了 2.83 dB 的峰值信噪比（PSNR）。代码可用： https://github.com/polyshape-lab/ShapeGuidedGaussian
cs.CV / 71 / 2609.10387
Enhanced Deformable Convolution with Center-invariant Offset and Edge-aware Mask
增强型可变形卷积与中心不变偏移和边缘感知掩码
Abstract
Deformable convolution networks have recently become popular for many computer vision tasks, especially for semantic segmentation, because of their exceptional capabilities in dynamic spatial modeling. However, due to the dense deformable offsets and the lack of longer-range dependencies, they can not fully adopt proper and precise deformations for feature representations. To tackle the issues, in this paper, we propose Enhanced Deformable ConvNets (EDCN) for semantic segmentation. Specifically, a novel Enhanced Deformable Convolution (EDC) is exploited in the decoder, which integrates the Center-invariant Offset Module (COM) and Edge-aware Mask Module (EMM). The COM employs larger kernels and eliminates deformations at the kernel center, obtaining offsets that are more in line with the target from richer spatial information. Concurrently, the EMM obtains the significance of image content via Sobel edge detection, then selectively applies deformations based on the content significance, minimizing unnecessary deformations associated with relatively less important information, thereby avoiding impact from less informative regions. Experiments show that EDC outperforms state-of-the-art deformable convolution variants, including Deformable ConvNets V1-V4 and Entire Deformable ConvNets, across mainstream segmentation datasets with various decoder settings. Moreover, ablation studies confirm the effectiveness of each component. In addition, visualizations illustrate that EDC enhances spatial adaptation and target focus. We further analyze the extendibility of EDC to larger kernels on the image classification benchmark. Code will be publicly released.
Chinese Translation
可变形卷积网络因其在动态空间建模方面的卓越能力，最近在许多计算机视觉任务中，尤其是语义分割中变得越来越流行。然而，由于密集的可变形偏移和缺乏长距离依赖，它们无法充分采用适当且精确的变形来表示特征。为了解决这些问题，本文提出了用于语义分割的增强型可变形卷积网络（Enhanced Deformable ConvNets，EDCN）。具体而言，在解码器中采用了一种新颖的增强型可变形卷积（Enhanced Deformable Convolution，EDC），该卷积集成了中心不变偏移模块（Center-invariant Offset Module，COM）和边缘感知掩码模块（Edge-aware Mask Module，EMM）。COM使用更大的卷积核，并消除了卷积核中心的变形，从而获得更符合目标的偏移，这些偏移来自更丰富的空间信息。同时，EMM通过Sobel边缘检测获取图像内容的重要性，然后根据内容的重要性选择性地应用变形，最小化与相对不重要信息相关的多余变形，从而避免对信息较少区域的影响。实验表明，EDC在主流分割数据集上，针对各种解码器设置，优于现有的可变形卷积变体，包括Deformable ConvNets V1-V4和Entire Deformable ConvNets。此外，消融研究证实了每个组件的有效性。此外，视觉化结果表明，EDC增强了空间适应性和目标聚焦。我们进一步分析了EDC在图像分类基准上扩展到更大卷积核的能力。代码将公开发布。
cs.CV / 72 / 2609.10456
Advanced Brain Tissue Imaging with Data-Consistent Diffusion Priors in Laminographic X-Ray Nanoimaging
基于数据一致性扩散先验的层析X射线纳米成像中的高级脑组织成像
Abstract
Nanoscale imaging of mammalian brains is critical for connectomics. X-ray laminography enables high-throughput imaging of extended, plate-like biological specimens. However, the tilted acquisition geometry leads to incomplete Fourier-space coverage, giving rise to a missing-cone of information. Conventional reconstruction methods cannot recover unmeasured information within the cone, resulting in artifacts that distort fine brain structures. While resolving these requires modeling 3D structure, direct 3D deep learning approaches are limited by data scarcity and computational cost. Here we introduce LUCID (Laminography with Unified Consistent Diffusion), a framework that combines multi-view diffusion priors with projection-domain data consistency. LUCID integrates complementary 3D structural information while enforcing strict alignment with the laminography forward model. On simulated datasets, LUCID substantially improves spatial fidelity and restores missing Fourier components, outperforming baseline methods. Applied to experimental laminography data, LUCID generalizes robustly despite being trained exclusively on fully sampled tomographic volumes, and effectively recovers unmeasured Fourier information.
Chinese Translation
对哺乳动物大脑的纳米级成像对于连接组学至关重要。X射线层析成像能够高通量地成像扩展的平板状生物样本。然而，倾斜的采集几何形状导致傅里叶空间覆盖不完整，产生信息缺失锥。传统重建方法无法恢复锥内未测量的信息，导致失真伪影，扭曲细微的脑结构。解决这些问题需要建模三维结构，而直接的三维深度学习方法受到数据稀缺和计算成本的限制。在此，我们引入LUCID（统一一致扩散的层析成像），一个将多视角扩散先验与投影域数据一致性相结合的框架。LUCID整合了互补的三维结构信息，同时严格遵循层析成像前向模型的对齐。在模拟数据集上，LUCID显著提高了空间保真度并恢复了缺失的傅里叶分量，优于基线方法。应用于实验层析成像数据，LUCID在仅基于完全采样的断层体积训练的情况下仍能稳健地推广，并有效恢复未测量的傅里叶信息。
cs.CV / 73 / 2609.10469
AgroVisNet: A lightweight Convolutional Network and the BD-PlantDX Expert-Validated Benchmark for Radish, Potato and Pointed Gourd Disease Classification
AgroVisNet：一种轻量级卷积网络及其针对萝卜、土豆和尖果南瓜病害分类的BD-PlantDX专家验证基准
Abstract
Automated plant disease diagnosis is increasingly deployed on farmer-held devices in regions where agronomic expertise is scarce and network connectivity is unreliable. Three obstacles limit its practical value: public benchmarks are dominated by a small set of non-native crops, region-specific datasets are rarely validated by domain experts, and the architectures that reach competitive accuracy carry parameter budgets that are unsuited to low-cost hardware. We propose AgroVisNet, a compact convolutional network trained from scratch, together with BD-PlantDX, an expert-validated benchmark of 12,432 field images spanning 12 classes of radish, potato and pointed gourd in healthy and diseased states, collected across the Bogura and Nilphamari districts of Bangladesh. AgroVisNet couples grouped bottleneck residual blocks carrying sequential channel and spatial attention with multi-scale depthwise blocks and a dual-pooling classification head, reaching 290,572 trainable parameters. On BD-PlantDX the model attains 99.52% test accuracy and 99.52% weighted F1, exceeding all six ImageNet-pretrained lightweight backbones evaluated under an identical protocol while using 8.7 to 16.8 times fewer parameters and 1.3 to 8.5 times fewer multiply-accumulate operations. Exported for deployment, the model quantises to a 0.46 MB full-integer network at a 0.22 percentage-point accuracy cost and classifies an image in 8.40 ms on a single CPU. Across five random seeds accuracy remains at 99.57 +- 0.10%, a ten-variant ablation isolates the contribution of each component, and the same architecture transfers without redesign to two independently collected datasets at 98.71% and 99.05% accuracy. Grad-CAM evidence indicates that predictions rest on lesion-bearing leaf regions rather than on background cues.
Chinese Translation
自动化植物病害诊断在农业专业知识匮乏和网络连接不可靠的地区越来越多地应用于农民手持设备上。然而，三个障碍限制了其实际价值：公共基准主要由少量非本地作物主导，特定地区的数据集很少经过领域专家验证，而达到竞争性准确率的架构所需的参数预算不适合低成本硬件。我们提出了AgroVisNet，一种从零开始训练的紧凑型卷积网络，以及BD-PlantDX，一个经过专家验证的基准，包含12,432张覆盖健康和病害状态下的萝卜、土豆和尖果南瓜的田间图像，数据收集自孟加拉国博古拉和尼尔法马里地区。AgroVisNet结合了带有顺序通道和空间注意力的分组瓶颈残差块、多尺度深度卷积块以及双池化分类头，达到290,572个可训练参数。在BD-PlantDX上，该模型获得了99.52%的测试准确率和99.52%的加权F1分数，超过了在相同协议下评估的六种ImageNet预训练轻量级骨干网络，同时使用的参数数量减少了8.7到16.8倍，乘加操作减少了1.3到8.5倍。导出用于部署时，该模型量化为0.46 MB的全整数网络，准确率损失为0.22个百分点，并在单个CPU上以8.40毫秒的时间对图像进行分类。在五个随机种子下，准确率保持在99.57 ± 0.10%，十种变体消融实验隔离了每个组件的贡献，相同架构在两个独立收集的数据集上转移时的准确率为98.71%和99.05%。Grad-CAM证据表明，预测依赖于带有病变的叶片区域，而非背景线索。
cs.CV / 74 / 2609.10489
Artificial Intelligence Literacy and Sustainable Development: An Ethical Governance and Development Goals Framework
人工智能素养与可持续发展：伦理治理与发展目标框架
Abstract
AI literacy provides foundational competencies that support ethical, transparent, and sustainable technological development, although higher-order capabilities such as governance, critical evaluation, and strategic decision-making extend beyond basic literacy into advanced levels of AI competency. This study positions AI literacy as a governance capacity that complements and strengthens all 17 SDGs. It introduces a six-level taxonomy of artificial intelligence reasoning and ethics that extends traditional learning models by incorporating ethical judgement and strategic foresight. This taxonomy forms the foundation of an integrated framework linking education, governance, and sustainable development. A survey of 300 participants from diverse professional backgrounds within a national context which reveals strong technical awareness but limited ethical and governance readiness, highlighting critical gaps in public capacity to manage artificial intelligence responsibly. Findings show that ethical reasoning and reflective thinking are the strongest predictors of sustainable and trustworthy artificial intelligence use. The study proposed to embed literacy-based competencies into curricula, institutional policies, and governance mechanisms to accelerate equitable and responsible progress toward sustainable development goals
Chinese Translation
人工智能素养提供了支持伦理、透明和可持续技术发展的基础能力，尽管治理、批判性评估和战略决策等更高阶能力超出了基本素养，进入了人工智能能力的高级水平。本研究将人工智能素养定位为一种治理能力，补充并加强所有17个可持续发展目标（SDGs）。它引入了一个六级的人工智能推理与伦理分类法，该分类法通过纳入伦理判断和战略前瞻性，扩展了传统学习模型。该分类法构成了一个将教育、治理与可持续发展相结合的综合框架的基础。对来自不同专业背景的300名参与者进行的调查显示，尽管技术意识强烈，但伦理和治理准备不足，突显了公众在负责任地管理人工智能方面的关键缺口。研究结果表明，伦理推理和反思性思维是可持续和可信赖的人工智能使用的最强预测因素。研究建议将基于素养的能力纳入课程、机构政策和治理机制，以加速朝向可持续发展目标的公平和负责任的进展。
cs.CV / 75 / 2609.10495
Cross-Model Agreement as a Deployment-Time Reliability Signal for Automatic Polyp Segmentation
跨模型一致性作为自动息肉分割的部署时可靠性信号
Abstract
In real-time colonoscopy, ground-truth annotations are unavailable at inference, so polyp segmentation models can fail silently. We propose Referee-Based Quality Estimation (RBQE), a reference-free framework measuring agreement between a primary segmentation model and an independently trained referee on the same image. RBQE is evaluated on a standardized 1,223-image external benchmark drawn from four public datasets, using four referee configurations chosen to separate two design axes: referee independence and architectural diversity. Using a common Agreement Dice descriptor, a same-architecture referee differing from the primary model only in random initialization already yields a useful reliability signal (ROC-AUC = 0.923), showing that independent training alone is sufficient. Cross-architecture referees improve further: SegFormer-B0 achieves the strongest performance (ROC-AUC = 0.960), significantly outperforming the same-architecture control and UNet++, and exceeding a representative Test-Time Augmentation baseline by 0.055 ROC-AUC under an identical protocol, whereas a prompt-coupled MedSAM referee underperforms despite maximal architectural diversity. Because empty-mask agreement is trivially separable, we also report a restricted evaluation excluding such cases: ROC-AUC falls to 0.876 (SegFormer-B0, 1,046 images) and 0.783 (same-architecture control, 975 images), yet RBQE's margin over both baselines widens on this identical subset. RBQE additionally increases the mean Dice of retained predictions as low-agreement cases are progressively rejected, supporting selective prediction, and requires only one additional deterministic referee forward pass at inference. Our study therefore supports cross-model agreement as a practical, interpretable reliability framework for automated polyp segmentation.
Chinese Translation
在实时结肠镜检查中，推理时无法获得真实标注，因此息肉分割模型可能会静默失败。我们提出了一种无参考的质量评估框架——基于裁判的质量评估（Referee-Based Quality Estimation, RBQE），该框架测量主分割模型与独立训练的裁判在同一图像上的一致性。RBQE在一个标准化的1,223幅图像的外部基准上进行评估，该基准来自四个公共数据集，使用四种裁判配置以区分两个设计轴：裁判独立性和架构多样性。使用通用的一致性Dice描述符，只有在随机初始化上与主模型不同的同架构裁判已经提供了有用的可靠性信号（ROC-AUC = 0.923），表明仅独立训练就足够了。跨架构裁判进一步提高了性能：SegFormer-B0达到了最强的表现（ROC-AUC = 0.960），显著优于同架构控制和UNet++，并在相同协议下超越了一个代表性的测试时增强基线0.055 ROC-AUC，而尽管具有最大架构多样性，结合提示的MedSAM裁判表现不佳。由于空掩膜一致性是显而易见的可分离，我们还报告了排除此类情况的限制评估：ROC-AUC降至0.876（SegFormer-B0，1,046幅图像）和0.783（同架构控制，975幅图像），然而RBQE在这一相同子集上的优势扩大。RBQE还提高了保留预测的平均Dice，因为低一致性案例被逐步拒绝，支持选择性预测，并且在推理时仅需一次额外的确定性裁判前向传播。因此，我们的研究支持跨模型一致性作为自动息肉分割的一个实用且可解释的可靠性框架。
cs.CV / 76 / 2609.10498
Field Converter: Geometry-Initialized Temporal Residual Refinement for World-Grounded Player Pose Estimation from Soccer Broadcasts
场地转换器：基于几何初始化的时间残差细化用于从足球广播中估计世界坐标系中的玩家姿态
Abstract
Recovering 3D human pose from monocular sports broadcasts remains challenging when players must be localized in a shared metric world coordinate system rather than only reconstructed relative to their own body. We introduce Field Converter, a geometry-initialized temporal residual framework for world-grounded 3D player pose estimation from calibrated soccer broadcasts. Our method first uses camera and pitch geometry to initialize the player root through ray-ground intersection, then predicts a temporal residual correction from pose, image, camera, and geometric cues. On match-disjoint evaluation sequences, residual refinement reduces root error from 49cm with geometry alone to 14cm with a frame-wise MLP and 10cm with a TCN, while a Transformer achieves a comparable 11cm. The resulting world-space MPJPE reaches 13.2cm, and ablations show that residual prediction clearly outperforms direct global-root regression while temporal context matters more than the specific temporal backbone. Failure analysis further identifies airborne motion as the main limitation of the ground-based geometric initialization.
Chinese Translation
从单目体育广播中恢复3D人类姿态仍然具有挑战性，因为玩家必须在共享的度量世界坐标系中定位，而不仅仅是相对于他们自己的身体进行重建。我们提出了场地转换器（Field Converter），这是一种基于几何初始化的时间残差框架，用于从校准的足球广播中进行世界坐标系中的3D玩家姿态估计。我们的方法首先利用相机和场地几何通过光线与地面的交点初始化玩家根部，然后从姿态、图像、相机和几何线索中预测时间残差修正。在比赛不重叠的评估序列中，残差细化将仅使用几何时的根误差从49厘米降低到使用逐帧多层感知器（MLP）时的14厘米，以及使用时间卷积网络（TCN）时的10厘米，而变压器（Transformer）则达到了可比的11厘米。最终的世界空间均方根关节位置误差（MPJPE）达到13.2厘米，消融实验表明，残差预测明显优于直接的全局根回归，而时间上下文比特定的时间骨干网络更为重要。失败分析进一步确定空中运动是基于地面的几何初始化的主要限制因素。
cs.CV / 77 / 2609.10518
BrainTaskonomy: Learning How to Pretrain and What to Transfer in fMRI Foundation Models
脑任务分类学：学习如何预训练及在 fMRI 基础模型中转移什么
Abstract
fMRI foundation models increasingly aggregate heterogeneous data across brain states, cohorts, and acquisition settings, yet pretraining domains are commonly treated as a flat mixture and downstream tasks are adapted independently. We study whether measured learning relations can organize both stages without modifying the backbone. During pretraining, a lightweight Brain-DiT proxy estimates difficulty and directed facilitation across ten fMRI domains, yielding a priority-guided cumulative domain curriculum combined with high-to-low-noise timestep scheduling and joint consolidation. During adaptation, controlled first- and higher-order transfer across fifteen tasks constructs a directed taskonomy, from which budgeted integer programming (BIP) selects directly supervised source tasks and target-specific routes. The joint priority-domain and high-to-low-timestep curriculum reduces v-NMSE, PSD-NMSE, and FC-MSE by 6.5%, 16.3%, and 10.5%, respectively, relative to uniform sampling over both dimensions, and shows strong downstream performance across six in- and out-of-domain tasks. The taskonomy reveals asymmetric, target-dependent transfer, while exploratory sealed-test evaluation shows larger descriptive gains for BIP policies when higher-order route spaces are available than for matched random controls. Together, these findings support organizing fMRI pretraining and adaptation by measured learning relations rather than treating domains and tasks as independent flat sets.
Chinese Translation
fMRI 基础模型越来越多地聚合来自不同脑状态、群体和采集设置的异质数据，但预训练领域通常被视为一个平坦的混合体，而下游任务则独立适应。我们研究了测量的学习关系是否可以在不修改主干的情况下组织这两个阶段。在预训练过程中，一个轻量级的 Brain-DiT 代理估计了十个 fMRI 领域中的难度和定向促进，产生了一个优先引导的累积领域课程，结合高到低噪声的时间步调度和联合巩固。在适应阶段，控制的一级和高阶转移跨越十五个任务构建了一个定向任务分类学，从中预算整数规划（BIP）选择直接监督的源任务和特定目标的路径。联合优先领域和高到低时间步课程相较于在两个维度上均匀采样，分别降低了 v-NMSE、PSD-NMSE 和 FC-MSE 6.5%、16.3% 和 10.5%，并在六个领域内外任务中显示出强大的下游性能。任务分类学揭示了不对称的、依赖目标的转移，而探索性封闭测试评估显示，当高阶路径空间可用时，BIP 策略的描述性增益大于匹配随机控制。综合来看，这些发现支持通过测量的学习关系来组织 fMRI 的预训练和适应，而不是将领域和任务视为独立的平坦集合。
cs.CV / 78 / 2609.10524
Precision in Rice Variety Classification using Stacking-Based Ensemble Learning
基于堆叠集成学习的水稻品种分类精度研究
Abstract
Rice, a staple food for a significant portion of the global population, exhibits remarkable diversity in its varieties, presenting substantial challenges for accurate identification by consumers, traders, and farmers. This complexity often facilitates fraudulent practices, such as the unauthorized mixing of rice types, which undermines quality and trust in the supply chain. Despite its critical importance, existing research falls short of providing robust and efficient methods for precise rice variety classification based on external characteristics like color, size, and texture. To address this gap, our study introduces a comprehensive rice variety identification framework designed to enhance transparency and quality assurance. We developed a stacked ensemble model tailored for rice variety classification and curated a comprehensive dataset comprising 20 rice varieties, each distinguished by unique visual attributes. The proposed approach achieved an unprecedented classification accuracy of 100%. Furthermore, we integrated our model into a mobile application, enabling even novice users to effortlessly identify rice varieties using grain images from a smartphone camera. These findings underscore the transformative potential of advanced machine learning techniques in mitigating fraudulent practices and ensuring stringent rice quality control. Our work holds significant implications for agricultural stakeholders, paving the way for automated crop identification systems and advancing precision agriculture practices.
Chinese Translation
水稻是全球大量人口的重要主食，具有显著的品种多样性，这给消费者、贸易商和农民的准确识别带来了重大挑战。这种复杂性往往助长了欺诈行为，例如未经授权的水稻品种混合，损害了供应链的质量和信任。尽管其重要性不言而喻，现有研究在基于颜色、大小和质地等外部特征进行精确水稻品种分类方面仍显不足。为了解决这一问题，我们的研究提出了一个全面的水稻品种识别框架，旨在增强透明度和质量保证。我们开发了一个针对水稻品种分类的堆叠集成模型，并整理了一个包含20个水稻品种的综合数据集，每个品种都具有独特的视觉特征。所提出的方法达到了前所未有的100%的分类准确率。此外，我们将模型集成到一个移动应用程序中，使得即使是初学者也能轻松使用智能手机相机拍摄的谷物图像识别水稻品种。这些发现强调了先进机器学习技术在减少欺诈行为和确保严格水稻质量控制方面的变革潜力。我们的工作对农业利益相关者具有重要意义，为自动化作物识别系统铺平了道路，并推动了精准农业实践的发展。
cs.CV / 79 / 2609.10531
Guiding Image-to-3D Generation with Test-Time Partial Observations
通过测试时部分观测引导图像到3D生成
Abstract
Image-to-3D models can generate visually compelling 3D assets from a single RGB image, but their geometry is often only loosely constrained by the available observations, limiting their use in applications that require geometric fidelity. In many real-world settings, however, partial geometric observations of the object may be available at test time. We introduce a training-free framework for incorporating such evidence into pretrained image-to-3D generative models without retraining or finetuning. To do this, we guide generation using a ray-consistent observation likelihood defined over the model's occupancy representation, combining surface occupancy and free-space evidence. Applied to SAM 3D and its multi-view extension, our approach substantially improves geometric fidelity across different levels of observability, as well as visual quality. Our results demonstrate that pretrained image-to-3D models can effectively integrate partial geometric observations through explicit test-time guidance, complementing their learned generative priors without modifying the underlying model.
Chinese Translation
图像到3D模型可以从单个RGB图像生成视觉上引人注目的3D资产，但其几何形状通常仅受到可用观测的松散约束，这限制了其在需要几何保真度的应用中的使用。然而，在许多现实世界的场景中，测试时可能会获得对象的部分几何观测。我们提出了一种无训练的框架，以将此类证据纳入预训练的图像到3D生成模型中，而无需重新训练或微调。为此，我们使用在模型的占用表示上定义的光线一致性观测似然性来引导生成，结合表面占用和自由空间证据。应用于SAM 3D及其多视图扩展，我们的方法在不同的可观测性水平上显著提高了几何保真度以及视觉质量。我们的结果表明，预训练的图像到3D模型可以通过明确的测试时引导有效地整合部分几何观测，补充其学习的生成先验，而无需修改基础模型。
cs.CV / 80 / 2609.10540
Programmable World Model
可编程世界模型
Abstract
Recent video world models generate increasingly realistic and interactive visual experiences, yet lack reliable mechanisms for maintaining persistent world state and enforcing programmable rules over extended interactions. We introduce Programmable World Model, a framework that decouples world-state evolution from visual observation generation. An agent translates natural-language instructions into executable programs that specify entity states and state-transition rules, enabling direct control over individual entities and their interactions. A lightweight engine executes these programs to update and maintain an explicit, persistent global world state, including off-screen entities and non-visual attributes. To connect world state with visual generation, we introduce state-augmented 3D oriented bounding boxes (OBBs) as an intermediate representation. This representation, together with the target camera trajectory, is deterministically compiled into pixel-aligned spatiotemporal conditioning signals for a pretrained video model serving as the generative renderer. This design allows users to create playable games with predefined mechanics, direct control over individual entities, and persistent world state throughout gameplay. We further introduce CombatStateBench, a benchmark for evaluating programmable world models. On CombatStateBench, our method achieves 94% Count Accuracy and 98% State Accuracy, substantially outperforming existing interactive video world models while supporting coherent long-horizon generation. These results demonstrate the effectiveness of separating explicit state evolution from generative rendering for building persistent, programmable worlds.
Chinese Translation
最近的视频世界模型生成越来越真实和互动的视觉体验，但在维护持久的世界状态和在扩展交互中强制可编程规则方面缺乏可靠的机制。我们提出了可编程世界模型（Programmable World Model），这是一个将世界状态演变与视觉观察生成解耦的框架。代理将自然语言指令转换为可执行程序，这些程序指定实体状态和状态转换规则，从而实现对单个实体及其交互的直接控制。一个轻量级引擎执行这些程序，以更新和维护一个明确的、持久的全局世界状态，包括屏幕外实体和非视觉属性。为了将世界状态与视觉生成连接起来，我们引入了状态增强的三维定向边界框（OBBs）作为中间表示。该表示与目标相机轨迹一起，确定性地编译成与像素对齐的时空条件信号，用于预训练的视频模型，作为生成渲染器。该设计使用户能够创建具有预定义机制的可玩游戏，直接控制单个实体，并在整个游戏过程中保持持久的世界状态。我们进一步引入了CombatStateBench，这是一个用于评估可编程世界模型的基准。在CombatStateBench上，我们的方法实现了94%的计数准确率和98%的状态准确率，显著超越现有的互动视频世界模型，同时支持一致的长时间生成。这些结果证明了将明确的状态演变与生成渲染分离在构建持久的、可编程的世界中的有效性。
cs.AI / 1 / 2609.09203
OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist Workflows
OpenDiscoveryTrace：用于评估人工智能科学家工作流程的过程追踪
Abstract
Existing benchmarks for autonomous AI scientists evaluate only final outputs---generated code, hypotheses, or papers---yet discard the reasoning process by which those outputs were obtained. This makes it impossible to audit scientific methodology, diagnose failure modes, or distinguish systematic reasoning from fortunate guessing. We present \textbf{OpenDiscoveryTrace}, a public dataset of 558 complete AI scientific agent trajectories that captures how models reason, not just what they produce. Each trajectory records a structured 9-field-per-step trace---including thoughts, tool calls, observations, errors, revision triggers, and self-reported confidence---as models execute 124 scientific tasks spanning drug discovery, materials science, genomics, and scientific literature analysis. The dataset covers seven models: three frontier models (GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro; 124 trajectories each, fully balanced across domains and difficulty levels) and four open-weight models (Qwen2.5-7B, Mistral-7B-v0.3, Phi-3.5-mini, and Qwen2.5-1.5B; 30 each), plus 60 live-retrieval variant trajectories. Pilot analysis on 363 LLM-judged trajectories reveals that process traces expose behavioral differences invisible to output-only evaluation: all three frontier models achieve comparable success rates (84--89%), yet Claude Opus 4.6 produces 30$\times$ more errors than GPT-5.4 (2.5 vs. 0.08 per trajectory, $p < 0.0001$, Cliff's $\delta = 0.613$), with qualitatively different error profiles---66.7% tool misuse for Claude versus 83.6% reasoning errors for GPT-5.4. We define five benchmark tasks with baselines from logistic regression, random forests, LSTMs, and Transformer models. The dataset, trace schema, agent harness, and benchmark definitions are publicly available under CC BY 4.0 to support research on process-level evaluation, scientific agent auditing, and AI governance.
Chinese Translation
现有的自主人工智能科学家基准仅评估最终输出——生成的代码、假设或论文——而忽略了获得这些输出的推理过程。这使得审计科学方法、诊断失败模式或区分系统性推理与偶然猜测变得不可能。我们提出了 extbf{OpenDiscoveryTrace}，这是一个包含558个完整人工智能科学代理轨迹的公共数据集，捕捉模型的推理过程，而不仅仅是它们的产出。每个轨迹记录了一个结构化的每步9字段追踪——包括思考、工具调用、观察、错误、修正触发器和自我报告的信心——当模型执行124个科学任务时，这些任务涵盖了药物发现、材料科学、基因组学和科学文献分析。该数据集涵盖了七个模型：三个前沿模型（GPT-5.4、Claude Opus 4.6和Gemini 3.1 Pro；每个模型124个轨迹，跨领域和难度水平完全平衡）和四个开放权重模型（Qwen2.5-7B、Mistral-7B-v0.3、Phi-3.5-mini和Qwen2.5-1.5B；每个模型30个轨迹），以及60个实时检索变体轨迹。对363个由大型语言模型（LLM）评估的轨迹的初步分析表明，过程追踪揭示了仅通过输出评估无法察觉的行为差异：所有三个前沿模型的成功率相当（84%--89%），然而Claude Opus 4.6产生的错误数量是GPT-5.4的30倍（每个轨迹2.5 vs. 0.08，$p < 0.0001$，Cliff's $eta = 0.613$），且错误特征 qualitatively不同——Claude的66.7%是工具误用，而GPT-5.4的83.6%是推理错误。我们定义了五个基准任务，基线来自逻辑回归、随机森林、LSTM和Transformer模型。该数据集、追踪模式、代理框架和基准定义在CC BY 4.0下公开，以支持对过程级评估、科学代理审计和人工智能治理的研究。
cs.AI / 2 / 2609.09226
Adaptive Entangled Game Modules in Artificial General Intelligence
人工通用智能中的自适应纠缠游戏模块
Abstract
We introduce a probability-wave framework for modeling the collective behavior of interacting adaptive agents, deriving testable eigenmodes through a generalized behavioral intelligence (GBI) nonlocal probability-wave equation. This framework captures a broad range of human intelligence behaviors with analytical mechanisms and offers an indirect method to examine the Liu-Chen-Ao (LCA) hypothesis of nonlocal entangled nerve fibers in the brain through collective trader behaviors. Our empirical analysis of Chinese intraday stock market data demonstrates that adaptive entangled game modes explain 82-94% (89% overall) of observed decision patterns, a sharp contrast to the predictions of neoclassical finance based on independent rational agents. Moreover, 2-12% of behaviors show adaption to intraday news, events, and environments, characterized by dual equilibrium states and abrupt reference point shifts, while purely independent modes occur in less than 5% of cases. These findings empirically support the LCA hypothesis, as observable trading behaviors reflect underlying brain mechanisms and internal intelligence decision-making in behavioral psychology. Our results highlight the necessity of incorporating adaptive entangled game modules into artificial general intelligence (AGI) architectures, addressing the limitations of conventional artificial neural network (ANN)-based AI, which relies on trillions of opaque parameters. By integrating ANN-based AI with probability-wave-based entangled-brain simulations, machine learning can enrich AGI foundation models (FMs) and facilitate the development of human-like processing units (HPUs) that leverage brain-inspired mechanisms. Such HPUs may ultimately create more compact, efficient, and robust AGI systems, particularly for embodied intelligence and robotics.
Chinese Translation
我们引入了一种概率波框架，用于建模相互作用的自适应智能体的集体行为，通过广义行为智能（Generalized Behavioral Intelligence, GBI）非局部概率波方程推导可测试的特征模式。该框架通过分析机制捕捉了广泛的人类智能行为，并提供了一种间接的方法，通过集体交易者行为来检验大刘-陈-敖（Liu-Chen-Ao, LCA）假设，即大脑中非局部纠缠神经纤维的存在。我们对中国日内股票市场数据的实证分析表明，自适应纠缠游戏模式解释了观察到的决策模式的82-94%（总体为89%），与基于独立理性主体的古典金融预测形成鲜明对比。此外，2-12%的行为显示出对日内新闻、事件和环境的适应，特征表现为双重均衡状态和突发的参考点转移，而纯粹独立的模式发生在不到5%的情况下。这些发现实证支持了LCA假设，因为可观察的交易行为反映了潜在的大脑机制和行为心理学中的内部智能决策。我们的结果强调了将自适应纠缠游戏模块纳入人工通用智能（AGI）架构的必要性，以解决传统人工神经网络（Artificial Neural Network, ANN）基础的人工智能的局限性，这种人工智能依赖于数万亿个不透明参数。通过将基于ANN的人工智能与基于概率波的纠缠大脑模拟相结合，机器学习可以丰富AGI基础模型（Foundation Models, FMs），并促进人类类处理单元（Human-like Processing Units, HPUs）的发展，这些单元利用大脑启发机制。这样的HPUs最终可能创造出更紧凑、高效和稳健的AGI系统，特别是在具身智能和机器人领域。
cs.AI / 3 / 2609.09233
Subagents vs Agent Skills: Executing Reusable Knowledge for Long-Horizon Agentic Tasks
子代理与代理技能：执行可重用知识以应对长时间跨度的代理任务
Abstract
How can language model agents effectively leverage libraries of reusable knowledge to solve long-horizon tasks? Recent work has increasingly focused on agent skills: reusable capabilities represented as skill packages, i.e., multi-file bundles containing instructions, scripts, and other resources that help agents perform specific tasks. Agent skills are typically executed by loading their skill instructions into an agent's context and relying on the agent to follow them. As task horizons grow, however, this approach becomes increasingly brittle, because reasoning quality degrades as more information accumulates in the context window. We investigate an alternative approach in which skill packages are instead invoked as subagents. Rather than loading skill instructions into the main context, subagent execution spawns fresh context windows dedicated to solving individual subtasks. We show that subagent execution outperforms agent-skill execution when skill packages expose clear input-output contracts and their instructions encode the procedural knowledge needed to fulfill those contracts. The tradeoff is additional communication overhead, as extra tokens are required to coordinate between the main agent and its subagents. Our results show that the benefit of reusable knowledge depends not only on its content, but also on how it is organized and invoked.
Chinese Translation
语言模型代理如何有效利用可重用知识库来解决长时间跨度的任务？近期的研究越来越关注代理技能：作为技能包表示的可重用能力，即包含指令、脚本和其他资源的多文件捆绑，帮助代理执行特定任务。代理技能通常通过将其技能指令加载到代理的上下文中并依赖代理遵循这些指令来执行。然而，随着任务时间跨度的增加，这种方法变得越来越脆弱，因为随着更多信息在上下文窗口中积累，推理质量会下降。我们研究了一种替代方法，其中技能包作为子代理被调用。子代理执行并不是将技能指令加载到主上下文中，而是为解决各个子任务生成新的上下文窗口。我们展示了当技能包暴露出明确的输入-输出契约且其指令编码了履行这些契约所需的程序性知识时，子代理执行的表现优于代理技能执行。其权衡在于额外的通信开销，因为需要额外的标记来协调主代理与其子代理之间的关系。我们的结果表明，可重用知识的好处不仅取决于其内容，还取决于其组织和调用方式。
cs.AI / 4 / 2609.09306
Gradland: On Phenomenal Experience, Differentiated Across Many Dimensions
Gradland：在多维度上区分的现象体验
Abstract
This paper investigates the hypothesis that the first-order structure of physical interactions, i.e. gradients or Jacobians, characterizes the structure of phenomenal experience. It does so in an idealized world inhabited by neural networks, Gradland, where the physics are known and the functions are (mostly) differentiable. The paper introduces two measures of Jacobian structure: effective rank and cohesion, based on Kirchhoff complexity. Applying the measures to a series of worked examples shows the hypothesis accounts for: (1) the duration of experience, that it can prolong over hundreds of milliseconds; (2) the difference between what is experienced vividly and obscurely; (3) the experience of texture; (4) the blooming buzzing confusion presumably experienced by newborns; (5) the difference between ideas that are held distinctly in mind and ideas that are confused; (6) what learning is like; and finally (7) the paper explains the function of rich, dense experience.
Chinese Translation
本文探讨了一个假设，即物理交互的一级结构，即梯度或雅可比矩阵，表征了现象体验的结构。研究在一个理想化的世界——Gradland中进行，该世界由神经网络构成，物理规律已知且函数（大多数情况下）是可微的。本文引入了两种雅可比结构的度量：有效秩和凝聚度，基于基尔霍夫复杂性。将这些度量应用于一系列实例表明，该假设可以解释：（1）体验的持续时间，可以延续数百毫秒；（2）生动体验与模糊体验之间的差异；（3）质感的体验；（4）新生儿可能经历的繁花似锦的混乱；（5）清晰记忆中的想法与混淆想法之间的差异；（6）学习的体验；最后，（7）本文解释了丰富、密集体验的功能。
cs.AI / 5 / 2609.09374
An Autonomous GeoAI Agent for Arctic Eco-Navigation
用于北极生态导航的自主GeoAI智能体
Abstract
Arctic maritime navigation is becoming increasingly important as changing sea-ice conditions expand seasonal accessibility while simultaneously introducing substantial operational, environmental, and community risks. Arctic route planning is inherently a multi-criteria problem: routes that improve vessel safety or efficiency may increase exposure to sea ice, sensitive ecosystems, or nearby communities. Existing routing methods prioritize travel time, fuel use, and navigational risk, often overlooking ecological and community impacts. We introduce a human-in-the-loop, multi-agent GeoAI system for Arctic eco-navigation that integrates operational, physical, ecological, and community-related criteria within a unified routing framework. Multiple specialized agents coordinate geospatial data acquisition and preparation, multi-objective route generation, and skyline-based decision support. The ecological criteria explicitly account for exposure to sensitive areas, including Essential Fish Habitat and seal critical habitat. By considering these ecosystem impacts and potential community burdens while keeping consequential value judgments under human control, the framework supports safer, more transparent, and socially responsible Arctic navigation. Project page and code are publicly available. https://samiraat.github.io/Arctic-Eco-Navigation-Agent/, https://github.com/samiraat/Arctic-Eco-Navigation-Agent
Chinese Translation
随着海冰条件的变化，北极海上导航变得愈加重要，季节性可达性不断扩大，同时也带来了显著的操作、环境和社区风险。北极航线规划本质上是一个多标准问题：改善船舶安全或效率的航线可能会增加对海冰、敏感生态系统或附近社区的暴露。现有的航线规划方法优先考虑旅行时间、燃料使用和导航风险，往往忽视生态和社区影响。我们提出了一种人机协作的多智能体GeoAI系统，用于北极生态导航，该系统在统一的航线框架内整合了操作、物理、生态和社区相关标准。多个专业智能体协调地理空间数据的获取和准备、多目标航线生成以及基于天际线的决策支持。生态标准明确考虑了对敏感区域的暴露，包括重要鱼类栖息地和海豹关键栖息地。通过考虑这些生态系统影响和潜在的社区负担，同时将重要的价值判断控制在人工控制之下，该框架支持更安全、更透明和更具社会责任感的北极导航。项目页面和代码已公开可用。https://samiraat.github.io/Arctic-Eco-Navigation-Agent/, https://github.com/samiraat/Arctic-Eco-Navigation-Agent
cs.AI / 6 / 2609.09395
The Menu Is an Execution Prior: State-Path Tool Menus for Online Agents
菜单是执行先验：在线智能体的状态路径工具菜单
Abstract
Language models act through tools, yet practical agents face libraries containing thousands of interfaces. We introduce the tool menu as the short, ordered subset of available tools shown to an agent before execution. The agent can call only tools in this menu. Multi-step tasks require the final action and the prerequisite tools that create its inputs in a usable order. Current constructors rank tools by request relevance, which can surface the final action while omitting or delaying less obvious producers. We introduce the state path, a pre-execution route from the observable request state to the desired outcome, and propose State-Path Tool Menu to learn it. Our framework treats the menu as an execution prior over these routes. Its encoder represents which tools can run from the current state, how their outputs satisfy later inputs, and which orders recur in training paths. A retriever covers an executable entry, the missing-input producers, and the final action. A reranker then places producers before consumers. On ToolBench, our menu raises online success from 0.737 to 0.898 and outperforms retrieval, reranking, generation, and routing baselines without changing the agent. The State-Path menu also covers more complete chains with 32 tools than the official list covers with 128, and its success gain persists across executor families with different model capacities. Our code is at https://github.com/Met2348/State-Path.
Chinese Translation
语言模型通过工具进行操作，但实际的智能体面临包含数千个接口的库。我们引入了工具菜单，作为在执行前向智能体展示的可用工具的简短、有序子集。智能体只能调用此菜单中的工具。多步骤任务需要最终动作以及以可用顺序创建其输入的先决工具。目前的构造器通过请求相关性对工具进行排名，这可能会突出最终动作，同时省略或延迟不太明显的生产者。我们引入了状态路径，这是从可观察的请求状态到期望结果的执行前路线，并提出了状态路径工具菜单以学习它。我们的框架将菜单视为这些路径上的执行先验。其编码器表示从当前状态可以运行哪些工具，它们的输出如何满足后续输入，以及在训练路径中哪些顺序会重现。检索器涵盖可执行条目、缺失输入的生产者和最终动作。然后，重新排序器将生产者放在消费者之前。在ToolBench上，我们的菜单将在线成功率从0.737提高到0.898，并在不更改智能体的情况下超越了检索、重新排序、生成和路由基线。状态路径菜单还涵盖了比官方列表使用128个工具更完整的32个工具链，其成功增益在不同模型能力的执行器家族中持续存在。我们的代码可在https://github.com/Met2348/State-Path找到。
cs.AI / 7 / 2609.09413
Decision-Focused Active Learning for Scale-Aware Critical-Materials Recovery
面向决策的主动学习用于规模感知的关键材料回收
Abstract
Choosing a recovery process for scale-up requires connecting laboratory results with product requirements, process costs, and scale effects. We analyze records from Pacific Northwest National Laboratory's Computer Intelligence for Critical Element Recovery and Optimization (CICERO) workflow for autonomous selective precipitation. Active learning uses prior results to choose experiments. In a conditional retrospective benchmark with fitted models and recycled neodymium-iron-boron (NdFeB) magnet records, active learning finds the best recorded result with fewer experiments than nonadaptive space filling. Enrichment is the selected rare-earth-to-iron ratio relative to that in the feed. Adaptive policies reach the recorded enrichment maximum by 16 to 24 wells (individual experiments), versus 48. Our two-stage reconstruction ties two adaptive alternatives at 16 wells. Conditional analyses of recycled samarium-cobalt (SmCo) magnets show a Round 2 tradeoff between purity and nominal yield, the recovery fraction calculated from an assumed starting amount - NdFeB Round 1 routes differ in enrichment. Rankings for produced water from oil and gas extraction depend on phase and dilution assumptions requiring confirmation. We propose choosing batches by their expected reduction in downstream Bayes risk: the minimum expected loss among available process decisions under current beliefs. In exploratory simulations, a hybrid that filters candidates has lower estimated loss than the implemented joint search across routes and conditions. Differences involving the synthetic two-stage policy are small relative to estimation uncertainty. We outline a pre-registered prospective test under a shared loss and logging standard, requiring clarified measurements and records, a defined process decision and relevant outputs, credible economic inputs, and validation at the intended scale.
Chinese Translation
选择适合规模化的回收过程需要将实验室结果与产品要求、过程成本和规模效应相连接。我们分析了太平洋西北国家实验室的计算智能关键元素回收与优化（CICERO）工作流程中的记录，该流程用于自主选择性沉淀。主动学习利用先前的结果来选择实验。在一个条件回顾基准中，结合拟合模型和回收的钕铁硼（NdFeB）磁体记录，主动学习以比非自适应空间填充更少的实验找到最佳记录结果。富集是相对于进料中稀土与铁的比率。自适应策略通过16到24个井（单个实验）达到了记录的富集最大值，而非自适应策略则需要48个井。我们的两阶段重建在16个井处将两个自适应选择联系在一起。对回收的钐钴（SmCo）磁体的条件分析显示，第二轮在纯度和名义产量之间存在权衡，回收分数是根据假定的起始量计算得出的 - NdFeB第一轮路径在富集方面存在差异。来自石油和天然气开采的生产水的排名依赖于相位和稀释假设，需要确认。我们建议根据其预期的下游贝叶斯风险减少来选择批次：在当前信念下可用过程决策中的最低预期损失。在探索性模拟中，过滤候选者的混合策略的估计损失低于实施的跨路径和条件的联合搜索。涉及合成两阶段策略的差异相对于估计不确定性较小。我们概述了在共享损失和记录标准下的预注册前瞻性测试，要求明确的测量和记录、定义的过程决策和相关输出、可信的经济输入，以及在预期规模下的验证。
cs.AI / 8 / 2609.09418
Valerant: An Automatic Navigable Game Map Generator via Action-Conditioned World Model Exploration
Valerant：一种通过动作条件的世界模型探索实现的自动可导航游戏地图生成器
Abstract
World Action Models (WAMs) couple predictive world modeling with action generation, allowing anticipated future states to guide agent behavior. Although WAMs are rapidly advancing embodied AI, general-purpose counterparts remain largely unexplored in games. Existing game-oriented approaches often combine action-conditioned world models with external policies and reward functions to realize WAM-like decision-making, yet they operate mainly in 2D visual observation space and do not instantiate persistent 3D geometry. Extending this paradigm to 3D games introduces a distinct challenge. In autonomous driving and robotics, the physical environment exists independently of the model, providing a persistent 3D world in which selected actions can be executed. Games have no such external substrate; the virtual world itself must be instantiated. Most playable games require a persistent and navigable space, while 3D games additionally require explicit geometry that supports movement and interaction. Action-conditioned video rollouts provide visual observations but not this spatial representation. We present \textsc{Valerant}, a training-free framework that transforms a pretrained action-conditioned world model into a WAM for exploring and constructing 3D game maps. By coupling predictive visual rollouts with SLAM-based spatial reconstruction and exploration-driven action selection, \textsc{Valerant} progressively transforms a single image into a persistent 3D game map. This framework extends WAM-based interaction beyond 2D visual simulation and offers a new approach to reducing manual effort in 3D game-map creation.
Chinese Translation
世界动作模型（WAMs）将预测性世界建模与动作生成相结合，使得预期的未来状态能够指导智能体行为。尽管WAMs在具身人工智能领域迅速发展，但其通用型对应物在游戏中仍然未得到充分探索。现有的面向游戏的方法通常将动作条件的世界模型与外部策略和奖励函数结合，以实现类似WAM的决策制定，但它们主要在二维视觉观察空间中运作，并未实例化持久的三维几何结构。将这一范式扩展到三维游戏中带来了独特的挑战。在自动驾驶和机器人技术中，物理环境独立于模型存在，提供了一个持久的三维世界，在其中可以执行选定的动作。而游戏并没有这样的外部基底；虚拟世界本身必须被实例化。大多数可玩游戏需要一个持久且可导航的空间，而三维游戏还需要支持移动和交互的显式几何结构。动作条件的视频回放提供了视觉观察，但并不具备这种空间表示。我们提出了 extsc{Valerant}，一个无训练框架，将预训练的动作条件世界模型转化为WAM，以探索和构建三维游戏地图。通过将预测性视觉回放与基于SLAM的空间重建和探索驱动的动作选择相结合， extsc{Valerant}逐步将单幅图像转化为持久的三维游戏地图。该框架将基于WAM的交互扩展到二维视觉模拟之外，并提供了一种减少三维游戏地图创建手动工作的新方法。
cs.AI / 9 / 2609.09428
XAI-Arena: Can LLMs Assess the Quality of XAI Explanations?
XAI-Arena：大型语言模型能否评估可解释人工智能解释的质量？
Abstract
Evaluating the quality of explanations produced by explainable AI (XAI) methods remains challenging because existing approaches often rely on subjective human judgment, limiting reproducibility, scalability, and comparability between studies. We examine whether LLMs can serve as a reproducible and scalable mechanism to make comparative assessments of the quality of XAI explanations. We introduce XAI-Arena, an LLM-as-a-judge framework for scalable, reproducible, multidimensional, and stakeholder-sensitive evaluation of XAI explanation quality. XAI-Arena then allows us to compare XAI explanations along various dimensions, namely, perceived simplicity, clarity, task adequacy, trust calibration, actionability, transparency, faithfulness, and overall interpretability. We then benchmark XAI explanation methods across various datasets, machine learning models, and stakeholder personas. Human validation shows a strong positive association between LLM-generated and human ratings (Spearman's rho=.693, p<.001). Together, LLM-based evaluations can capture systematic differences in XAI explanation quality and provide a scalable and reproducible framework for comparative assessment of XAI explanations.
Chinese Translation
评估可解释人工智能（XAI）方法所产生的解释质量仍然具有挑战性，因为现有方法往往依赖于主观的人类判断，这限制了研究的可重复性、可扩展性和可比性。我们研究了大型语言模型（LLMs）是否可以作为一种可重复且可扩展的机制，对XAI解释的质量进行比较评估。我们提出了XAI-Arena，一个基于LLM的评判框架，用于对XAI解释质量进行可扩展、可重复、多维度和利益相关者敏感的评估。XAI-Arena使我们能够在多个维度上比较XAI解释，即感知的简单性、清晰性、任务适应性、信任校准、可操作性、透明度、忠实性和整体可解释性。接着，我们在多个数据集、机器学习模型和利益相关者角色上基准测试XAI解释方法。人类验证显示LLM生成的评分与人类评分之间存在强正相关（Spearman's rho=.693, p<.001）。总之，基于LLM的评估能够捕捉XAI解释质量的系统性差异，并提供一个可扩展和可重复的框架，用于对XAI解释的比较评估。
cs.AI / 10 / 2609.09448
Do Agents Know When They Succeed? Calibrating Agent Confidence from Internal Representations
代理是否知道何时成功？从内部表征中校准代理信心
Abstract
As agentic systems getting adopted rapidly in safety critical applications, it is vital to measure the confidence associated with the agentic actions. In comparison to the traditional machine learning systems, agentic workflows have complex failure modes with planning, tool invocation and dynamic environment interactions. In this paper, we investigate whether model's internal representations provide stronger signals of eventual task success in multi-turn agentic setups. We introduce two complementary methods: Latent Trajectory Dynamics (LTD), which summarizes changes in residual-stream representations across an an interaction trajectory, and the Action Representation Probe (ARP), which predicts success from representations formed at action decisions. Across three interactive benchmarks (Bash, SQL, Python) and three model families (Qwen14B, Qwen7B, DeepSeek6.7B), our methods consistently outperform surface level generation and sequence-based calibration baselines providing a zero-overhead reliability monitor that requires neither prompt alterations nor multi-sample rollouts.
Chinese Translation
随着代理系统在安全关键应用中的快速采用，测量与代理行为相关的信心变得至关重要。与传统的机器学习系统相比，代理工作流程具有复杂的失败模式，包括规划、工具调用和动态环境交互。本文探讨了模型的内部表征是否在多轮代理设置中提供了更强的最终任务成功信号。我们提出了两种互补的方法：潜在轨迹动态（Latent Trajectory Dynamics, LTD），该方法总结了交互轨迹中残差流表征的变化，以及动作表征探测器（Action Representation Probe, ARP），该方法根据在动作决策时形成的表征预测成功。在三个交互基准（Bash、SQL、Python）和三个模型系列（Qwen14B、Qwen7B、DeepSeek6.7B）中，我们的方法始终优于表面生成和基于序列的校准基线，提供了一种零开销的可靠性监测器，无需提示修改或多样本回滚。
cs.AI / 11 / 2609.09458
ContractEval: Query-Conditioned Execution Matching for Procedural Instruction Conformance
ContractEval：基于查询条件的执行匹配用于程序指令一致性
Abstract
As LLM agents move from answering questions to carrying out procedures, failures can be unwarranted rather than visibly wrong: the final response looks acceptable even though the system skipped the check, branch, dependency, or invariant that made the answer justified. Output-only evaluation sees the answer, and trace-aware judging sees activity, but neither identifies which obligations were active for the query. We introduce CONTRACTEVAL, a diagnostic framework for making those active obligations explicit. It represents procedural instructions as query-active obligations and matches them against response or trace evidence, turning omissions, wrong branches, ordering errors, extra actions, invariant breaches, and output-contract violations into distinct conformance failures. On a controlled suite of audited procedural contracts, output-only and trace-aware LLM judges miss many injected structural failures; under gold expected and observed graphs, ContractEval detects and localizes all of them. LLM-backed extraction preserves much of this signal but remains calibration-sensitive. ContractEval is therefore not a compliance guarantee; it makes procedural conformance auditable rather than implicit in final-answer quality.
Chinese Translation
随着大型语言模型（LLM）代理从回答问题转向执行程序，失败可能是无故的，而不是明显错误的：最终响应看起来是可接受的，即便系统跳过了使答案合理的检查、分支、依赖关系或不变式。仅输出评估只看到答案，而跟踪感知评估则看到活动，但两者都无法识别哪些义务在查询中是活跃的。我们提出了CONTRACTEVAL，一个使这些活跃义务显式化的诊断框架。它将程序指令表示为查询活跃义务，并将其与响应或跟踪证据进行匹配，将遗漏、错误分支、顺序错误、额外操作、不变式违反和输出合同违规转化为明显的一致性失败。在一组经过审计的程序合同的受控测试中，仅输出和跟踪感知的LLM评估者错过了许多注入的结构性失败；在黄金预期和观察图下，ContractEval能够检测并定位所有这些失败。基于LLM的提取保留了大量信号，但仍然对校准敏感。因此，ContractEval并不是合规保证；它使程序一致性成为可审计的，而不是隐含在最终答案质量中。
cs.AI / 12 / 2609.09565
Multi-Agent Agentic Graph Learning via Structural Signatures
基于结构特征的多智能体图学习
Abstract
Agentic graph learning (AGL) has recently achieved promising results on graph reasoning tasks, where an agent powered by a large language model (LLM) sequentially samples the graph as evidence to support its final prediction. Existing methods either employ a single agent or orchestrate multiple role-based agents to reason and learn over the entire graph, but both essentially rely on a shared reasoning policy across different graph regions, which can be suboptimal for graphs with heterogeneous structural and semantic patterns. Inspired by the progress of multi-agent collaboration on complex reasoning tasks, a natural remedy is to let multiple agents own different memory and collaborate; however, applying this paradigm to graphs directly faces two challenges. First, existing AGL methods typically verbalize graph structures into natural-language descriptions for LLM agents, making the reasoning process sensitive to the ordering of structural information and thereby breaking the permutation-invariant nature of graphs. Second, incorporating increasingly large sampled neighborhoods leads to rapidly growing contexts. To address these challenges, this paper introduces a multi-agent agentic graph learning (i.e., MAAGL) framework. MAAGL partitions the graph into communities and assigns an independent agent to each community for region-specific specialization. MAAGL represents structural and semantic evidence separately. Structural evidence is summarized by a dynamically updated structural signature that is permutation-invariant and fixed in size, while semantic evidence is filtered to the top-k nodes ranked by relevance. Based on historical trajectories with similar signatures, agents estimate their confidence and trigger debate-style collaboration when needed. Extensive experiments on four benchmark datasets show that MAAGL outperforms SOTA AGL methods.
Chinese Translation
智能体图学习（AGL）最近在图推理任务中取得了令人鼓舞的成果，其中一个由大型语言模型（LLM）驱动的智能体顺序地对图进行采样，以支持其最终预测。现有方法要么采用单一智能体，要么协调多个基于角色的智能体对整个图进行推理和学习，但这两种方法本质上都依赖于跨不同图区域的共享推理策略，这对于具有异构结构和语义模式的图来说可能并不理想。受到多智能体在复杂推理任务中合作进展的启发，一个自然的解决方案是让多个智能体拥有不同的记忆并进行协作；然而，将这一范式直接应用于图面临两个挑战。首先，现有的AGL方法通常将图结构转化为自然语言描述供LLM智能体使用，使得推理过程对结构信息的顺序敏感，从而破坏了图的置换不变性。其次，纳入越来越大的采样邻域会导致上下文迅速增长。为了解决这些挑战，本文提出了一种多智能体智能图学习框架（即MAAGL）。MAAGL将图划分为社区，并为每个社区分配一个独立的智能体以实现区域特定的专业化。MAAGL分别表示结构和语义证据。结构证据通过动态更新的结构特征进行总结，该特征具有置换不变性且大小固定，而语义证据则过滤出按相关性排名的前k个节点。基于具有相似特征的历史轨迹，智能体估计其信心，并在需要时触发辩论式协作。在四个基准数据集上的大量实验表明，MAAGL优于现有的AGL方法。
cs.AI / 13 / 2609.09578
CityPlanner: A Sandbox Agent for Executable Urban Planning
城市规划者：可执行城市规划的沙盒代理
Abstract
Urban planning is a real-world spatial optimization problem that requires selecting feasible actions from large candidate spaces under practical objectives such as cost and service quality. Existing optimization and reinforcement learning methods are effective for fixed formulations, but often depend on task-specific representations and constraint handling. We propose \emph{CityPlanner}, a sandbox-agent framework for executable urban planning. CityPlanner introduces \emph{UrbanSandbox}, a unified file-based environment where agents inspect task files, generate plans, run evaluators, and revise decisions based on executable feedback. To make learning tractable, we further propose atomic-task reinforcement learning, which decomposes long sandbox trajectories into \emph{BuildPlan} for initial construction and \emph{ImprovePlan} for feedback-based refinement. Experiments on a real-world benchmark show that CityPlanner consistently outperforms heuristic, task-specific RL, and general LLM-agent baselines. Ablations verify the contributions of UrbanSandbox, atomic-task RL, and iterative deployment. We release the code and dataset at https://anonymous.4open.science/r/co-agent-C1C8
Chinese Translation
城市规划是一个现实世界的空间优化问题，需要在成本和服务质量等实际目标下，从大量候选空间中选择可行的行动。现有的优化和强化学习方法在固定的公式化问题上有效，但往往依赖于特定任务的表示和约束处理。我们提出了 extit{城市规划者}（CityPlanner），一个用于可执行城市规划的沙盒代理框架。城市规划者引入了 extit{城市沙盒}（UrbanSandbox），这是一个统一的基于文件的环境，代理可以在其中检查任务文件、生成计划、运行评估器，并根据可执行反馈修正决策。为了使学习变得可行，我们进一步提出了原子任务强化学习（atomic-task reinforcement learning），它将长的沙盒轨迹分解为用于初始构建的 extit{构建计划}（BuildPlan）和用于基于反馈的改进的 extit{改进计划}（ImprovePlan）。在一个真实世界基准上的实验表明，城市规划者始终优于启发式、特定任务的强化学习和通用大语言模型代理基线。消融实验验证了城市沙盒、原子任务强化学习和迭代部署的贡献。我们在 https://anonymous.4open.science/r/co-agent-C1C8 发布了代码和数据集。
cs.AI / 14 / 2609.09589
A Function-Space Approach to the Statistical Mechanics of Learning Dynamics
一种基于函数空间的学习动态统计力学方法
Abstract
Deep neural networks exhibit regular macroscopic behavior despite highly nonlinear dynamics in vast parameter spaces. We develop a statistical-mechanical description of learning directly in function space, treating parameter configurations as microscopic realizations and functions with their dynamical operators as macroscopic variables. For mean-squared loss, the exact error dynamics are governed by the learning operator \(M=JJ^\ast\). Combining the dynamical Boltzmann weight of the conditional stochastic dynamics with the parameter-space density of states, whose local curvature defines a statistical operator \(B\), and integrating over local fluctuations yields $$ \Phi_{\mathrm{fluc}}(M;B)=\frac{\sigma_\xi^2}{2}\log\det(M^{-1}+B)+\mathrm{const}. $$ At fixed spectrum, this term is rotationally stationary when \([M,B]=0\), is minimized by pairing large eigenvalues of \(M\) with small eigenvalues of \(B\), and generates a local restoring contribution against rotational mismatch. For ReLU-type function spaces under mild stable statistical conditions, \(B=\sigma_\xi^2L^\ast\mathcal K L\), where \(L\) measures coarse-grained second-order structure. Thus the low-\(B\) sector corresponds, up to bounded anisotropy of \(\mathcal K\), to low structural curvature, implying a preference for faster relaxation along smooth, data-adaptive directions. These results identify function space as a natural macroscopic level for studying stable collective organization in learning.
Chinese Translation
深度神经网络在广阔的参数空间中尽管具有高度非线性的动态行为，却展现出规律的宏观行为。我们在函数空间中直接发展了一种学习的统计力学描述，将参数配置视为微观实现，将具有其动态算子的函数视为宏观变量。对于均方损失，精确的误差动态由学习算子（learning operator）$M=JJ^ullet$控制。将条件随机动态的动态玻尔兹曼权重与参数空间的态密度结合，其中局部曲率定义了统计算子（statistical operator）$B$，并对局部波动进行积分，得到$$ ext{Φ}_{ ext{fluc}}(M;B)= rac{ ext{σ}_ξ^2}{2} ext{log} ext{det}(M^{-1}+B)+ ext{const}. $$ 在固定谱下，当$[M,B]=0$时，该项是旋转不变的，通过将$M$的大特征值与$B$的小特征值配对来最小化，并产生对抗旋转不匹配的局部恢复贡献。对于在温和稳定统计条件下的ReLU类型函数空间，$B= ext{σ}_ξ^2L^ullet ext{K}L$，其中$L$测量粗粒化的二阶结构。因此，低-$B$区间对应于$ ext{K}$的有界各向异性，表明在平滑、数据自适应方向上更快的松弛偏好。这些结果将函数空间识别为研究学习中稳定集体组织的自然宏观层次。
cs.AI / 15 / 2609.09625
From State Synchronization to Cognitive Self-Evolution: An Operational Architecture for Cognitive Digital Twins
从状态同步到认知自我演化：认知数字双胞胎的操作架构
Abstract
As Digital Twin (DT) systems evolve beyond state synchronization toward task-oriented and knowledge-driven operation, Cognitive Digital Twins (CDTs) have emerged as an extension that incorporates cognitive capabilities into twin operation. Existing CDT studies often focus on specific enabling techniques, such as learning modules, knowledge graphs, and large language models, while providing limited insight into how cognition can be systematically integrated into DT architectures. To address this issue, this paper proposes a four-layer CDT architecture consisting of the physical layer, digital-twin layer, cognitive layer, and task layer. The proposed architecture establishes a self-evolving closed operational loop spanning these four layers, in which physical states are synchronized into digital representations, cognition constructs task-specific cognitive models through knowledge, memory, and attention, and task-level decisions are generated under practical constraints. Operational feedback further refines cognitive experience and updates relationships and annotations in the digital representation, enabling subsequent task interpretation, initiation, and reasoning to evolve with system operation. Based on this framework, two representative operation modes are characterized: user-request-driven cognition and self-driven cognition. We further discuss key enabling mechanisms and deployment challenges associated with semantic communication, knowledge querying, task orchestration, and closed-loop synchronization. A lightweight simulation study illustrates reliable closed-loop task feasibility under limited semantic information and improved operational efficiency through accumulated task experience. The proposed framework provides a structured foundation for the design and development of future CDT systems.
Chinese Translation
随着数字双胞胎（Digital Twin, DT）系统的发展，超越了状态同步，朝着任务导向和知识驱动的操作演进，认知数字双胞胎（Cognitive Digital Twins, CDT）作为一种扩展，融入了认知能力于双胞胎操作中。现有的CDT研究通常集中于特定的支持技术，如学习模块、知识图谱和大型语言模型，而对如何系统性地将认知整合到DT架构中提供的见解有限。为了解决这一问题，本文提出了一种由物理层、数字双胞胎层、认知层和任务层组成的四层CDT架构。该架构建立了一个跨越这四个层次的自我演化闭环操作，其中物理状态被同步为数字表示，认知通过知识、记忆和注意力构建任务特定的认知模型，并在实际约束下生成任务级决策。操作反馈进一步细化认知经验，并更新数字表示中的关系和注释，使得后续的任务解释、启动和推理能够随着系统操作而演变。基于这一框架，特征化了两种代表性的操作模式：用户请求驱动的认知和自驱动的认知。我们进一步讨论了与语义通信、知识查询、任务协调和闭环同步相关的关键支持机制和部署挑战。一项轻量级的仿真研究展示了在有限语义信息下可靠的闭环任务可行性，并通过积累的任务经验提高了操作效率。所提出的框架为未来CDT系统的设计和开发提供了结构化基础。
cs.AI / 16 / 2609.09627
Seven Sources of Physical AI Capability Formation
物理人工智能能力形成的七个来源
Abstract
Capabilities relevant to Physical AI can arise from materially different formation histories, yet existing taxonomies organized by morphology, architecture, learning algorithm, task, or domain do not directly answer what gives rise to a capability. We define a capability-formation source as a factor materially contributing to capability formation, distinct from components or construction steps. We identify seven non-exclusive sources: Recorded-Experience (RE), Predictive-Modeling (PM), Evaluative-Interaction (EI), Surrogate-Environment (SE), Mechanism-Grounded (MG), Embodied-Coupling (EC), and Evolution-Driven (ED) Formation. Using reconstructive induction with theoretical saturation, we traced a research matrix to primary studies, deduplicated the literature, set coding rules, and conducted three rounds of maximum-difference and negative-case sampling. Challenges included curriculum and self-supervised learning, active inference, open-ended and developmental learning, planning and search, neuro-symbolic architectures, digital twins, generative physical world models, and morphology-control co-design. Within the scope and criteria fixed as of September 4, 2026, all 49 evidence records were explainable by the seven sources individually or in combination. No R1-R3 challenge produced an irreducible eighth source, and R3 required no new core definition or substantive boundary rule. We therefore claim theoretical saturation within the stated scope, not logical completeness or exhaustive future coverage. The framework distinguishes similarity in observed capability from similarity in how it was formed, supporting analysis of explanation, transfer, replication, dependencies, governance evidence, and geoeconomic foundations.
Chinese Translation
与物理人工智能相关的能力可以源于截然不同的形成历史，但现有的按形态、架构、学习算法、任务或领域组织的分类法并未直接回答能力的形成原因。我们将能力形成来源定义为对能力形成有实质性贡献的因素，区别于组件或构建步骤。我们识别出七个非排他性的来源：记录经验（Recorded-Experience, RE）、预测建模（Predictive-Modeling, PM）、评估互动（Evaluative-Interaction, EI）、替代环境（Surrogate-Environment, SE）、机制基础（Mechanism-Grounded, MG）、体现耦合（Embodied-Coupling, EC）和进化驱动（Evolution-Driven, ED）形成。通过重建归纳法与理论饱和，我们追踪了一个研究矩阵到主要研究，去重文献，设定编码规则，并进行了三轮最大差异和负案例抽样。挑战包括课程与自监督学习、主动推理、开放式与发展学习、规划与搜索、神经符号架构、数字双胞胎、生成物理世界模型以及形态控制共同设计。在截至2026年9月4日的范围和标准内，所有49个证据记录均可通过这七个来源单独或组合进行解释。没有任何R1-R3挑战产生不可简化的第八个来源，R3也不需要新的核心定义或实质性边界规则。因此，我们在所述范围内声称理论饱和，而非逻辑完整或详尽的未来覆盖。该框架区分了观察到的能力的相似性与其形成方式的相似性，支持对解释、转移、复制、依赖性、治理证据和地缘经济基础的分析。
cs.AI / 17 / 2609.09646
RobustSGPO: Search-Space Control for Agent Harness Evolution
RobustSGPO：代理工具演化的搜索空间控制
Abstract
Semantic-gradient-based prompt optimization (SGPO) improves agent harnesses using execution feedback, but its local update rule leaves the choice of edit scope and operation unresolved. We introduce RobustSGPO, which specifies the requested edit, constructs and checks the patch, and continues search from either the incumbent or retained snapshots. We evaluate permission scheduling, cumulative controls, and task-family transfer in the AgentX brainstorming workflow using 120 tasks, 95 runs, and 7,350 candidate attempts. Periodic $1\to2\to3$ scheduling exceeds fixed maximum permission by 0.28 test-score points. RobustSGPO increases completion on 30 held-out tasks from 60.0% to 80.0% and improves test quality from 3.77 to 4.14 under a 20-million-token budget. Category retention reduces source-task degradation after a shift, whereas random retention reaches a higher destination endpoint. Search-space control benefits quality through executable edits and alternative starting points, with measurable retention overhead.
Chinese Translation
基于语义梯度的提示优化（SGPO）通过执行反馈来改善代理工具，但其局部更新规则未能解决编辑范围和操作的选择问题。我们提出了RobustSGPO，它明确指定所需的编辑，构建并检查补丁，并从现有快照或保留快照继续搜索。我们在AgentX头脑风暴工作流程中使用120个任务、95次运行和7,350个候选尝试评估权限调度、累积控制和任务家族转移。周期性的$1 o2 o3$调度超出了固定最大权限0.28测试分数点。RobustSGPO将30个保留任务的完成率从60.0%提高到80.0%，并在2000万标记预算下将测试质量从3.77提高到4.14。类别保留在转移后减少了源任务的退化，而随机保留则达到了更高的目标端点。搜索空间控制通过可执行编辑和替代起始点提升了质量，并具有可测量的保留开销。
cs.AI / 18 / 2609.09647
Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
代理人工智能的黑箱红队测试：基于分类法的自动化风险发现框架
Abstract
Agentic systems are rapidly moving to production, where they read untrusted inputs, call tools with real permissions, and act autonomously, expanding the security surface beyond chat-only models. Yet standard evaluations remain single-turn and fail to capture multi-step agent vulnerabilities. We present a systematic black-box framework for risk-aware agent evaluation requiring only basic system descriptions. Our approach introduces: (1) a seven-domain taxonomy mapping observable behaviors to risk categories, (2) fully automated SAGE-RT red teaming producing 120 adversarial scenarios per domain, and (3) human-validated evaluation using LLM judges. Empirical validation across two agent architectures (CrewAI and AutoGen) with four base models reveals alarming patterns: 56.25\% average governance risk, 65\% privacy risk in multi-agent configurations, and agent behavior vulnerabilities reaching 85\%. Our black-box approach effectively identifies critical architectural vulnerabilities without privileged access, providing a scalable path toward safer agent deployments.
Chinese Translation
代理系统正在迅速投入生产，它们读取不可信的输入，调用具有实际权限的工具，并自主行动，扩展了安全面超出仅限聊天的模型。然而，标准评估仍然是单轮的，未能捕捉多步骤代理的脆弱性。我们提出了一个系统的黑箱框架，用于风险意识的代理评估，仅需基本的系统描述。我们的方法引入了：（1）一个七领域分类法，将可观察行为映射到风险类别；（2）完全自动化的SAGE-RT红队测试，每个领域生成120个对抗场景；（3）使用大型语言模型（LLM）评审的人工验证评估。对两种代理架构（CrewAI和AutoGen）及四个基础模型的实证验证揭示了令人担忧的模式：平均治理风险为56.25%，多代理配置中的隐私风险为65%，代理行为脆弱性达到85%。我们的黑箱方法有效识别关键的架构脆弱性，无需特权访问，为更安全的代理部署提供了可扩展的路径。
cs.AI / 19 / 2609.09657
RESCUE-BENCH: Towards Relation-Aware Multi-Party Emotional Support Conversation Systems
RESCUE-BENCH：面向关系感知的多方情感支持对话系统
Abstract
Existing emotional support conversation systems mainly focus on one-on-one seeker-supporter interactions and individual emotional states, leaving interpersonal relations in multi-party scenarios underexplored. In this work, we introduce relation-aware emotional support conversation, a new task that evaluates whether LLMs can capture and utilize the evolving dynamics of relationships to offer more effective emotional support. We construct RESCUE (Relation-aware Emotional Support Conversation Understanding and Evaluation Benchmark) from real couple and family interview conversations, containing 191 samples, 7,079 annotated turns, and 1,064.8 minutes of video. Based on rich annotations of socio-emotional and support-related dynamics, RESCUE defines six tasks that evaluate two core capabilities required for relation-aware emotional support: Relational Understanding and Relation-Sensitive Support. Experiments with ten LLMs show that current models perform relatively well on tasks relying on local emotional or intervention cues, but struggle with relation-intensive tasks such as relation pattern prediction, viewpoint prediction, and support strategy prediction. These findings reveal the limitations of current LLMs in modeling interpersonal relations and making relation-sensitive support decisions.
Chinese Translation
现有的情感支持对话系统主要集中于一对一的求助者与支持者之间的互动以及个体情感状态，导致多方场景中的人际关系尚未得到充分探索。在本研究中，我们引入了关系感知的情感支持对话，这是一项新的任务，旨在评估大型语言模型（LLMs）是否能够捕捉和利用关系的动态变化，从而提供更有效的情感支持。我们从真实的情侣和家庭访谈对话中构建了RESCUE（关系感知情感支持对话理解与评估基准），包含191个样本、7,079个标注回合和1,064.8分钟的视频。基于丰富的社会情感和支持相关动态的标注，RESCUE定义了六项任务，以评估关系感知情感支持所需的两个核心能力：关系理解和关系敏感支持。对十个LLMs的实验表明，当前模型在依赖局部情感或干预线索的任务上表现相对良好，但在关系密集型任务（如关系模式预测、观点预测和支持策略预测）上表现不佳。这些发现揭示了当前LLMs在建模人际关系和做出关系敏感支持决策方面的局限性。
cs.AI / 20 / 2609.09664
PRAGMA: Evaluating Personalized Guidance with Memory Alignment in Lifelong Conversations
PRAGMA：在终身对话中评估个性化指导与记忆对齐
Abstract
Large language models (LLMs) are increasingly deployed as personalized assistants that interact with users over extended periods of time. As conversations grow longer, relying on full interaction histories becomes increasingly inefficient and unreliable: long contexts introduce substantial computational overhead, making it difficult for models to consistently identify and utilize the most relevant information for the current request. These challenges have motivated memory systems that structure and retrieve user-specific information. In realistic interactions, users often seek practical guidance such as recommendations, planning, and decision support. Unlike factual recall tasks, personalized guidance requires models to integrate information across multiple past conversations and reason about changing user preferences and experiences. However, existing conversational memory evaluations mainly focus on retrieval and factual recall. To study this challenge, we introduce PRAGMA, a benchmark for evaluating personalized guidance in long-term conversations. PRGAMA contains curated longitudinal conversation histories, evidence annotations, and guidance scenarios grounded in evolving user contexts and incorrect user assumptions. Experiments across retrieval systems, memory systems, and long-context models reveal that current systems struggle both to recover the appropriate conversational evidence and to effectively use it for personalized guidance. Our results highlight the need for memory architectures that support robust conversational retrieval and memory-grounded reasoning beyond evidence recall.
Chinese Translation
大型语言模型（LLMs）越来越多地被部署为个性化助手，与用户进行长期互动。随着对话时间的延长，依赖完整的互动历史变得越来越低效和不可靠：长上下文引入了巨大的计算开销，使得模型难以始终如一地识别和利用当前请求的最相关信息。这些挑战促使了记忆系统的出现，以结构化和检索用户特定的信息。在现实的互动中，用户通常寻求实用的指导，例如推荐、规划和决策支持。与事实回忆任务不同，个性化指导要求模型整合多个过去对话中的信息，并推理用户偏好和经验的变化。然而，现有的对话记忆评估主要集中在检索和事实回忆上。为研究这一挑战，我们引入了PRAGMA，一个用于评估长期对话中个性化指导的基准。PRAGMA包含经过策划的纵向对话历史、证据注释以及基于不断变化的用户背景和错误用户假设的指导场景。在检索系统、记忆系统和长上下文模型的实验中，结果表明当前系统在恢复适当的对话证据和有效利用这些证据进行个性化指导方面均面临困难。我们的结果突显了需要支持稳健对话检索和基于记忆的推理的记忆架构，超越单纯的证据回忆。
cs.AI / 21 / 2609.09678
Safe to Stop? Risk-Constrained Stopping for Sequential Clinical Diagnosis Agents
安全停止？针对顺序临床诊断代理的风险约束停止
Abstract
Clinical diagnosis agents must decide not only what test to request next, but also when to diagnose or defer. Existing agent benchmarks largely evaluate accuracy after fixed or unconstrained interaction, leaving autonomous stopping reliability implicit. We present Cros, a risk-constrained stopping layer combining state-wise error ranking, policy design on disjoint development splits, and LTT-style exact tests of selective diagnostic error and minimum autonomous coverage for complete sequential policies. Its finite-sample guarantee requires the candidate family, testing rule, and any randomization to be frozen before calibration labels are accessed. On a 1,834-episode MIMIC-derived abdominal-pain benchmark, the full ranker achieves exploratory state-error AUROC 0.853, compared with 0.715 for maximum class probability and 0.552 for the backbone's native stop score. On the previously viewed 367-episode evaluation split, analytically averaging over the frozen Cros weights yields 16.9% selective error at 78.8% coverage, cost 5.57, and 0.68 tests, versus 30.8% error at 100% coverage, cost 8.14, and 1.53 tests under native stopping. Forced continuation is non-monotone: error is 28.3% with HPI alone and 34.3% after full workup. However, the uniform-weight mixture ablation is cheaper on this viewed split despite missing the locked development margins, and Cros nominally satisfies the joint criterion in only 6 of 20 development resplits. Because evaluation labels were inspected during earlier development, these findings provide exploratory feasibility and audit evidence, not a confirmatory safety certificate.
Chinese Translation
临床诊断代理不仅需要决定下一步请求什么测试，还需要决定何时进行诊断或推迟。现有的代理基准大多在固定或不受限制的交互后评估准确性，未明确自主停止的可靠性。我们提出了Cros，这是一种风险约束的停止层，结合了状态误差排名、在不相交的开发分割上进行的策略设计，以及LTT风格的选择性诊断误差和完整顺序策略的最小自主覆盖的精确测试。其有限样本保证要求候选家族、测试规则以及任何随机化在访问校准标签之前被冻结。在一个包含1,834个回合的MIMIC衍生腹痛基准中，完整排名器实现了探索性状态误差的AUROC为0.853，而最大类别概率为0.715，主干的原生停止分数为0.552。在之前查看的367个回合评估分割中，分析性地对冻结的Cros权重进行平均，得到选择性误差为16.9%，覆盖率为78.8%，成本为5.57，测试次数为0.68，而在原生停止下，误差为30.8%，覆盖率为100%，成本为8.14，测试次数为1.53。强制继续是非单调的：仅使用HPI时误差为28.3%，而在全面检查后为34.3%。然而，尽管缺少锁定的开发边际，均匀权重混合消融在这个查看的分割上成本更低，并且Cros在20个开发重新分割中仅在6个中名义上满足联合标准。由于在早期开发过程中检查了评估标签，这些发现提供了探索性的可行性和审计证据，而不是确认的安全证书。
cs.AI / 22 / 2609.09702
Decision Shifts, Lost Label Functionality, and an Inconclusive Grounding Audit in Correctness-Gated Multi-Teacher Distillation
决策转变、标签功能丧失与正确性门控多教师蒸馏中的不确定基础审计
Abstract
Candidate decision correctness and rationale grounding are different objectives. We examine correctness-gated multi-teacher distillation in a fixed experiment. Eight arms share 4,330 sources, a 63.9M-parameter student, 12,990 optimization rows, 406 updates, evidence inputs, and a decoder; seven teacher-based arms use one fixed three-response pool. Three seeds are evaluated on 267 held-out examples. Relative to unfiltered distillation, the correctness-weighted arm differed in accuracy by +0.1660 (95% observed-matrix interval [0.0670, 0.2455]), five-label macro-F1 by +0.1323 ([0.0916, 0.1731]), and task-defined conditional unsafe-action rate by -0.4979 ([-0.5926, -0.3686]). These shifts do not imply uniformly better behavior. Source-label SFT had the highest mean macro-F1 (0.586). The weighted arm had zero Refuted recall in every seed, and two seeds assigned NotEnoughInfo to all 167 claim examples. In an availability-amended audit at one reference seed, weighted and unfiltered outputs had 0/20 versus 1/20 evidence-supported positives and 20/20 versus 19/20 positives containing unsupported material. Samples were non-paired, source overlap was not serialized, and the amendment followed automatic summarization but preceded annotation. The audit therefore cannot estimate a common-source grounding effect and is inconclusive about system-level improvement or harm. Hard filtering already achieved 0.660 accuracy, 0.530 macro-F1, and 0.135 conditional unsafe rate. The implemented weighted arm showed no demonstrated incremental decision benefit over hard filtering. This fixed-matrix failure analysis shows decision redistribution with lost label functionality; the available human audit does not establish a grounding gain.
Chinese Translation
候选决策的正确性和推理基础是不同的目标。我们在固定实验中考察了正确性门控的多教师蒸馏。八个实验组共享4,330个来源，一个63.9M参数的学生模型，12,990个优化行，406次更新，证据输入和一个解码器；七个基于教师的实验组使用一个固定的三响应池。三个种子在267个保留示例上进行评估。与未过滤的蒸馏相比，正确性加权的实验组在准确性上提高了+0.1660（95%观察矩阵区间[0.0670, 0.2455]），五标签宏F1提高了+0.1323（[0.0916, 0.1731]），任务定义的条件不安全行动率降低了-0.4979（[-0.5926, -0.3686]）。这些变化并不意味着行为普遍更好。源标签的SFT具有最高的平均宏F1（0.586）。加权实验组在每个种子中都没有显示出任何被驳回的召回率，且两个种子将所有167个声明示例标记为信息不足。在一个参考种子的可用性修正审计中，加权和未过滤的输出分别有0/20和1/20的证据支持的正例，以及20/20和19/20的包含不支持材料的正例。样本是非配对的，来源重叠没有序列化，修正是在自动摘要之后但在注释之前进行的。因此，该审计无法估计共同来源的基础效应，并且对系统级的改进或损害没有结论。严格过滤已经达到了0.660的准确率，0.530的宏F1和0.135的条件不安全率。实施的加权实验组未显示出相对于严格过滤的增量决策收益。这一固定矩阵的失败分析显示了决策的重新分配与标签功能的丧失；可用的人类审计未能建立基础收益。
cs.AI / 23 / 2609.09707
Which Tokens Should SFT Actually Learn? A Token-Trimming Perspective on Mathematical Reasoning
SFT 实际上应该学习哪些标记？一种基于标记修剪的数学推理视角
Abstract
Supervised fine-tuning (SFT) applies a uniform cross-entropy loss to all target tokens, even though different tokens provide unequal learning signals for mathematical reasoning. This uniform treatment can over-sharpen already mastered tokens while amplifying learning pressure on uncertain, low-confidence tokens, leading to suboptimal training dynamics. We propose Trimmed Logit-Gap SFT (TrimSFT), a simple token-level reweighting method that scales the SFT loss according to the logit gap between the gold token and its strongest competitor. TrimSFT trims supervision away from both extremes: tokens already mastered (large logit gap) and tokens weakly supported by the current model (small or negative logit gap), concentrating learning within an intermediate logit-gap region between them. We instantiate this principle with a Gaussian weight centered at margin m with bandwidth {\tau}, requiring no reference model or additional forward pass. We evaluate TrimSFT on six base models from the Llama, Qwen, and DeepMath families across five mathematical reasoning benchmarks. TrimSFT consistently improves over standard SFT, achieving the best average performance on five out of six models, with gains of up to +26.9 points over SFT on MATH500. Further analyses show that the bandwidth {\tau} matters more than the exact margin location, and that half-trim variants that remove supervision pressure from only one side yield inferior trade-offs. A token-level logit-gap distribution analysis suggests that TrimSFT reshapes model confidence in a more balanced way than uniform SFT or monotonic reweighting methods. These results suggest that reasoning SFT can benefit from trimming both extremes rather than treating all tokens uniformly.
Chinese Translation
监督微调（SFT）对所有目标标记应用统一的交叉熵损失，尽管不同标记为数学推理提供的不平等学习信号。这种统一处理可能会过度强化已经掌握的标记，同时加大对不确定、低信心标记的学习压力，导致次优的训练动态。我们提出了修剪逻辑差（Trimmed Logit-Gap）SFT（TrimSFT），这是一种简单的标记级重加权方法，根据金标记与其最强竞争者之间的逻辑差缩放 SFT 损失。TrimSFT 从两个极端修剪监督：已经掌握的标记（大逻辑差）和当前模型支持较弱的标记（小或负逻辑差），将学习集中在它们之间的中间逻辑差区域。我们用以边际 m 为中心、带宽为 { au} 的高斯权重来实例化这一原则，无需参考模型或额外的前向传递。我们在来自 Llama、Qwen 和 DeepMath 家族的六个基础模型上评估了 TrimSFT，涵盖五个数学推理基准。TrimSFT 在标准 SFT 上始终表现出改善，在六个模型中的五个上实现了最佳平均性能，在 MATH500 上相较于 SFT 提升了多达 +26.9 分。进一步分析表明，带宽 { au} 比确切的边际位置更为重要，而仅从一侧去除监督压力的半修剪变体则产生了较差的权衡。标记级逻辑差分布分析表明，TrimSFT 以比统一 SFT 或单调重加权方法更为平衡的方式重塑了模型信心。这些结果表明，推理 SFT 可以通过修剪两个极端而不是对所有标记进行统一处理来获益。
cs.AI / 24 / 2609.09735
Can Artificial Intelligence Support Healthcare and Mental Health Through Early Cyberbullying Detection ? The Impact of Emotion-Aware AI on Proactive Online Safety
人工智能能否通过早期网络欺凌检测支持医疗保健和心理健康？情感感知人工智能对主动在线安全的影响
Abstract
Healthcare systems, mental health, and public well-being are increasingly affected by cyberbullying and harmful online interactions. This paper presents CareGuard, an early-warning framework designed to support healthcare-driven mental health protection and proactive online safety through the detection of cyberbullying-related content using advanced natural language processing techniques. CareGuard integrates zero-shot semantic labeling with fine-tuned transformer-based models, including BERT, DistilBERT, and RoBERTa, to enable robust and context-aware classification across sensitive cyberbullying categories. To improve efficiency and reduce unnecessary computation in healthcare-oriented monitoring settings, the framework incorporates an emotion-aware filtering mechanism alongside cosine similarity-based semantic screening, allowing the system to focus on semantically relevant and emotionally salient content. Experimental results on benchmark datasets demonstrate that CareGuard effectively balances detection accuracy and computational efficiency, highlighting its potential for scalable deployment in healthcare systems, mental health monitoring, and online safety applications.
Chinese Translation
医疗保健系统、心理健康和公共福祉正日益受到网络欺凌和有害在线互动的影响。本文提出了CareGuard，一个早期预警框架，旨在通过使用先进的自然语言处理技术检测与网络欺凌相关的内容，从而支持以医疗为驱动的心理健康保护和主动在线安全。CareGuard将零样本语义标注与经过微调的基于变换器的模型（包括BERT、DistilBERT和RoBERTa）相结合，以实现对敏感网络欺凌类别的强大且上下文感知的分类。为了提高效率并减少医疗监测环境中的不必要计算，该框架结合了情感感知过滤机制和基于余弦相似度的语义筛选，使系统能够专注于语义相关且情感突出的内容。在基准数据集上的实验结果表明，CareGuard有效地平衡了检测准确性和计算效率，突显了其在医疗系统、心理健康监测和在线安全应用中的可扩展部署潜力。
cs.AI / 25 / 2609.09754
LexAgentHallu: A Hierarchical Benchmark for Profiling Hallucinations in Legal Agents
LexAgentHallu：用于评估法律代理人幻觉的分层基准
Abstract
As large language models are increasingly deployed as tool-augmented legal agents, they introduce agentic hallucinations where tool-call and reasoning errors cascade into fabricated holdings and miscited authority. However, existing legal benchmarks evaluate only single-turn QA with outcome-level metrics, while agentic hallucination benchmarks lack legal-specific diagnostic capability. Neither answers to what extent and how a legal agent hallucinates along its trajectory. To address these limitations, we introduce LexAgentHallu, a legal agentic hallucination benchmark designed to evaluate to what extent and how legal agents fail along multi-step trajectories. Built through a four-stage expert-in-the-loop pipeline, LexAgentHallu contains 3414 instances across 17 legal categories and 6 task types. Each instance is annotated under a dual-layer hallucination taxonomy of 7 high-level categories and 27 fine-grained subclasses, covering both substantive errors and agent-procedural failures. We further design fine-grained metrics that quantify to what extent and localize how each failure occurs along an agent's execution path. Our evaluation across 18 proprietary and open-source agents uncovers a Right-Answer-Wrong-Reason effect and reveals that hallucination subclasses cluster rather than scatter, forming distinct agentic framework, legal task, and category profiles. These findings, invisible to outcome-level evaluation, validate the diagnostic power of LexAgentHallu for evaluating agentic hallucination in law.
Chinese Translation
随着大型语言模型越来越多地被用作工具增强的法律代理人，它们引入了代理幻觉，其中工具调用和推理错误级联成虚构的裁决和错误引用的权威。然而，现有的法律基准仅评估单轮问答（QA）并使用结果级指标，而代理幻觉基准缺乏法律特定的诊断能力。现有研究未能回答法律代理人在其执行过程中幻觉的程度和方式。为了解决这些局限性，我们推出了LexAgentHallu，这是一个旨在评估法律代理人在多步骤执行过程中失败的程度和方式的法律代理幻觉基准。LexAgentHallu通过一个四阶段的专家参与流程构建，包含3414个实例，涵盖17个法律类别和6种任务类型。每个实例根据7个高层类别和27个细分子类的双层幻觉分类法进行标注，涵盖实质性错误和代理程序性失败。我们进一步设计了细粒度指标，以量化每个失败发生的程度并定位其在代理执行路径上的发生方式。我们对18个专有和开源代理的评估揭示了“正确答案错误理由”效应，并显示幻觉子类聚集而非分散，形成独特的代理框架、法律任务和类别特征。这些发现对于结果级评估是不可见的，验证了LexAgentHallu在评估法律领域代理幻觉的诊断能力。
cs.AI / 26 / 2609.09774
Procedural Memory Under Change: Reuse and Interference in Controlled Web Tasks
变化中的程序性记忆：受控网络任务中的重用与干扰
Abstract
Procedural memory lets language agents reuse successful routines, but reuse presumes that a stored routine remains applicable. We study what happens when that presumption is deliberately violated. The study combines a retrospective, human-assisted interface-adaptation case from BrowserGym TimeWarp with controlled frozen-memory comparisons on synthetic shopping decisions. During the documented WebShop V1-V6 development path, interface-specific code was adapted while the separately stored high-level procedure was not reported to change; this phase does not constitute an autonomous memory-agent evaluation. In the controlled phase, an early pilot produced one task on which two memory conditions selected a more expensive item while the no-memory condition selected the reference minimum. Follow-up probes did not establish a recurring row-order or identity-binding pattern. We then tested four forms of mismatch: changed quantities, a different evidence representation, a conflict between local and global optimization, and distributed promotion evidence, across 32 formal cells. Each cell used one temperature-0 generation with the same local qwen3:8b configuration and no adaptive retry. Across these pairs, none of the predefined diagnostic interference signatures appeared on the tasks for which they were defined when current-task evidence was explicit and sufficient. The result identifies a tested region of non-interference: a procedural memory can be mismatched without becoming behaviorally disruptive. It does not establish general safety or a mechanism. The remaining question is which additional conditions turn applicability mismatch into observable, memory-caused error.
Chinese Translation
程序性记忆使语言代理能够重用成功的例程，但重用假定存储的例程仍然适用。我们研究当这一假设被故意违反时会发生什么。该研究结合了来自 BrowserGym TimeWarp 的回顾性人机协作界面适应案例，以及对合成购物决策的受控冻结记忆比较。在记录的 WebShop V1-V6 开发路径中，特定于界面的代码被适应，而单独存储的高层程序并未报告发生变化；这一阶段并不构成自主记忆代理的评估。在受控阶段，早期的试点产生了一个任务，在该任务中，两种记忆条件选择了更昂贵的项目，而无记忆条件选择了参考最低值。后续探测未能建立重复的行顺序或身份绑定模式。随后，我们测试了四种形式的不匹配：数量变化、不同的证据表示、局部与全局优化之间的冲突，以及分布式推广证据，共32个正式单元。每个单元使用一种温度为0的生成，具有相同的局部 qwen3:8b 配置且没有自适应重试。在这些对中，当当前任务证据明确且充分时，定义的干扰特征在其定义的任务中均未出现。结果确定了一个经过测试的非干扰区域：程序性记忆可以不匹配而不导致行为上的干扰。这并未建立一般的安全性或机制。剩下的问题是哪些额外条件将适用性不匹配转变为可观察的、由记忆引起的错误。
cs.AI / 27 / 2609.09776
Proof-Carrying Cognition: Closing the Verification Gap with Reality-Settled Reward
证明携带认知：通过现实确定的奖励缩小验证差距
Abstract
Frontier gains in language-model reasoning come from reinforcement learning on reasoning traces and are concentrated in domains with a cheap, sound verifier. We argue the field's binding constraint is the verification gap: no scalable, incorruptible reward for reasoning outside formal domains. We make four contributions. (1) Theory: in a joint-Gaussian model of best-of-N selection, verifier-gold correlation rho is the exact exchange rate between test-time compute and capability, and an unsound verifier pays a polynomial penalty N^(1/rho^2); a margin-free copula form predicts realized soundness of real LLM judges to 4% median error. (2) Demonstration: in program-synthesis testbeds with executable ground truth, including a pre-registered scaled replication, unsound verifiers lose Soundness-under-Pressure as optimization grows (0.94 to 0.32 at N=4096) while a sound verifier improves monotonically; reality-anchored settlement beats a frozen verifier under i.i.d. and adversarial pressure, driving the hacking gap from ~0.27 to ~0; soundness scales log-linearly with settled labels, with on-policy settlement ~10x more label-efficient than random labeling. With real LLM judges and unit-test execution as gold, a weak judge loses soundness under best-of-N (p<0.001), a stronger judge is more robust, and selection alone manufactures +0.53 hacking gaps from honest samples. Under real GRPO training, a frozen reward model traces the full overoptimization curve (executed reward collapses 90%) while the same model refit on a 10% settlement stream preserves 6x the executed reward. (3) Paradigm: proof-carrying cognition, where reasoning steps are typed probabilistic claims priced by a self-built world model trained only on held-out reality and settled by proper scoring rules. (4) Benchmark: we specify Soundness-under-Pressure as the headline metric for a reality-settled reasoning benchmark.
Chinese Translation
语言模型推理的前沿进展来自于对推理轨迹的强化学习，并集中在具有廉价、可靠验证器的领域。我们认为该领域的约束条件是验证差距：在非正式领域中没有可扩展、不可腐蚀的推理奖励。我们做出四项贡献。(1) 理论：在最佳选择的联合高斯模型中，验证器-黄金相关性 rho 是测试时计算与能力之间的确切汇率，而不可靠的验证器支付多项式惩罚 N^(1/rho^2)；无边际的 copula 形式预测真实 LLM 判断者的实现可靠性中位误差为 4%。(2) 演示：在具有可执行真实值的程序合成测试平台中，包括一个预注册的规模复制，不可靠的验证器在优化增长时失去压力下的可靠性（在 N=4096 时从 0.94 降至 0.32），而可靠的验证器则单调改善；现实锚定的结算在独立同分布和对抗压力下优于冻结验证器，将黑客差距从 ~0.27 降至 ~0；可靠性与已结算标签呈对数线性扩展，政策内结算的标签效率约为随机标记的 10 倍。以真实 LLM 判断者和单元测试执行作为黄金标准，弱判断者在最佳选择中失去可靠性（p<0.001），而强判断者则更具鲁棒性，仅选择就从诚实样本中制造出 +0.53 的黑客差距。在真实 GRPO 训练下，冻结奖励模型追踪完整的过度优化曲线（执行奖励崩溃 90%），而在 10% 结算流上重新拟合的同一模型则保留了 6 倍的执行奖励。(3) 范式：证明携带认知，其中推理步骤被类型化为由仅在保留现实上训练的自建世界模型定价的概率性声明，并通过适当的评分规则进行结算。(4) 基准：我们将压力下的可靠性指定为现实结算推理基准的主要指标。
cs.AI / 28 / 2609.09815
UnitBoost: Managing Compound LLM Systems with a Merge Operator, Not a Model
UnitBoost：通过合并操作符管理复合 LLM 系统，而非模型
Abstract
Compound LLM systems often solve a coordination problem by adding a higher-level LLM. The resulting meta-agent reads workers' outputs, writes the final answer, allocates later calls, and decides when to stop. It is expressive, but it also concentrates three control decisions in an opaque, order-sensitive model call. We ask whether the manager needs to be generative at all. UnitBoost replaces that model with a defined meta-level operator: a task-given unit map turns worker outputs into slot-value proposals, a constrained argmax assembles the output, and the slots left unfilled or unsupported become an explicit residual for the next round. The operator is order-free, records unit provenance, and gives a simple guarantee: without coupling constraints, unit-wise maximization under the same admission score dominates selection of any complete candidate. On three held-out benchmarks, it exceeds the best single candidate chosen with gold labels by 0.060-0.195 absolute task-score points and input-matched generative managers by 0.048-0.076. Replacing only the management step improves six compound-system configurations by 0.013-0.182. Residual-directed rounds raise FanOutQA cell F1 from 0.4778 to 0.5524; matched controls show that the true residual outperforms random targets and ordinary rereading, while a label-free supply signal flags exhaustion after one unproductive round. The same analysis measures three conditions in which no such gain is available (one indivisible unit, unavailable unit identity, and an endpoint that charges for every emitted unit) and quantifies cross-unit coupling as a repair cost. The manager gives up semantic freedom and gains order invariance, unit provenance, and testable failure conditions.
Chinese Translation
复合 LLM 系统通常通过添加一个更高层次的 LLM 来解决协调问题。生成的元代理读取工作者的输出，撰写最终答案，分配后续调用，并决定何时停止。虽然这种方法具有表现力，但它也将三个控制决策集中在一个不透明且对顺序敏感的模型调用中。我们提出管理者是否真的需要生成性。UnitBoost 用一个定义明确的元级操作符替代了该模型：一个任务给定的单位映射将工作者的输出转化为槽值提案，一个受限的 argmax 组装输出，而未填充或不支持的槽则成为下一轮的显式残差。该操作符不受顺序限制，记录单位来源，并提供一个简单的保证：在没有耦合约束的情况下，基于相同接纳分数的单位最大化优于任何完整候选的选择。在三个保留的基准测试中，它的表现超过了通过黄金标签选择的最佳单一候选者 0.060-0.195 个绝对任务分数点，以及与输入匹配的生成管理者 0.048-0.076。仅替换管理步骤就使六个复合系统配置的性能提高了 0.013-0.182。残差导向的轮次将 FanOutQA 单元 F1 从 0.4778 提升至 0.5524；匹配控制显示真实残差优于随机目标和普通重读，而无标签的供应信号在一次无效轮次后标记耗尽。相同的分析测量了三种没有可用增益的条件（一个不可分割的单位、不可用的单位身份，以及一个对每个发出的单位收费的终点），并量化跨单位耦合作为修复成本。管理者放弃了语义自由，获得了顺序不变性、单位来源和可测试的失败条件。
cs.AI / 29 / 2609.09853
The Era by Eon Benchmark: A Generated Enterprise Estate with Exact Ground Truth for Benchmarking LLM Agents
Eon基准时代：一个具有精确真实数据的生成企业环境，用于评估LLM代理
Abstract
LLM agents for enterprise systems of record cannot be evaluated on customer production data, and no existing substitute provides ground truth. We present the Era by Eon Benchmark for evaluating LLM agents that use enterprise tools. The benchmark is built around a complete fictional company. It includes product simulators, company-specific internal databases, benchmark questions, and computed answer keys. Industry, company size, business model, application portfolio, and a seed define each company. One seeded entity graph supplies shared company data to simulators of Salesforce, Zendesk, Slack, Gong, and other products. A questionconditioned generator creates the schemas and records for internal databases. It takes shared entities, keys, and values from the same graph before generating database-specific facts. Both mechanisms therefore describe one consistent enterprise estate. Every expected answer is computed from the final records, so grading is exact. Design and answer-key checks validate the internal databases. A realism scorecard and adversarial detector validate the entity graph. Across 23 generated companies, the mean realism score rose from 61.8 to 97.0, with zero records flagged as synthetic. In the reported simulator-track comparison, nine models answered the same 33 questions three times each. Accuracy estimates ranged from 42.4% to 76.8%, and three of 36 pairwise differences remained supported after correction.
Chinese Translation
企业记录系统的LLM代理无法在客户生产数据上进行评估，现有的替代方案也无法提供真实数据。我们提出了Eon基准，用于评估使用企业工具的LLM代理。该基准围绕一个完整的虚构公司构建。它包括产品模拟器、公司特定的内部数据库、基准问题和计算出的答案键。行业、公司规模、商业模式、应用组合和种子定义了每个公司。一个种子实体图为Salesforce、Zendesk、Slack、Gong等产品的模拟器提供共享公司数据。一个基于问题的生成器创建内部数据库的模式和记录。它从同一图中提取共享实体、键和值，然后生成特定于数据库的事实。因此，这两种机制描述了一个一致的企业环境。每个预期答案都是从最终记录中计算得出的，因此评分是精确的。设计和答案键检查验证了内部数据库。现实性评分卡和对抗检测器验证了实体图。在23个生成的公司中，平均现实性评分从61.8上升至97.0，且没有记录被标记为合成。在报告的模拟器轨迹比较中，九个模型对同一33个问题各回答三次。准确率估计范围从42.4%到76.8%，在修正后，36个成对差异中有三个仍然得到支持。
cs.AI / 30 / 2609.09864
Shifting Relational Paradigms for Affective Computing: Affective Resonance, Vitality Affects, and Vocal Interaction Fields
情感计算的关系范式转变：情感共鸣、活力情感与声学互动场
Abstract
Affective computing has largely followed an individual-state paradigm, extracting discrete emotion labels or arousal/valence from isolated speakers. We argue this framing is incomplete for interaction. Drawing on affective resonance and vitality-contour accounts, we propose a relational framework in which the primary unit of affective analysis is the interactional field constituted within vocal dynamics. As a proof of concept, we present a preliminary empirical study using continuous self-supervised speech representations to detect directional expressive coupling in multi-party conversation. Coupling is regime-specific, concentrated at sub-second timescales, and collapses under exclusive-speech negative controls, consistent with a relational account of affective dynamics. We introduce design frameworks for Artificial Affective Resonance Intelligence grounded in Affective Resonance Dynamic Ontologies, supported by null-calibrated directional coupling analyses across interaction regimes.
Chinese Translation
情感计算在很大程度上遵循个体状态范式，从孤立的说话者中提取离散的情感标签或唤醒/价值感。我们认为这种框架对于互动来说是不完整的。基于情感共鸣和活力轮廓的理论，我们提出了一种关系框架，其中情感分析的主要单位是由声学动态构成的互动场。作为概念验证，我们展示了一项初步的实证研究，使用连续自监督的语音表示来检测多方对话中的方向性表现耦合。耦合是特定于状态的，集中在亚秒级时间尺度，并在独占性发言的负控制下崩溃，这与情感动态的关系解释一致。我们引入了基于情感共鸣动态本体的人工情感共鸣智能设计框架，并通过跨互动状态的零校准方向耦合分析提供支持。
cs.AI / 31 / 2609.09875
AgentAudit: An Open, Extensible Framework for Full-Lifecycle Trust Evaluation of AI Agents
AgentAudit：一个开放的、可扩展的框架，用于人工智能代理的全生命周期信任评估
Abstract
Existing evaluation frameworks mostly assess only one part of AI agents, such as task completion (AgentBench) or security robustness (AgentDojo, ASB), rather than the complete pipeline of planning, tool selection, tool execution, memory and reasoning. Failures can occur at any stage, yet existing benchmarks rarely identify their precise source. AgentAudit evaluates the entire execution trace across ten capability, grounding, security and behavioural dimensions, namely instruction integrity, planner, memory, tool selection, tool invocation, tool correctness, alignment, tool faithfulness, security and execution integrity, combined with behavioural classification and failure attribution to pinpoint the exact stage responsible for an observed failure. AgentAudit can evaluate any LLM-based AI agent, since it attaches to the agent instead of replacing it. It reads only the recorded execution trace and does not interfere with how the agent runs, so it places no constraint on the agent's internal implementation. We evaluate five language models (OpenAI GPT-5, Claude Sonnet 5, Sarvam 105B, Llama 3.3 70B and Gemini 2.5 Flash) across nine capability and adversarial tasks. Claude Sonnet 5 and GPT-5 obtain the highest mean Composite Trust Scores (95.1 and 80.6 out of 100, respectively), while Sarvam 105B, Llama 3.3 70B and Gemini 2.5 Flash trail substantially (57.6, 45.7 and 22.6). All traces were scored by a single fixed judge model, which was itself one of the evaluated models, a limitation discussed in Section VII.E. More importantly, models with similar task-completion behaviour can diverge sharply in trustworthiness, as several non-frontier models are repeatedly classified Unsafe_Compliance on adversarial tasks rather than merely failing them, a distinction that pass/fail benchmarks cannot surface.
Chinese Translation
现有的评估框架大多仅评估人工智能代理的某一部分，例如任务完成（AgentBench）或安全性稳健性（AgentDojo, ASB），而不是规划、工具选择、工具执行、记忆和推理的完整流程。失败可能在任何阶段发生，但现有基准很少能准确识别其来源。AgentAudit 评估整个执行轨迹，涵盖十个能力、基础、安保和行为维度，即指令完整性、规划者、记忆、工具选择、工具调用、工具正确性、对齐、工具忠实性、安全性和执行完整性，并结合行为分类和失败归因，以准确找出导致观察到的失败的具体阶段。AgentAudit 可以评估任何基于 LLM 的人工智能代理，因为它附加在代理上而不是替代它。它仅读取记录的执行轨迹，不干扰代理的运行方式，因此对代理的内部实现没有任何限制。我们在九个能力和对抗任务中评估了五个语言模型（OpenAI GPT-5、Claude Sonnet 5、Sarvam 105B、Llama 3.3 70B 和 Gemini 2.5 Flash）。Claude Sonnet 5 和 GPT-5 获得了最高的平均综合信任分数（分别为 95.1 和 80.6 分，满分 100 分），而 Sarvam 105B、Llama 3.3 70B 和 Gemini 2.5 Flash 的分数则显著较低（分别为 57.6、45.7 和 22.6）。所有轨迹均由一个固定的评判模型评分，该模型本身也是被评估的模型之一，这一限制在第 VII.E 节中进行了讨论。更重要的是，具有相似任务完成行为的模型在可信度上可能存在显著差异，因为一些非前沿模型在对抗任务中被反复分类为不安全合规（Unsafe_Compliance），而不仅仅是未能完成任务，这一区别是通过合格/不合格基准无法显现的。
cs.AI / 32 / 2609.09882
Scored vs. Generated Readouts in Behavioral Language Models: An Empirical Study of Elicitation Format
行为语言模型中的评分与生成读出：引出格式的实证研究
Abstract
Language models fine-tuned on customer behavior can predict outcomes and generate explanations, but these readouts are often treated as interchangeable. Holding model checkpoint and prompt content fixed, we compare probabilities obtained by scoring answer tokens with predictions generated after a written rationale. Across 13 model-domain cells covering four retail tasks in three markets, including two using fully public data and checkpoints, the scored readout ranks outcomes more accurately in 12 of 13 cells (two-sided sign test, p approximately 0.003), by 1.5 to 14.5 points in area under the receiver operating characteristic curve (AUC). Paired bootstrap confidence intervals exclude zero in every newly measured cell. The gap varies with task-specific supervision and mismatch between training and serving formats, ranging from -2.2 points for an untuned base model to +13.7 for rationale-format supervision. Analysis of approximately 9,000 rationales identifies two correlates: reduced reliance on the dominant predictive feature and convergence on stock formulations. Probability saturation does not track the gap. A third readout, eliciting a probability before any verdict, improves calibration (Brier score from 0.47 to 0.15) while ranking within noise of scoring, but only for outcome rates represented in training; it is worse than scoring when the scored head is already calibrated. We interpret these differences through the objectives matched by each readout, identify training choices that narrow the gap, and propose retaining generated rationales while sourcing ranking from the scored head.
Chinese Translation
针对客户行为微调的语言模型能够预测结果并生成解释，但这些读出通常被视为可互换的。在固定模型检查点和提示内容的情况下，我们比较了通过评分答案标记获得的概率与在书面理由之后生成的预测。在覆盖三个市场中四个零售任务的13个模型-领域单元中，包括两个使用完全公开数据和检查点的单元，评分读出在13个单元中的12个中更准确地排名结果（双侧符号检验，p 约为 0.003），在接收者操作特征曲线（AUC）下的面积提高了1.5到14.5分。配对自助法置信区间在每个新测量的单元中均排除了零。这个差距与任务特定的监督和训练与服务格式之间的不匹配有关，范围从未调优的基础模型的-2.2分到理由格式监督的+13.7分。对约9,000个理由的分析识别出两个相关因素：对主导预测特征的依赖减少以及对标准公式的趋同。概率饱和度并未跟踪这一差距。第三种读出在任何裁决之前引出概率，改善了校准（Brier 分数从 0.47 降至 0.15），同时在评分的噪声中排名，但仅适用于训练中表示的结果率；当评分头已经校准时，它的表现不如评分。我们通过每种读出所匹配的目标来解释这些差异，识别出缩小差距的训练选择，并建议保留生成的理由，同时从评分头获取排名。
cs.AI / 33 / 2609.09885
Decision Transformer for UAV-Mounted RIS-Assisted Dynamic D2D Communications
基于决策变换器的无人机搭载可重构智能表面辅助动态设备间通信
Abstract
This paper studies unmanned aerial vehicle (UAV)-mouted reconfigurable intelligent surface (RIS)-assisted device-to-device (D2D) communication with stochastic link activation. It models UAV motion and attitude, time-varying Rician angles, and angle-dependent RIS reflection. A joint optimization of UAV trajectory, attitude, and RIS phases is formulated to maximize average sum rate under mobility, energy, and hardware constraints. The problem is addressed using deep reinforcement learning and a Decision Transformer trained on expert trajectories from multiple scenarios. Results demonstrate effective cross-scenario generalization, with zero-shot transfer outperforming direct DRL transfer and online fine-tuning achieving competitive performance with fewer interactions.
Chinese Translation
本文研究了无人机（UAV）搭载的可重构智能表面（RIS）辅助的设备间（D2D）通信，考虑了随机链路激活。我们对无人机的运动和姿态、时变的瑞利角以及与角度相关的RIS反射进行了建模。提出了一种联合优化方案，旨在最大化在移动性、能量和硬件约束下的平均总速率，优化无人机的轨迹、姿态和RIS相位。该问题采用深度强化学习和基于多个场景专家轨迹训练的决策变换器（Decision Transformer）进行解决。结果表明，该方法在跨场景泛化方面表现出色，零样本迁移优于直接的深度强化学习迁移，而在线微调则在较少的交互下实现了具有竞争力的性能。
cs.AI / 34 / 2609.09898
Grounded Evaluation and Repair for NL-to-PDDL Problem Generation
基于实证的自然语言到PDDL问题生成的评估与修复
Abstract
Large Language Models (LLMs) have shown promise for translating Natural Language (NL) planning descriptions into PDDL problem instances. However, standard evaluation criteria such as syntactic validity or planner success can substantially overestimate faithfulness to the described task: a generated problem may be parseable and solvable while misrepresenting the intended initial state, goal, object structure, or optimization target. This paper studies an end-to-end NL-to-PDDL pipeline that combines LLM generation, checks in terms of PDDL parsing, planning and validation, a domain-conformance checker, an LLM critic, and iterative repair. Fine-grained repair feedback is constructed from the domain description, the generated problem, the natural language problem description, and operational diagnostics. Reference-based comparisons against curated benchmark PDDL problem descriptions are used for post-hoc benchmark analysis, and these offline checks include renaming-invariant structural matching and semantic equivalence, where domain support is available. Across Planetarium, AutoPlanBench, and curated PDDL~2.1 problems, results show that operational success and benchmark-reference reconstruction can diverge substantially. Results also show that structured repair can be useful, and that PDDL~2.1 remains challenging for reference reconstruction, even when operational success improves.
Chinese Translation
大型语言模型（LLMs）在将自然语言（NL）规划描述转换为PDDL问题实例方面显示出了潜力。然而，标准评估标准如语法有效性或规划者成功率可能会大幅高估对所描述任务的忠实度：生成的问题可能是可解析和可解决的，但却错误地表现了预期的初始状态、目标、对象结构或优化目标。本文研究了一种端到端的NL到PDDL管道，该管道结合了LLM生成、PDDL解析、规划和验证方面的检查、领域一致性检查器、LLM评论器以及迭代修复。细粒度的修复反馈是基于领域描述、生成的问题、自然语言问题描述和操作诊断构建的。基于参考的比较与策划的基准PDDL问题描述用于事后基准分析，这些离线检查包括重命名不变的结构匹配和语义等价性（在领域支持可用的情况下）。在Planetarium、AutoPlanBench和策划的PDDL~2.1问题中，结果显示操作成功与基准参考重建之间可能存在显著差异。结果还表明，结构化修复是有用的，并且即使在操作成功改善的情况下，PDDL~2.1在参考重建方面仍然具有挑战性。
cs.AI / 35 / 2609.09925
Time-Frequency Geometric Cross-Attention for Chunked Vision-Language-Action Models
用于分块视觉-语言-动作模型的时频几何交叉注意力
Abstract
Modern vision-language-action (VLA) policies predict a whole chunk of actions: one to two seconds of coordinated motion emitted in a single forward pass. Yet an action chunk is essentially a short multivariate trajectory, but inside these models it is a sequence of generic per-timestep hidden tokens decoded by a linear head. This under-serves two motion structures. First, frequency: a chunk superimposes a smooth global trend and fine corrective motion across time scales, and a single token entangles them. Second, cross-phase geometry: motions of different phases (reach, contact, grasp adjustment, settling) unfold along very different, near-orthogonal directions in representation space, yet are tightly related for the task and arise across the time axis. Dot-product attention scores alignment by an inner product, so it favors aligned tokens and is least sensitive near orthogonality, leaving such relationships for the network to recover through a detour. We introduce Time-Frequency Geometric Cross-Attention (TFGCA), a drop-in module repairing both blind spots. TFGCA uses a per-dimension learnable stationary wavelet transform to decompose the action chunk into time-frequency tokens, and each time token retrieves information from them via a cross-attention that fuses the dot product (similarity) with the wedge-product magnitude (sensitive to near-orthogonality) through a learnable weight. A zero-initialized residual reproduces the base behavior at initialization, so it can be dropped onto a pretrained VLA and fine-tuned jointly. Relative to the same-source base, TFGCA improves in-distribution LIBERO by +1.5 on average, the OOD LIBERO-Plus by +6.3, the randomized average under RoboTwin domain randomization by +28.5, and the overall success rate on three real-robot AgiBot A2 tasks by +11.67 points, with larger gains out of distribution.
Chinese Translation
现代视觉-语言-动作（VLA）策略预测一整块动作：在一次前向传递中发出的持续一到两秒的协调运动。然而，一个动作块本质上是一个短的多变量轨迹，但在这些模型中，它是由线性头解码的一系列通用的每时间步隐藏标记。这未能充分利用两种运动结构。首先，频率：一个块在时间尺度上叠加了平滑的全局趋势和细微的修正运动，而单个标记将它们纠缠在一起。其次，交叉相位几何：不同相位的运动（到达、接触、抓握调整、安置）在表示空间中沿着非常不同、近正交的方向展开，但对于任务而言它们是紧密相关的，并且在时间轴上产生。点积注意力通过内积来评估对齐，因此它偏向于对齐的标记，并且在近正交时最不敏感，从而将这种关系留给网络通过绕道恢复。我们引入了时频几何交叉注意力（TFGCA），一个修复这两个盲点的模块。TFGCA使用每个维度可学习的平稳小波变换将动作块分解为时频标记，每个时间标记通过交叉注意力从中检索信息，该交叉注意力将点积（相似性）与楔积的大小（对近正交敏感）通过可学习的权重融合。一个零初始化的残差在初始化时重现基础行为，因此可以直接应用于预训练的VLA并进行联合微调。相对于相同来源的基础，TFGCA在分布内的LIBERO上平均提高了1.5，在OOD LIBERO-Plus上提高了6.3，在RoboTwin领域随机化下的随机平均提高了28.5，在三个真实机器人AgiBot A2任务上的整体成功率提高了11.67个百分点，且在分布外的增益更大。
cs.AI / 36 / 2609.09928
Structural Process Supervision for Latent Chain-of-Thought Reasoning
潜在思维链推理的结构过程监督
Abstract
Latent reasoning approaches enhance token-level efficiency and robustness by replacing verbose, explicit chain-of-thought (CoT) tokens with compact continuous-space embeddings. However, existing methods lack direct process supervision over these latent embeddings, which often leads to representation collapse and uneven information distribution. To address this, we propose Prototype-Mediated Process Supervision (PMPS), which introduces learnable reasoning prototypes as semantic anchors to provide structural process-level supervision for latent reasoning. PMPS projects latent embeddings and explicit CoT embeddings into a shared prototype space, achieving many-to-many soft alignment between unequal-length representations through prototype assignment. Meanwhile, we introduce a Progressive Sequential Alignment (PSA) module to further guide training: positional priors initially encourage sequential alignment structure, then gradually relax to permit adaptive matching. Experimental results show that PMPS compresses output token length to under 50% of explicit CoT on GSM8K-Aug. Compared to leading baseline SIM-CoT, our method achieves average accuracy gains of 2.08% across different model families. On GPT-2, PMPS even surpasses CoT-SFT. On larger models and a more challenging task, PMPS consistently attains the highest accuracy among all latent reasoning methods with comparable output length.
Chinese Translation
潜在推理方法通过用紧凑的连续空间嵌入替代冗长的显式思维链（Chain-of-Thought, CoT）标记，从而提高了标记级别的效率和鲁棒性。然而，现有方法缺乏对这些潜在嵌入的直接过程监督，这往往导致表示崩溃和信息分布不均。为了解决这个问题，我们提出了原型介导的过程监督（Prototype-Mediated Process Supervision, PMPS），该方法引入可学习的推理原型作为语义锚点，为潜在推理提供结构化的过程级监督。PMPS将潜在嵌入和显式CoT嵌入投影到共享的原型空间，通过原型分配实现不等长表示之间的多对多软对齐。同时，我们引入了渐进式序列对齐（Progressive Sequential Alignment, PSA）模块以进一步指导训练：位置先验最初鼓励序列对齐结构，然后逐渐放宽以允许自适应匹配。实验结果表明，PMPS将输出标记长度压缩至显式CoT的50%以下。在与领先基线SIM-CoT的比较中，我们的方法在不同模型系列中平均提高了2.08%的准确率。在GPT-2上，PMPS甚至超越了CoT-SFT。在更大模型和更具挑战性的任务上，PMPS在所有潜在推理方法中始终保持最高的准确率，同时输出长度相当。
cs.AI / 37 / 2609.10036
Belief-State Engine: Augmenting LLMs for Principled Planning Under Partial Observability
信念状态引擎：增强大型语言模型在部分可观测环境下的原则性规划
Abstract
Large language model agents produce fluent action sequences across a wide range of tasks, yet they fail in characteristic ways once the environment becomes partially observable. Ambiguous feedback pushes them into premature commitments. A single informative observation can collapse their uncertainty onto the wrong hypothesis. Policies drift as the history grows. We trace these symptoms to a common structural cause. An LLM agent, as commonly deployed, is a history-conditioned policy with no explicit belief over hidden state. We propose an architectural fix. The Belief-State Engine (BSE) is an inference module placed outside the LLM. It maintains a Bayesian posterior over the latent states of a given POMDP (Partially Observable Markov Decision Process) model, and at each decision step it exposes only that posterior to the LLM. The raw action-observation log is not shown. We set out a minimal four-axiom specification of what a belief-consistent internal state must satisfy, and prove that the LLM paired with the BSE is a sound Markov policy on the belief MDP induced by the underlying POMDP. It therefore inherits the Bellman optimality guarantees of classical POMDP theory, provided the LLM is never exposed to the raw history. We evaluate the architecture on the Tiger POMDP and a red-team attack-graph task, against six baselines: a reactive LLM, Chain-of-Thought, ReAct, a natural-language belief tracker, QMDP, and POMCP. Across both domains, the BSE-augmented agent improves task return, belief calibration, and decision consistency. Ten targeted ablations isolate the contribution of each architectural choice confirms that the effect is not specific to any one model. Code, environment specifications, prompt templates, and seed logs accompany this paper.
Chinese Translation
大型语言模型代理能够在广泛的任务中生成流畅的行动序列，但一旦环境变得部分可观测，它们便会以特有的方式失败。模糊的反馈使它们过早地做出承诺。单一的信息观察可能会将它们的不确定性压缩到错误的假设上。随着历史的增长，策略会发生漂移。我们将这些症状追溯到一个共同的结构性原因。通常部署的LLM代理是一个历史条件的策略，但对隐藏状态没有明确的信念。我们提出了一种架构修正。信念状态引擎（Belief-State Engine, BSE）是一个置于LLM外部的推理模块。它在给定的部分可观测马尔可夫决策过程（Partially Observable Markov Decision Process, POMDP）模型的潜在状态上维持一个贝叶斯后验，并在每个决策步骤中仅将该后验暴露给LLM。原始的行动-观察日志不会被展示。我们提出了一个最小的四公理规范，说明信念一致的内部状态必须满足的条件，并证明与BSE配对的LLM是在由基础POMDP诱导的信念马尔可夫决策过程（MDP）上的一个有效马尔可夫策略。因此，只要LLM从未接触原始历史，它就继承了经典POMDP理论的贝尔曼最优性保证。我们在老虎POMDP和红队攻击图任务上评估该架构，针对六个基线进行比较：反应式LLM、思维链（Chain-of-Thought）、反应（ReAct）、自然语言信念追踪器、QMDP和POMCP。在这两个领域中，增强的BSE代理改善了任务回报、信念校准和决策一致性。十项有针对性的消融实验隔离了每个架构选择的贡献，确认该效果并不特定于任何单一模型。代码、环境规格、提示模板和种子日志随本文附上。
cs.AI / 38 / 2609.10055
OntologyAligner: Ontology-Aligned Retrieval and Hierarchy-Guided Large Language Model Reranking for Biomedical Ontology Normalization
OntologyAligner：用于生物医学本体规范化的本体对齐检索与层次引导的大型语言模型重排序
Abstract
Biomedical ontology normalization maps free-text expressions to standardized concepts, enabling consistent integration and analysis of biomedical data. This task remains challenging because lexical variation and subtle distinctions among hierarchically related concepts can obscure concept boundaries. We present OntologyAligner, a three-stage framework that combines ontology-aligned retrieval, large language model candidate reranking, and selective hierarchy-guided refinement. We also construct PhenoNormBench, a unified benchmark comprising 13,390 samples from seven Human Phenotype Ontology datasets. OntologyAligner achieved state-of-the-art performance on HPO normalization, with 88.78% Macro Top-1 Accuracy and 86.75% Micro Top-1 Accuracy, exceeding the strongest baseline by 4.85 and 5.07 percentage points, respectively. Ablation analyses showed complementary contributions from all three stages, and sensitivity analyses demonstrated stability across candidate-set sizes and model backbones. Applications to MONDO, MEDIC, and NCBITaxon further established portability to other ontologies. OntologyAligner offers a generalizable framework for accurate mapping of biomedical text to structured ontology concepts. PhenoNormBench and the code are publicly available at https://github.com/zhelishisongjie/OntologyAligner.
Chinese Translation
生物医学本体规范化将自由文本表达映射到标准化概念，从而实现生物医学数据的一致集成和分析。由于词汇变异和层次相关概念之间的细微差别可能模糊概念边界，这一任务仍然具有挑战性。我们提出了OntologyAligner，一个三阶段框架，结合了本体对齐检索、大型语言模型候选重排序和选择性层次引导的细化。我们还构建了PhenoNormBench，一个统一的基准，包含来自七个人体表型本体数据集的13,390个样本。OntologyAligner在HPO规范化上实现了最先进的性能，Macro Top-1准确率为88.78%，Micro Top-1准确率为86.75%，分别超越了最强基线4.85和5.07个百分点。消融分析显示三个阶段的互补贡献，敏感性分析证明了在候选集大小和模型骨干上的稳定性。对MONDO、MEDIC和NCBITaxon的应用进一步确立了其对其他本体的可移植性。OntologyAligner提供了一个可推广的框架，用于将生物医学文本准确映射到结构化本体概念。PhenoNormBench和代码可在https://github.com/zhelishisongjie/OntologyAligner公开获取。
cs.AI / 39 / 2609.10060
Reference-Based Bias Detection in LLMs via Relative Representations of Hidden States
基于参考的LLMs偏见检测：通过隐藏状态的相对表示
Abstract
Existing bias auditing methods typically rely on model outputs, requiring costly benchmarks or judge models and potentially missing internal shifts that never appear in generated text. We propose a reference-based method that audits bias in hidden-state representations across related model variants, for example before and after fine-tuning. Because fine-tuning reshapes representation geometry, absolute hidden states are not directly comparable, so we encode each sentence by its similarities to a fixed set of anchor sentences, yielding relative representations in a shared comparison space. There we measure how target groups shift in their association with positive and negative attributes, a quantity we call the Representational Bias Shift $\Delta B$. Across three model families and the WildGuardMix, DecodingTrust and ToxiGen benchmarks, $\Delta B$ correlates with output-level bias change in 15 of the 18 settings we test, reaching $|r| = 0.84$ ($p < 0.001$) under full fine-tuning and becoming more model-dependent under parameter-efficient adaptation. Thresholding $\Delta B$ detects checkpoints whose bias increased with ROC AUC between $0.65$ and $0.99$, and on WildGuardMix and DecodingTrust it separates them better than a SEAT-based baseline for all three families. $\Delta B$ is also stable under changes to the anchor set, attribute sets and target templates. Our method requires no task-specific evaluation data and audits a model in about three minutes, using $3$-$50\times$ less compute than the output-level benchmarks considered here. We view it as complementary to output-based auditing rather than a replacement for it.
Chinese Translation
现有的偏见审计方法通常依赖于模型输出，需耗费大量成本进行基准测试或评判模型，并可能错过那些在生成文本中从未出现的内部变化。我们提出了一种基于参考的方法，通过相关模型变体（例如在微调前后）审计隐藏状态表示中的偏见。由于微调重塑了表示几何，因此绝对隐藏状态不可直接比较，因此我们通过与固定锚句集的相似性对每个句子进行编码，从而在共享比较空间中产生相对表示。在这里，我们测量目标群体在与正面和负面属性的关联中如何变化，这一量称为表示偏见变化（Representational Bias Shift）$ riangle B$。在三个模型家族以及WildGuardMix、DecodingTrust和ToxiGen基准测试中，$ riangle B$与我们测试的18个设置中15个的输出级偏见变化相关，达到$|r| = 0.84$（$p < 0.001$）在完全微调下，并在参数高效适应下变得更加依赖于模型。通过阈值化$ riangle B$，可以检测到偏见增加的检查点，其ROC AUC介于$0.65$和$0.99$之间，并且在WildGuardMix和DecodingTrust上，它比基于SEAT的基线在所有三个家族中表现更好。$ riangle B$在锚集、属性集和目标模板变化下也保持稳定。我们的方法不需要特定任务的评估数据，审计一个模型大约只需三分钟，使用的计算量是这里考虑的输出级基准的$3$-$50 imes$。我们认为这是一种与基于输出的审计互补的方法，而不是对其的替代。
cs.AI / 40 / 2609.10092
RAP: Research Attention Prediction Reveals Target-Conditioned Evidence Acquisition Biases
RAP：研究注意力预测揭示目标条件下的证据获取偏差
Abstract
Large language models (LLMs) increasingly act as research agents, yet their ability to track shifts in research attention is difficult to evaluate because reviews and research ideas lack uniquely verifiable outcomes. We introduce Research Attention Prediction (RAP), a rolling benchmark covering 278 AI/ML fields and 1,390 episodes. At each cut-off, an LLM agent searches a temporally restricted arXiv corpus and predicts the next six months' paper shares across eight frozen research directions. Search generally helps, but all four diagnostic models perform worse than an exact-count exponentially weighted moving average (EWMA) baseline in compositional accuracy. We identify two linked bottlenecks. Under cumulative-history access, State carry-forward outperforms direct Forecast for all four diagnostic models; frozen-evidence replay links a shared component of this reversal to Forecast-oriented policies retrieving a smaller share of recent evidence. Even with exact historical activity, future-specific updating remains limited, with only GPT-5.5 plus reopened Search slightly surpassing EWMA. Fine-tuning on realised outcomes improves Qwen3-4B's forecast Spearman correlation by 0.105 on held-out fields at later origins, with gains also on change-rich episodes.
Chinese Translation
大型语言模型（LLMs）越来越多地充当研究代理，但由于评论和研究想法缺乏独特的可验证结果，因此评估它们跟踪研究注意力变化的能力变得困难。我们引入了研究注意力预测（Research Attention Prediction，RAP），这是一个涵盖278个人工智能/机器学习（AI/ML）领域和1,390个事件的滚动基准。在每个截止点，LLM代理搜索一个时间限制的arXiv文献库，并预测未来六个月在八个固定研究方向上的论文分享。搜索通常有帮助，但所有四个诊断模型在组合准确性上均表现不及精确计数的指数加权移动平均（EWMA）基线。我们识别出两个相关的瓶颈。在累积历史访问下，状态转移（State carry-forward）在所有四个诊断模型中均优于直接预测（Forecast）；冻结证据重播（frozen-evidence replay）将这一逆转的共享组成部分与以预测为导向的政策联系起来，这些政策检索到较少的近期证据。即使有精确的历史活动，未来特定的更新仍然有限，只有GPT-5.5加上重新开启的搜索稍微超过了EWMA。在实现结果上的微调使Qwen3-4B在后期起源的保留领域上预测的Spearman相关性提高了0.105，并且在变化丰富的事件上也有增益。
cs.AI / 41 / 2609.10135
Agent-Based ML-LLM Fusion with Self-Optimizing Prompts for Plateau Weather Alerts
基于代理的机器学习-大语言模型融合与自优化提示在高原天气预警中的应用
Abstract
To address insufficient contextualization, weak generalization, and poor scenario adaptation in tourism meteorological services, we propose SmartWeatherAgent--a unified three-stage architecture integrating intent recognition, hazard prediction, and reasoning-enhanced generation. The system fuses rule-based methods with large language models to parse queries at multiple granularities and employs a LightGBM model enriched with highland-specific features (e.g., wind speed abruptness rate), achieving an F1-Macro score of 0.605 with 1.60 ms latency on high-wind, precipitation, and low-temperature events. A 12-round micro-step prompt self-optimization loop boosts the composite warning quality score S_final from 4.2 (B01) to 8.9 (B12, +112%). Key improvements include a sharp rise in B08 from data source citation (6.5 -> 8.5), sustained high performance in B10 via physical mechanism explanation, and a peak scientific rigor score of 9.2 in B12 through explicit uncertainty statements. The system autonomously generates structured warnings that integrate causal mechanisms, spatiotemporal evolution, quantitative evidence, regulatory references, and confidence statements--enhancing professional depth, logical rigor, and scientific soundness, and advancing meteorological services toward proactive perception, explainable decision-making, and intelligent agency.
Chinese Translation
为了解决旅游气象服务中上下文不足、泛化能力弱和场景适应性差的问题，我们提出了SmartWeatherAgent——一个统一的三阶段架构，集成了意图识别、危险预测和增强推理生成。该系统将基于规则的方法与大语言模型相结合，以多粒度解析查询，并采用丰富高原特征（如风速突变率）的LightGBM模型，在高风、降水和低温事件中实现了0.605的F1-Macro得分，延迟为1.60毫秒。12轮微步提示自优化循环将综合预警质量得分S_final从4.2（B01）提升至8.9（B12，+112%）。关键改进包括B08中数据源引用的显著提升（6.5 -> 8.5），通过物理机制解释在B10中保持高性能，以及在B12中通过明确的不确定性声明达到9.2的科学严谨性得分。该系统自主生成结构化预警，整合因果机制、时空演变、定量证据、监管参考和置信声明——增强了专业深度、逻辑严谨性和科学合理性，推动气象服务朝向主动感知、可解释决策和智能代理的方向发展。
cs.AI / 42 / 2609.10144
Kernel-Managed Shared Memory for System-Wide Personalization
内核管理的共享内存用于系统范围的个性化
Abstract
AI systems become more useful when they can adapt to the people using them, but in multi-agent systems, useful context learned by one agent often remains unavailable to others. We present kernel-managed shared memory, a system-level abstraction in which specialized agents write structured, tagged memories while the agent-system kernel, not individual agents, governs retrieval, privacy enforcement, and prompt injection. We implement and evaluate this design on AIOS and compare it against three alternatives across three assistant models (GPT-4o, Llama-3.1:8B, Qwen-2.5:7B) and 1,800 total trials. Against an unmanaged external memory backend (Mem0) using identical underlying storage, kernel-managed retrieval and injection improve personalization scores by 2.4-4.0 points on a 5-point scale (e.g., 1.05 to 4.69 profile usage on GPT-4o), with every comparison significant at p < 10^-18. Against standard retrieval-augmented injection, gains are similarly large and consistent across all three models. Against full, unfiltered context concatenation, a soft ceiling on available context rather than on response quality, kernel-managed injection statistically matches performance on two of three models and shows a small, model-specific deficit on the third, while using substantially shorter prompts: end-to-end latency is 15-61% lower across all three models, with corresponding reductions in per-call token usage and inference cost. These results indicate that centralizing memory management in the agent-system kernel, rather than leaving retrieval and privacy enforcement to individual agents, delivers most of the personalization benefit of unconstrained context at a fraction of its cost.
Chinese Translation
当人工智能系统能够适应使用它们的人时，它们变得更加有用，但在多智能体系统中，一个智能体学习到的有用上下文往往对其他智能体不可用。我们提出了内核管理的共享内存，这是一种系统级抽象，其中专门的智能体写入结构化、带标签的记忆，而由智能体系统内核（而非单个智能体）管理检索、隐私执行和提示注入。我们在AIOS上实现并评估了该设计，并将其与三种助手模型（GPT-4o、Llama-3.1:8B、Qwen-2.5:7B）下的三种替代方案进行了比较，共进行了1,800次试验。与使用相同底层存储的未管理外部内存后端（Mem0）相比，内核管理的检索和注入在5分制上提高了个性化评分2.4-4.0分（例如，GPT-4o上的个人资料使用从1.05提高到4.69），每次比较的显著性均为p < 10^-18。与标准的检索增强注入相比，增益在所有三种模型中同样显著且一致。与完整的、未过滤的上下文连接相比，内核管理的注入在可用上下文的软上限上而非响应质量上，在三种模型中的两种上统计上匹配性能，并在第三种模型上显示出小的、特定于模型的不足，同时使用显著更短的提示：在所有三种模型中，端到端延迟降低了15-61%，每次调用的令牌使用和推理成本也相应减少。这些结果表明，将内存管理集中在智能体系统内核中，而不是将检索和隐私执行留给单个智能体，可以在其成本的一小部分内实现不受限制的上下文所带来的大部分个性化收益。
cs.AI / 43 / 2609.10177
Beyond Surface Imitation: Contrastive Modeling for Reasoning Path Alignment in Multimodal In-Context Learning
超越表面模仿：用于多模态上下文学习中推理路径对齐的对比建模
Abstract
In-context learning (ICL) is widely used in multimodal large language models (MLLMs) and achieves strong performance across a wide range of multimodal tasks. However, existing multimodal ICL methods often rely on surface level imitation of in-context demonstrations, making it difficult for MLLMs to align their responses with the reasoning path required by the given multimodal input. This limitation becomes more pronounced in complex multimodal tasks, thereby restricting further improvements in MLLM performance. To address this issue, we propose a new multimodal ICL framework that combines contrastive demonstration modeling with the self-refinement capability of MLLMs. Specifically, our framework reformulates each demonstration by explicitly contrasting a suboptimal response with a better response under the same input, together with a reasoning path that reveals how the response should be refined. This contrastive formulation makes the reasoning path toward the desired response more explicit and guides the MLLM beyond superficial imitation. Furthermore, because effective refinement depends on the current response, we introduce a response-conditioned retrieval mechanism to select demonstrations whose reasoning paths are more relevant to the current response. In addition, we use a lightweight alignment controller to predict response quality and determine whether further refinement is needed. Experiments on three types of multimodal tasks show that the proposed framework consistently improves MLLM performance, with particularly notable gains on visual question answering (VQA).
Chinese Translation
上下文学习（ICL）在多模态大型语言模型（MLLMs）中被广泛应用，并在多种多模态任务中取得了良好的表现。然而，现有的多模态ICL方法往往依赖于对上下文示例的表面模仿，这使得MLLMs难以将其响应与给定多模态输入所需的推理路径对齐。这一限制在复杂的多模态任务中尤为明显，从而限制了MLLM性能的进一步提升。为了解决这一问题，我们提出了一种新的多模态ICL框架，该框架将对比示例建模与MLLM的自我精炼能力相结合。具体而言，我们的框架通过明确对比同一输入下的次优响应与更优响应，重新表述每个示例，并提供一个揭示如何精炼响应的推理路径。这种对比表述使得通向期望响应的推理路径更加明确，并引导MLLM超越表面的模仿。此外，由于有效的精炼依赖于当前响应，我们引入了一种响应条件检索机制，以选择与当前响应更相关的示例推理路径。此外，我们使用轻量级对齐控制器来预测响应质量，并确定是否需要进一步的精炼。在三种类型的多模态任务上的实验表明，所提出的框架始终提高了MLLM的性能，尤其在视觉问答（VQA）任务上取得了显著的提升。
cs.AI / 44 / 2609.10221
Why Sample What You Can Enumerate? Exact Policy Optimization for Genomic Tool Selection
为什么要对可枚举的样本进行采样？基于精确策略优化的基因组工具选择
Abstract
Reinforcement learning over a frozen reasoner has become a common recipe for teaching a policy which external tools to invoke. We show that this recipe becomes structurally mismatched in specialist scientific settings where the complete tool-subset space is enumerable. There, a small set of recurring computational capabilities covers the domain, so the space of tool subsets is combinatorial yet small enough to enumerate, and GRPO still estimates an action expectation from a handful of sampled rollouts. Worse, the approximation degrades as training succeeds: as the policy concentrates on preferred subsets it resamples them, sampled rewards collide, and the group-normalized advantage vanishes. On genomic reasoning the fraction of questions yielding no reward signal rises from 0.2% under a uniform reference policy to 20.8% after GRPO training. As a remedy, we introduce FGPO (Full-Group Policy Optimization), which (1) scores every tool subset and optimizes the exact action expectation, so each update sees the complete action space, and (2) precomputes the reward of each question--subset pair into an exhaustive table, removing frozen-reasoner calls from the training loop entirely. Across five frozen reasoners and three genomic benchmarks, FGPO outperforms GRPO in all 15 settings by 6.75 points on average and up to 14.20, while a standard on-demand GRPO schedule would require 2.4 times as many frozen-reasoner reward evaluations and, on GenomeQA, FGPO cuts invoked tools per question from 2.36 to 1.40.
Chinese Translation
在静态推理器上进行强化学习已成为教导策略调用外部工具的常见方法。我们展示了在专业科学环境中，当完整的工具子集空间是可枚举时，这种方法在结构上变得不匹配。在这种情况下，一小组重复的计算能力覆盖了整个领域，因此工具子集的空间是组合性的，但足够小以便进行枚举，而 GRPO（组归一化策略优化）仍然从少量采样的回合中估计动作期望。更糟糕的是，随着训练的成功，近似效果会恶化：当策略集中于优选子集时，它会重新采样这些子集，采样奖励发生冲突，组归一化优势消失。在基因组推理中，产生无奖励信号的问题比例从均匀参考策略下的 0.2% 上升到 GRPO 训练后的 20.8%。为了解决这个问题，我们引入了 FGPO（全组策略优化），它（1）对每个工具子集进行评分并优化精确的动作期望，因此每次更新都能看到完整的动作空间；（2）将每个问题-子集对的奖励预先计算到一个详尽的表格中，从而完全消除训练循环中的静态推理器调用。在五个静态推理器和三个基因组基准测试中，FGPO 在所有 15 种设置中平均比 GRPO 提高了 6.75 分，最高可达 14.20，而标准的按需 GRPO 调度需要 2.4 倍的静态推理器奖励评估，并且在 GenomeQA 上，FGPO 将每个问题调用的工具数量从 2.36 降低到 1.40。
cs.AI / 45 / 2609.10263
What Should an Agent Forget? Separating What Is Stored from What Is Used
代理应该遗忘什么？将存储的内容与使用的内容分开
Abstract
Persistent language agents need stored experience to remain available across time, while each answer requires evidence suited to a particular question. A superseded fact can mislead a current-state answer and still be essential for a historical query. We present RD-Forget, a training-free framework that separates what an agent stores from what it uses. A retained source archive preserves observations, and a query-conditioned memory view controls their influence on the current answer. A frozen language-model curator extracts relevant evidence, groups facts into semantic slots, and preserves the relations needed for multi-hop reasoning. Same-slot replacement links suppress superseded values in current-state contexts, while intent-aware retrieval makes earlier evidence eligible again. A rate-distortion formulation guides construction of the answer-time view within a memory budget. Experiments span conversational memory, knowledge updating, fact consolidation, long-context reasoning, and personalization under a shared answering pipeline. The results associate accurate answers with both query-relevant evidence construction and control over obsolete alternatives. Configurations without forgetting or query conditioning have the largest score deficits, while slot grouping, historical access, and relation preservation contribute complementary functions. Retaining history while selectively controlling its use offers a practical way to accommodate changing facts and future questions.
Chinese Translation
持久的语言代理需要存储的经验以便在时间上保持可用，而每个答案都需要适合特定问题的证据。被取代的事实可能会误导当前状态的答案，但对于历史查询仍然至关重要。我们提出了 RD-Forget，这是一个无训练的框架，用于将代理存储的内容与其使用的内容分开。保留的源档案保存观察结果，而基于查询的记忆视图控制它们对当前答案的影响。一个冻结的语言模型策展人提取相关证据，将事实分组到语义槽中，并保留多跳推理所需的关系。同槽替换链接在当前状态上下文中抑制被取代的值，而意图感知检索使早期证据再次符合资格。一个率失真公式指导在内存预算内构建答案时间视图。实验涵盖了对话记忆、知识更新、事实整合、长上下文推理和在共享回答管道下的个性化。结果将准确答案与查询相关的证据构建和对过时替代品的控制联系在一起。没有遗忘或查询条件的配置得分差距最大，而槽分组、历史访问和关系保留则提供了互补功能。在选择性控制其使用的同时保留历史，为适应变化的事实和未来的问题提供了一种实用的方法。
cs.AI / 46 / 2609.10315
TRACE: Training Reasoning Agents for Causal Exploration with Synthesized Rewards
TRACE：使用合成奖励训练因果探索推理代理
Abstract
Reinforcement learning with verifiable rewards (RLVR) has advanced language-model reasoning in domains such as mathematics and code, where objective answers are inexpensive to check. Diagnostic reasoning over complex data lacks this advantage: establishing the true cause of an anomaly often requires costly expert investigation and may remain ambiguous after the fact. We ask whether this asymmetry of verification can instead be engineered. We sample an intervention, inject it into a controlled simulator, and generate the observations it would produce. The hidden intervention provides an oracle label and objective reward, while the agent must still investigate noisy, confounded, and distributed evidence. We instantiate this approach in TRACE, a digital-advertising diagnostic environment with 12 root causes and fine-grained segment attribution. Agents investigate each episode using Python and SQL and must identify both the root cause and, when applicable, the affected segment assignment. On a held-out 235-episode test set, the strongest prompted baseline, Claude Opus 5, reaches 0.686 FullAttr@1. Supervised fine-tuning raises Qwen3.5-35B-A3B from 0.159 to 0.637, and subsequent RL with synthesized rewards reaches 0.757, outperforming all evaluated prompted baselines, including frontier closed-source models and a prompted Qwen3.5-122B-A10B model. The resulting policy also uses substantially fewer tool calls than the prompted 35B base. These results provide evidence that access to a scalable, objective training signal can be a more important constraint than model scale alone. More broadly, simulation-based verification can make otherwise ambiguous diagnostic reasoning tasks amenable to scalable reinforcement learning.
Chinese Translation
具有可验证奖励的强化学习（RLVR）在数学和代码等领域推动了语言模型推理的发展，这些领域的客观答案检查成本较低。然而，复杂数据上的诊断推理缺乏这一优势：确定异常的真实原因通常需要昂贵的专家调查，并且事后可能仍然模糊不清。我们探讨是否可以工程化这种验证的不对称性。我们对一个干预进行采样，将其注入到一个受控的模拟器中，并生成其可能产生的观察结果。隐藏的干预提供了一个神谕标签和客观奖励，而代理仍需调查嘈杂、混淆和分布的证据。我们在TRACE中实现了这一方法，这是一个具有12个根本原因和细粒度段归因的数字广告诊断环境。代理使用Python和SQL调查每个事件，并必须识别根本原因以及在适用时受影响的段分配。在一个保留的235事件测试集中，最强的提示基线Claude Opus 5达到了0.686 FullAttr@1。监督微调将Qwen3.5-35B-A3B的性能从0.159提高到0.637，随后使用合成奖励的强化学习达到了0.757，超越了所有评估的提示基线，包括前沿的闭源模型和提示的Qwen3.5-122B-A10B模型。最终的策略还使用了比提示的35B基础模型显著更少的工具调用。这些结果提供了证据，表明获得可扩展的、客观的训练信号可能比模型规模本身更为重要的约束。更广泛地说，基于模拟的验证可以使原本模糊的诊断推理任务适应可扩展的强化学习。
cs.AI / 47 / 2609.10335
From Symbolic Perception to Logical Deduction: A Framework for Guiding Language Models in Geometric Reasoning
从符号感知到逻辑推理：引导语言模型进行几何推理的框架
Abstract
Plane geometry remains a significant challenge in AI, requiring the integration of visual perception and mathematical reasoning. While Large Multimodal Models (LMMs) naturally handle visuo-linguistic inputs, they are often computationally intensive and opaque. We demonstrate that a pure Large Language Model (LLM), when equipped with specialized modules, can rival state-of-the-art LMMs on complex geometry problems. Our framework integrates a Geometric Vision Parser, which translates diagrams into symbolic form, with a Symbolic Solver that performs formal deductions, thereby mitigating hallucinations and promoting interpretable reasoning. To enable rigorous evaluation, we curate a benchmark of challenging problems from the 2025 Chinese Zhongkao examinations, ensuring data novelty and testing deeper deductive skills. Experiments demonstrate that our approach achieves performance comparable to Gemini 2.5 Pro while delivering clearer, human-like solutions.
Chinese Translation
平面几何在人工智能领域仍然是一个重大挑战，需要视觉感知与数学推理的结合。尽管大型多模态模型（Large Multimodal Models, LMMs）能够自然处理视听语言输入，但它们通常计算密集且不透明。我们证明，当配备专门模块时，纯大型语言模型（Large Language Model, LLM）能够在复杂几何问题上与最先进的LMMs相媲美。我们的框架整合了几何视觉解析器（Geometric Vision Parser），将图示转换为符号形式，以及符号求解器（Symbolic Solver），执行形式推理，从而减轻幻觉现象并促进可解释的推理。为了实现严格的评估，我们策划了一套来自2025年中国中考的挑战性问题基准，确保数据的新颖性并测试更深层次的推理能力。实验表明，我们的方法在性能上可与Gemini 2.5 Pro相媲美，同时提供更清晰、类人化的解决方案。
cs.AI / 48 / 2609.10350
Cyber-Financial Contagion: Modeling the Propagation of an AI Vendor Compromise Through the Banking System
网络金融传染：建模人工智能供应商妥协在银行系统中的传播
Abstract
The banking system now depends on a small set of shared artificial intelligence vendors for fraud screening, credit decisioning, anti-money-laundering triage, customer analytics, and internal decision support. This paper studies how a compromise inside one of those vendors can propagate along a chain of operational, informational, and financial linkages until it triggers losses that look, from the outside, like a classical banking crisis. We build a four-layer heterogeneous network that couples AI vendors, financial institutions, interbank exposures, and customer accounts, and we propose CFC-Prop, a stochastic epidemic-and-clearing model that runs on that network. On a synthetic dataset with 60 vendors, 220 banks, roughly 2,500 vendor-bank service edges, and 1,400 interbank exposures, CFC-Prop reproduces the heavy-tailed loss distributions and the sharp dependence on patch latency that are consistent with prior cyber-financial evidence. We also train an early-warning model, CFC-GNN, that uses vendor-side incident telemetry and graph structure to flag high-cascade-risk vendors before impact. Across four baselines the proposed model reaches AUROC 0.82 and AUPRC 0.60 while keeping calibration errors bounded. We release the full code, synthetic data, and reproducible scripts. The results argue that cyber concentration among AI vendors is a first-order financial-stability problem and give supervisors a concrete quantitative tool for reasoning about it.
Chinese Translation
银行系统现在依赖于一小部分共享的人工智能供应商进行欺诈筛查、信用决策、反洗钱分流、客户分析和内部决策支持。本文研究了这些供应商中的一个发生妥协如何沿着操作、信息和金融链条传播，直到触发的损失在外部看起来像是经典的银行危机。我们构建了一个四层异构网络，连接了人工智能供应商、金融机构、银行间风险敞口和客户账户，并提出了CFC-Prop，这是一种在该网络上运行的随机流行病与清算模型。在一个包含60个供应商、220家银行、约2500个供应商-银行服务边缘和1400个银行间风险敞口的合成数据集上，CFC-Prop再现了与先前网络金融证据一致的重尾损失分布和对补丁延迟的强依赖性。我们还训练了一个早期预警模型CFC-GNN，该模型利用供应商侧事件遥测和图结构在影响发生前标记高级联风险的供应商。在四个基准测试中，所提出的模型达到了AUROC 0.82和AUPRC 0.60，同时保持了校准误差的界限。我们发布了完整的代码、合成数据和可重复的脚本。结果表明，人工智能供应商之间的网络集中是一个一阶金融稳定性问题，并为监管者提供了一个具体的定量工具来进行推理。
cs.AI / 49 / 2609.10413
Fortunate Recall: Ontology-Driven Memory Lifecycle Management for Persistent Coherence in LLMs
幸运回忆：基于本体的记忆生命周期管理以实现大型语言模型中的持久一致性
Abstract
Current LLM memory systems treat all personal facts identically, so stores grow without bound while retrieval precision degrades. The core challenge is lifecycle management: which memories should persist, which should be replaced, and at what rate, conditioned on the behavioral type of each fact. Fortunate Recall (FR) is a composable policy layer that classifies personal facts into a 10+1 behavioral ontology and applies category-specific lifecycle policies (differential temporal decay, slot-key supersession, event-time validity, and category-aware retrieval routing) as deterministic functions over LLM-extracted metadata. FR-Bank, our infrastructure-independent implementation, reaches a 76.9% pass rate on LifecycleBench, a new 516-question temporal-disambiguation benchmark, ahead of Mem0, A-MEM, Memory-R1, and MemoryOS (61% to 70.5%), and 75.2% on the full LongMemEval-S under the canonical Wu et al. judge protocol, so lifecycle policies impose no measurable cost on standard retrieval. A pre-registered ablation locates the gains: replacing the typed layer with three generic lifecycle primitives leaves correctness statistically unchanged (-1.7pp, 95% CI [-6.0, +2.7]), so the generic lifecycle metadata carries the correctness advantage, while the behavioral ontology carries calibration, halving downstream confabulation (12.0% vs 24.2%, p<0.001). End-to-end, FR-Bank cuts confabulation from Mem0's 45.1% to 22.4% over answered queries and from 32.2% to 13.0% over all queries while answering more of them correctly (31.2% vs 18.6%); the ranking replicates on the open-weight Kimi K2.5. The decomposition transfers to BEAM, an independently built benchmark: 46.8% correct vs Mem0's 32.9% over 280 questions, with the ontology's benefit concentrated in contradiction resolution and saturating near seven policy clusters. The ontology, benchmark, and code are released.
Chinese Translation
当前的大型语言模型（LLM）记忆系统对所有个人事实的处理是相同的，因此存储不断增长，而检索精度却在下降。核心挑战在于生命周期管理：哪些记忆应该持久化，哪些应该被替换，以及以何种速度进行替换，这取决于每个事实的行为类型。幸运回忆（Fortunate Recall, FR）是一个可组合的策略层，将个人事实分类为10+1种行为本体，并根据类别特定的生命周期策略（差异时间衰减、插槽键替代、事件时间有效性和类别感知检索路由）作为确定性函数应用于LLM提取的元数据。我们的基础设施无关实现FR-Bank在LifecycleBench上达到了76.9%的通过率，这是一个新的516题时间消歧义基准，领先于Mem0、A-MEM、Memory-R1和MemoryOS（61%到70.5%），并在标准的Wu等人评审协议下在完整的LongMemEval-S上达到了75.2%，因此生命周期策略对标准检索没有可测量的成本。一个预注册的消融实验找到了收益所在：用三个通用生命周期原语替换类型层，正确性在统计上没有变化（-1.7pp，95% CI [-6.0, +2.7]），因此通用生命周期元数据保持了正确性优势，而行为本体则提供了校准，减少了下游的虚构（12.0%对比24.2%，p<0.001）。整体而言，FR-Bank将Mem0的虚构率从45.1%降至22.4%（在回答过的查询中），并将所有查询的虚构率从32.2%降至13.0%，同时正确回答的比例增加（31.2%对比18.6%）；在开放权重的Kimi K2.5上，排名结果得以复制。该分解转移到BEAM，一个独立构建的基准：在280个问题上，正确率为46.8%，而Mem0为32.9%，本体的益处集中在矛盾解决上，并在接近七个策略集群时饱和。我们发布了本体、基准和代码。
cs.AI / 50 / 2609.10441
ConvMem: Convolutional Memory for Long-Context Reasoning
ConvMem：用于长上下文推理的卷积记忆
Abstract
While Large Language Models (LLMs) have demonstrated impressive capabilities, they often struggle with extremely long contexts due to fixed context limits. To address this, sequential approaches like MemAgent extend the effective context by reading text in segments and iteratively updating a fixed-size memory. However, this sequential paradigm suffers from high latency and requires costly reinforcement learning (RL) training, which can lead to overfitting on specific datasets. To overcome these limitations, we propose ConvMem, a training-free, highly parallelizable framework that reformulates long-context reasoning as a hierarchical convolution. Inspired by CNNs, ConvMem treats an LLM prompted with a specific query as a convolutional kernel. This kernel summarizes text segments hierarchically, shortening the reasoning path from a linear chain into a logarithmic tree. Specifically, ConvMem integrates \textit{Configurable Strides} and \textit{Skip Connections} to ensure robust evidence capture and propagation, while employing \textit{Multi-Kernel Convolution} to decompose complex queries into disentangled semantic channels. This design not only mitigates error accumulation but also enables massive parallelization across both text segments and reasoning threads. Experiments on RULER-HotpotQA and RULER-2WikiMultiHopQA demonstrate that ConvMem outperforms training-free baselines and avoids the risk of overfitting to parametric priors often observed in RL-trained models on out-of-distribution tasks.
Chinese Translation
尽管大型语言模型（LLMs）展现了令人印象深刻的能力，但由于固定的上下文限制，它们在处理极长上下文时常常面临挑战。为了解决这个问题，像MemAgent这样的顺序方法通过分段读取文本并迭代更新固定大小的记忆来扩展有效上下文。然而，这种顺序范式存在高延迟的问题，并且需要昂贵的强化学习（RL）训练，这可能导致在特定数据集上的过拟合。为了克服这些局限性，我们提出了ConvMem，一个无训练、可高度并行化的框架，将长上下文推理重新表述为层次卷积。受到卷积神经网络（CNNs）的启发，ConvMem将针对特定查询的LLM视为卷积核。该卷积核以层次方式总结文本片段，将推理路径从线性链缩短为对数树。具体而言，ConvMem集成了可配置步幅（Configurable Strides）和跳跃连接（Skip Connections），以确保稳健的证据捕获和传播，同时采用多核卷积（Multi-Kernel Convolution）将复杂查询分解为解耦的语义通道。这一设计不仅减轻了错误累积的影响，还实现了文本片段和推理线程之间的大规模并行化。在RULER-HotpotQA和RULER-2WikiMultiHopQA上的实验表明，ConvMem的表现优于无训练基线，并避免了在分布外任务中常见的对参数先验的过拟合风险，这种现象通常出现在经过RL训练的模型中。
cs.AI / 51 / 2609.10451
JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition
JarvisGUI：迈向具有动态任务组合的跨设备图形用户界面代理
Abstract
Real-world GUI usage frequently involves workflows that span multiple devices and platforms, requiring the transfer of intermediate results, maintenance of shared state, and coordination across heterogeneous environments. However, existing GUI benchmarks overwhelmingly evaluate agents on single-device, statically defined tasks, thus leaving such cross-device capabilities largely unexamined, resulting in an overly optimistic assessment of agents' readiness for real-world usage. We introduce JarvisGUI, a dynamic benchmark that evaluates GUI agents on cross-device workflows requiring coordinated interaction across heterogeneous platforms, including Android, Windows, and Ubuntu. Specifically, JarvisGUI formulates GUI tasks as input-output transformations under a lightweight type system, which allows us to automatically compose multi-step, cross-device workflows and dynamically evaluate agent performance within a unified framework. By evaluating agents in virtual environments spanning multiple operating systems, JarvisGUI reveals that state-of-the-art open-source GUI agents struggle with the state-transfer awareness, cross-platform contextual reasoning, and long-horizon dependency management required for real-world workflows, exposing a critical capability gap invisible to existing benchmarks.
Chinese Translation
现实世界中的图形用户界面（GUI）使用通常涉及跨多个设备和平台的工作流程，这需要中间结果的传递、共享状态的维护以及在异构环境中的协调。然而，现有的GUI基准测试主要在单设备、静态定义的任务上评估代理，这使得跨设备能力在很大程度上未得到检验，从而导致对代理在现实世界使用准备情况的过于乐观的评估。我们引入了JarvisGUI，这是一个动态基准测试，评估GUI代理在跨设备工作流程中的表现，这些工作流程需要在包括Android、Windows和Ubuntu在内的异构平台之间进行协调交互。具体而言，JarvisGUI将GUI任务形式化为轻量级类型系统下的输入-输出转换，这使我们能够自动组合多步骤的跨设备工作流程，并在统一框架内动态评估代理性能。通过在跨多个操作系统的虚拟环境中评估代理，JarvisGUI揭示了最先进的开源GUI代理在状态转移意识、跨平台上下文推理和长时间依赖管理方面面临的挑战，这些都是现实世界工作流程所需的能力，暴露了现有基准测试无法察觉的关键能力差距。
cs.CL / 1 / 2609.09166
X-CoSD: Communication-Efficient Cross-Vocabulary Collaborative Speculative Decoding
X-CoSD：通信高效的跨词汇协作推测解码
Abstract
This paper investigates collaborative speculative decoding (CoSD), a distributed large language model (LLM) inference framework in which an on-device small language model (SLM) drafts candidate tokens and a server LLM verifies them. Existing CoSD methods assume a shared vocabulary between the SLM and the LLM and incur substantial communication load because residual resampling requires token distribution exchange between the user device and the edge server. To address these limitations, we propose cross-vocabulary CoSD (X-CoSD), a lossless and communication-efficient CoSD framework for heterogeneous SLM-LLM vocabularies. X-CoSD is built on hybrid resampling (HR), which splits residual resampling across the common-vocabulary region on the device and the LLM-only region on the server, so that distribution transmission is required only for the common-vocabulary region. We further propose X-CoSD-E, an enhanced variant based on server resampling with device verification (SR-DV), in which the server sends only replacement candidates sampled from the server LLM and their corresponding probabilities for local verification at the device. We prove that both X-CoSD and X-CoSD-E preserve the server LLM distribution, and experiments show that they significantly improve token generation speed while maintaining generation quality comparable to that of the server LLM.
Chinese Translation
本文研究了协作推测解码（CoSD），这是一种分布式大语言模型（LLM）推理框架，其中设备上的小语言模型（SLM）草拟候选标记，而服务器LLM对其进行验证。现有的CoSD方法假设SLM和LLM之间共享词汇，这会导致显著的通信负担，因为残差重采样需要在用户设备和边缘服务器之间交换标记分布。为了解决这些限制，我们提出了跨词汇CoSD（X-CoSD），这是一种无损且通信高效的CoSD框架，适用于异构SLM-LLM词汇。X-CoSD建立在混合重采样（HR）之上，它将残差重采样分割为设备上的公共词汇区域和服务器上的LLM专用区域，因此仅在公共词汇区域需要进行分布传输。我们进一步提出了基于服务器重采样与设备验证（SR-DV）的增强变体X-CoSD-E，其中服务器仅发送从服务器LLM中采样的替换候选项及其对应的概率，以便在设备上进行本地验证。我们证明了X-CoSD和X-CoSD-E都保持了服务器LLM的分布，实验表明它们显著提高了标记生成速度，同时保持了与服务器LLM相当的生成质量。
cs.CL / 2 / 2609.09264
StochBench: A Domain-Specific Benchmark for Stochastic Processes in Lean
StochBench：一个针对随机过程的领域特定基准测试
Abstract
Leading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam, that poorly represent field-specific applications. We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at varying abstraction levels, each paired with its natural-language source. Addressing a field underrepresented in Mathlib, it covers finite and countable Markov chains, renewal processes, random walks, martingales, stopping times, queues, Brownian motion, stochastic calculus, weak convergence, and Poisson and continuous-time Markov processes. Our Opus 4.8-based agent achieves a 34.9% proof rate (157/450) under a 15-minute per-problem limit. StochBench better represents domain-specific applied mathematics while remaining challenging for advanced provers.
Chinese Translation
目前针对大型语言模型的形式定理证明的主要基准测试是从竞赛数学中提取的小型集合，例如国际数学奥林匹克（IMO）和普特南竞赛（Putnam），这些基准在特定领域应用方面表现不佳。我们介绍了StochBench，这是一个基于Lean 4的基准测试，包含450个不同抽象层次的研究生随机过程问题，每个问题都配有其自然语言来源。该基准测试涵盖了在Mathlib中代表性不足的领域，包括有限和可数的马尔可夫链、更新过程、随机游走、马尔可夫过程、停止时间、排队论、布朗运动、随机微积分、弱收敛以及泊松过程和连续时间马尔可夫过程。我们的基于Opus 4.8的代理在每个问题限制为15分钟的情况下，达到了34.9%的证明率（157/450）。StochBench更好地代表了领域特定的应用数学，同时对高级证明者仍然具有挑战性。
cs.CL / 3 / 2609.09338
Osprey: Target-agnostic Pre-training Makes Stronger Drafters in Speculative Decoding
Osprey：目标无关的预训练使得推测解码中的生成器更强大
Abstract
Speculative decoding is critical for accelerating LLM inference. However, the speedup is fragile: drafters are typically trained against a narrow distribution for a single target model, and their acceptance rate collapses under workload shifts. This is a striking inversion of modern LLM development, where target models are valued precisely for the broad generalization they acquire through large-scale pretraining. We argue that the natural remedy, pretraining, has been hard to apply to drafters because existing recipes are target-specific: the drafter consumes the target's hidden states and is distilled on the target's logits, so pretraining must be repeated for each target. We introduce Osprey, which instead bootstraps drafters from off-the-shelf pretrained small language models, treating broad pretraining as a reusable, target-agnostic asset and reducing per-target work to a lightweight adaptation step. Realizing this requires overcoming two challenges: small LMs are far deeper than a latency-bound drafter can afford, and their pretrained computation must remain intact while the drafter learns to ingest target hidden states and emit tokens in the target's vocabulary. Osprey addresses both by pruning to a shallow backbone, restoring its language-modeling capability with target-agnostic next-token pretraining, and adapting it to each target through vocabulary alignment, zero-initialized QKV expansion, and distillation from the target model's output distribution. Empirically, a single pretrained Osprey backbone transfers across targets and improves mean acceptance length by 16.1% for Qwen3-8B, 21.2% for Llama-3.3-70B-Instruct, and 22.7% for the 229B MiniMax-M2.5 (with 17.5% higher tokens per second), with the largest gains on out-of-domain and multilingual data. Our code is available at https://github.com/LeanModels/Osprey.
Chinese Translation
推测解码对于加速大型语言模型（LLM）的推理至关重要。然而，这种加速是脆弱的：生成器通常是在针对单一目标模型的狭窄分布下进行训练的，当工作负载发生变化时，其接受率会崩溃。这与现代LLM的发展形成了鲜明的对比，后者正是因为通过大规模预训练获得的广泛泛化能力而受到重视。我们认为，天然的解决方案——预训练，难以应用于生成器，因为现有的方法是针对特定目标的：生成器消耗目标的隐状态，并在目标的logits上进行蒸馏，因此必须为每个目标重复预训练。我们提出了Osprey，它从现成的预训练小型语言模型中引导生成器，将广泛的预训练视为可重用的、目标无关的资产，并将每个目标的工作减少到轻量级的适应步骤。实现这一目标需要克服两个挑战：小型语言模型的深度远超延迟受限的生成器所能承受的，同时在生成器学习摄取目标隐状态并在目标词汇中发出标记时，其预训练计算必须保持不变。Osprey通过修剪到一个浅层骨干网络来解决这两个问题，利用目标无关的下一个标记预训练恢复其语言建模能力，并通过词汇对齐、零初始化的QKV扩展以及从目标模型的输出分布进行蒸馏来适应每个目标。实证结果表明，单个预训练的Osprey骨干网络在不同目标之间迁移，并在Qwen3-8B上提高了16.1%的平均接受长度，在Llama-3.3-70B-Instruct上提高了21.2%，在229B MiniMax-M2.5上提高了22.7%（每秒标记数提高了17.5%），在域外和多语言数据上获得了最大的增益。我们的代码可在 https://github.com/LeanModels/Osprey 获取。
cs.CL / 4 / 2609.09349
SWORD: Wikidata-based Distortions Reveal Hidden Cross-Lingual Inconsistencies in LLM Factual Error Rejection
SWORD：基于Wikidata的扭曲揭示大型语言模型事实错误拒绝中的隐含跨语言不一致性
Abstract
Modern LLMs demonstrate impressive multilingual performance, yet standard benchmarks primarily reward selecting correct answers rather than evaluating genuine factual understanding. We introduce Systematic Wikidata-based Object-Relation Distortion (SWORD), a benchmark that evaluates whether models consistently reject factual errors across languages. SWORD generates syntactically well-formed but factually incorrect statements in eight widely spoken languages through controlled perturbations of Wikidata triples, ranging from random entity substitutions to semantically plausible property-based selections. Our distortion-based evaluation surfaces two critical insights that remain entirely obscured by conventional benchmarks. First, models counterintuitively achieve higher accuracy on semantically plausible distortions than on nonsensical random substitutions, suggesting reliance on distributional familiarity rather than genuine factual verification. Second, models exhibiting comparable baseline accuracy across languages show substantial performance degradation specifically on (East) Asian languages when presented with distorted statements, with cross-lingual performance gaps reaching up to 28 percentage points (49\% relative reduction) in some models. These findings demonstrate that multilingual factual reasoning involves asymmetric capabilities that aggregate accuracy metrics systematically obscure.
Chinese Translation
现代大型语言模型（LLMs）展现出令人印象深刻的多语言表现，然而标准基准主要奖励选择正确答案，而非评估真正的事实理解。我们引入了系统化的基于Wikidata的对象-关系扭曲（SWORD），这是一个评估模型是否在不同语言中一致拒绝事实错误的基准。SWORD通过对Wikidata三元组的控制扰动生成在八种广泛使用语言中语法上结构良好但事实错误的陈述，这些扰动包括随机实体替换和语义上合理的基于属性的选择。我们的基于扭曲的评估揭示了两个在传统基准下完全被掩盖的重要见解。首先，模型在语义上合理的扭曲上意外地实现了比在无意义的随机替换上更高的准确率，这表明模型依赖于分布熟悉性而非真正的事实验证。其次，尽管模型在不同语言中的基线准确率相当，但在面对扭曲陈述时，特别是在（东）亚语言上表现出显著的性能下降，跨语言性能差距在某些模型中达到28个百分点（相对减少49%）。这些发现表明，多语言事实推理涉及不对称的能力，而这些能力在系统性地掩盖准确性指标时被忽视。
cs.CL / 5 / 2609.09356
Auditable Emergency Triage for Maternal and Newborn Care in India
印度母婴护理的可审计紧急分诊
Abstract
At Noora Health, our nurses answer more than 50,000 medical queries per month on our WhatsApp-based service that provides caregivers with on-demand support. Their most time-critical task is emergency triage: deciding which queries need immediate in-person attention. To support them, we built a system that uses a large language model (LLM) to classify whether a message is an emergency and provide a rationale for interpretability. But the system was opaque: analyzing mistakes meant reading reasoning chains for each message, which is infeasible at our scale. Prompt changes meant re-running a full evaluation to prevent regressions, which was both costly and operationally challenging. Clinicians follow a decision tree to make this call, but it was never documented or passed to the model, which relied on a flat list of danger signs. To address these issues, we decomposed triage into two steps: an LLM extracts canonical symptoms and patient context from the query using a clinician-authored vocabulary, and a deterministic rule engine captures the scenarios that indicate an emergency. We show that the new system raised recall from 0.565 to 0.810 and F1 from 0.606 to 0.702, with structured rules driving most of the accuracy gains while the decomposition provides auditability: clinical experts can inspect each stage of the new system to see whether the query was mistranslated, symptoms were incorrectly extracted, patient context was wrongly inferred, or the necessary rules were missing. They can add new rules independently without causing regressions and avoid running costly evaluations. Since deployment, the new system has triaged 152,421 patient queries and flagged 28,535 (18.7%) as emergencies. The over-escalation rate has been 17.8%, without any increase in missed emergencies. Clinicians have also added 48 new rules since deployment, evidence of the faster correction loop we set out to build.
Chinese Translation
在Noora Health，我们的护士每月通过基于WhatsApp的服务回答超过50,000个医疗咨询，为护理人员提供按需支持。他们最紧迫的任务是紧急分诊：决定哪些咨询需要立即面对面的关注。为了支持他们，我们构建了一个系统，该系统使用大型语言模型（LLM）来分类消息是否为紧急情况，并提供可解释性的理由。然而，该系统缺乏透明度：分析错误意味着需要逐条阅读每条消息的推理链，这在我们的规模下是不可行的。提示的变化意味着需要重新进行全面评估以防止回归，这既昂贵又在操作上具有挑战性。临床医生遵循决策树来做出这一判断，但这一过程从未被记录或传递给模型，模型依赖于一份平面的危险信号列表。为了解决这些问题，我们将分诊分解为两个步骤：LLM使用临床医生撰写的词汇从咨询中提取典型症状和患者背景，而确定性规则引擎捕捉指示紧急情况的场景。我们展示了新系统将召回率从0.565提高到0.810，F1值从0.606提高到0.702，结构化规则推动了大部分准确性提升，而分解提供了可审计性：临床专家可以检查新系统的每个阶段，以查看咨询是否被错误翻译、症状是否被错误提取、患者背景是否被错误推断，或必要的规则是否缺失。他们可以独立添加新规则而不会导致回归，并避免进行昂贵的评估。自部署以来，新系统已对152,421个患者咨询进行了分诊，并将28,535个（18.7%）标记为紧急情况。过度升级率为17.8%，且未增加漏报的紧急情况。自部署以来，临床医生还添加了48条新规则，证明了我们旨在建立的更快的纠正循环。
cs.CL / 6 / 2609.09363
Do LLMs Make More Mistakes If They Do Not Believe the Input Data?
如果大型语言模型（LLMs）不相信输入数据，它们会犯更多错误吗？
Abstract
Large language models (LLMs) are prone to hallucinating or misinterpreting facts, which impairs their usability in retrieval-augmented generation or data-to-text systems. We analyse how faithfulness of LLMs to provided context depends on how plausible they perceive the context to be (context-memory conflict). To better identify error patterns, we make use of the increased difficulty of non-English and low-resource language text generation and input data based on local knowledge, only partially captured in models' parametric knowledge. We let the models generate text in English, Czech, Slovak and Upper Sorbian from factual (FA), counterfactual (CFA) and fictional (FI) RDF triples containing local Czech and Slovak data. Contrary to our expectations, we observe only a weak context-memory conflict on the human-annotated sample. For Kimi K3 as an LLM judge, which agrees well with human annotations on the sample, counterfactual inputs receive only slightly lower faithfulness scores than factual ones (-0.05 on a 1-5 scale). We also find that a suboptimal choice of LLM judge would lead to overestimating the strength of the context-memory conflict.
Chinese Translation
大型语言模型（LLMs）容易产生幻觉或误解事实，这影响了它们在检索增强生成或数据到文本系统中的可用性。我们分析了LLMs对提供的上下文的忠实度如何依赖于它们对上下文的可信度（上下文-记忆冲突）。为了更好地识别错误模式，我们利用了非英语和低资源语言文本生成的难度增加，以及基于地方知识的输入数据，这些知识在模型的参数知识中仅部分被捕获。我们让模型从包含当地捷克和斯洛伐克数据的事实（FA）、反事实（CFA）和虚构（FI）RDF三元组中生成英语、捷克语、斯洛伐克语和上索拉比亚语的文本。与我们的预期相反，我们在人工标注的样本中仅观察到微弱的上下文-记忆冲突。对于Kimi K3作为LLM评判者，它与样本上的人工标注高度一致，反事实输入的忠实度评分仅比事实输入稍低（在1-5的评分中为-0.05）。我们还发现，选择一个次优的LLM评判者会导致对上下文-记忆冲突强度的高估。
cs.CL / 7 / 2609.09410
Benchmarking Hybrid Deep Research Across Database Querying and Web Search
跨数据库查询和网络搜索的混合深度研究基准测试
Abstract
While autonomous agents have made significant strides in "deep research" by iteratively navigating the open web to synthesize information, real-world problem-solving is rarely confined to a single environment. Complex analytical tasks inherently require agents to weave together evidence from both ambiguous unstructured text (e.g., the open web) and highly precise structured data (e.g., relational databases). However, existing benchmarks evaluate these modalities in isolation, failing to capture the critical "handoff" - the ability to preserve constraints when moving evidence between systems. We introduce HybridDeepResearch, to our knowledge the first deep-research benchmark that requires both web search and SQL to form a complete, verifiable answer. The benchmark contains 380 tool-dependent tasks grounded in LiveSQLBench-Base-Lite databases and public web corpora, validated through automated checks and human review, and covering three reasoning patterns: SQL2S, S2SQL, and Parallel. Evaluations across proprietary and open-weight models under various agentic scaffolds reveal that even state-of-the-art models like GLM-5.2, Claude-Sonnet-4.6 and GPT-5 achieve only about 50-54% Pass@8 on the hard subset. Notably, results show that directional reasoning is substantially more difficult than parallel intersection, highlighting that bridging structured and unstructured information spaces without losing constraints remains a major open challenge for agentic systems. Code and datasets are publicly available at GitHub (https://github.com/Snowflake-AI-Research/HybridDeepResearch) and Hugging Face (https://huggingface.co/datasets/Snowflake/HybridDeepResearch).
Chinese Translation
虽然自主智能体在“深度研究”方面通过迭代导航开放网络以合成信息取得了显著进展，但现实世界的问题解决很少局限于单一环境。复杂的分析任务本质上要求智能体将来自模糊的非结构化文本（例如，开放网络）和高度精确的结构化数据（例如，关系数据库）的证据结合起来。然而，现有基准测试孤立地评估这些模式，未能捕捉到关键的“交接”——在系统之间移动证据时保持约束的能力。我们引入了HybridDeepResearch，据我们所知，这是第一个需要同时进行网络搜索和SQL查询以形成完整、可验证答案的深度研究基准。该基准包含380个依赖工具的任务，基于LiveSQLBench-Base-Lite数据库和公共网络语料库，通过自动检查和人工审核进行验证，并涵盖三种推理模式：SQL2S、S2SQL和并行推理。在各种智能体框架下对专有和开放权重模型的评估显示，即使是最先进的模型如GLM-5.2、Claude-Sonnet-4.6和GPT-5在困难子集上的Pass@8也仅达到约50-54%。值得注意的是，结果表明，定向推理显著比并行交集更具挑战性，这突显了在不失去约束的情况下连接结构化和非结构化信息空间仍然是智能系统面临的重大开放挑战。代码和数据集已在GitHub（https://github.com/Snowflake-AI-Research/HybridDeepResearch）和Hugging Face（https://huggingface.co/datasets/Snowflake/HybridDeepResearch）上公开发布。
cs.CL / 8 / 2609.09425
Edu-QuRating: Multi-Dimensional Educational Data Curation with Distilled Pairwise Judgements
Edu-QuRating：基于提炼的成对判断的多维教育数据策展
Abstract
Educational data filters have become a practical way to improve language-model pre-training, but most filters treat educational value as a single scalar property. This may be too broad for some applications, especially if the data set already features a high density of educational material. Useful learning material needs to be accurate, engaging, well structured, and appropriate for the intended audience and application (e.g. learner- vs teacher-facing). Following QuRating (Wettig et al. 2024), we introduce Edu-QuRating: a pipeline for multi-dimensional educational data scoring and curation. Edu-QuRating defines education-specific rubrics, uses an LLM judge to label sampled document pairs and distills those pairwise preferences into reusable Edu-QuRaters, which can score individual text chunks on a set of educational criteria. Across two sequence-classification base models and six educational criteria, the best Edu-QuRater recovers held-out GPT-4.1-mini pairwise judgements with mean accuracy 0.917. We then apply the resulting scorers in two applications. First, we investigate the potential of Edu-QuRaters for corpus filtering to improve pretraining of small language models. We scored 322.25M FineWeb-Edu-Fortified documents to obtain a filtered pre-training mixture. In matched single-run pre-training comparisons, models trained with Edu-QuRating-based mixtures reached higher observed aggregate accuracy across nine benchmarks than the FineWeb-Edu baseline, with gains concentrated in particular tasks. Second, we used Edu-QuRater scores as reward terms for GRPO post-training. In held-out pairwise judge evaluations, combining Edu-QuRater and answer-structure rewards produced responses preferred to the Qwen3-4B base model on both pedagogical quality and instruction following.
Chinese Translation
教育数据过滤器已成为改善语言模型预训练的实用方法，但大多数过滤器将教育价值视为单一的标量属性。这对于某些应用可能过于宽泛，尤其是当数据集已经具有高密度的教育材料时。有效的学习材料需要准确、有趣、结构良好，并且适合预期的受众和应用（例如，面向学习者或教师）。基于 QuRating (Wettig et al. 2024)，我们引入了 Edu-QuRating：一个用于多维教育数据评分和策展的流程。Edu-QuRating 定义了特定于教育的评分标准，使用 LLM judge 对采样的文档对进行标注，并将这些成对偏好提炼为可重用的 Edu-QuRaters，这些 Edu-QuRaters 可以根据一组教育标准对单个文本块进行评分。在两个序列分类基础模型和六个教育标准中，最佳的 Edu-QuRater 以平均准确率 0.917 恢复了保留的 GPT-4.1-mini 成对判断。随后，我们在两个应用中应用了这些评分器。首先，我们研究了 Edu-QuRaters 在语料过滤中的潜力，以改善小型语言模型的预训练。我们对 3.2225 亿个 FineWeb-Edu-Fortified 文档进行了评分，以获得过滤后的预训练混合物。在匹配的单次预训练比较中，使用基于 Edu-QuRating 的混合物训练的模型在九个基准测试中达到了比 FineWeb-Edu 基线更高的观察到的整体准确率，增益集中在特定任务上。其次，我们将 Edu-QuRater 评分作为 GRPO 后训练的奖励项。在保留的成对评估中，结合 Edu-QuRater 和答案结构奖励的响应在教学质量和遵循指令方面优于 Qwen3-4B 基础模型。
cs.CL / 9 / 2609.09496
The Mutations of Machine Speech
机器言语的变异
Abstract
Algorithmic outputs now populate the digital environments through which contemporary life is organized. The role of law in facilitating and constituting (rather than merely responding to) these processes is gaining increasing traction across scholarly accounts. This inquiry traces the evolution of algorithmic outputs attending to their legal underpinnings and social implications, surfacing the mutations of machine speech. The first mutation redefined speech as data to be queried: search engines transformed the web from a space of information retrieval into an economic regime of algorithmic visibility. The second mutation reframed speech as engagement: social media platforms fused moderation with amplification, turning expression into a metric of attention, governed by corporate architectures. The third mutation emerges in conversational systems and interfaces, where generative text displaces information retrieval, bringing with it dense technolegal entanglements and profound epistemic consequences. Scholars of freedom of expression, informational privacy, and communication studies have long grappled with these dynamics, yet their implications for broader legal thought have also become urgent. This piece seeks to organize and clarify the evolving debate around algorithmic speech, making this critical but often fragmented discourse more accessible to wider legal and interdisciplinary audiences. In doing so, it bridges the gap between observing technological transformation and critically assessing the constitutive role of law within it, offering a conceptual resource for researchers, students, policymakers, and practitioners navigating and contesting this evolving landscape.
Chinese Translation
算法输出如今充斥着当代生活所依赖的数字环境。法律在促进和构成这些过程中的角色（而不仅仅是对其反应）在学术界越来越受到关注。本研究追溯了算法输出的演变，关注其法律基础和社会影响，揭示了机器言语的变异。第一次变异重新定义了言语为可查询的数据：搜索引擎将网络从信息检索的空间转变为算法可见性的经济体制。第二次变异将言语重新框定为参与：社交媒体平台将内容管理与放大结合在一起，将表达转变为一种注意力的度量，由企业架构所主导。第三次变异出现在对话系统和界面中，生成文本取代了信息检索，带来了复杂的技术法律纠缠和深刻的认识论后果。言论自由、信息隐私和传播学的学者们长期以来一直在探讨这些动态，但其对更广泛法律思想的影响也变得紧迫。本文旨在组织和澄清围绕算法言语的不断演变的辩论，使这一关键但常常支离破碎的论述对更广泛的法律和跨学科受众更具可及性。在此过程中，它弥合了观察技术变革与批判性评估法律在其中构成角色之间的鸿沟，为研究人员、学生、政策制定者和从业者在这一不断演变的领域中导航和争论提供了概念资源。
cs.CL / 10 / 2609.09552
TEFM: Token-Efficient Faithful Modeling for Structured Data
TEFM：面向结构化数据的高效可信建模
Abstract
In this paper, we solve two fundamental obstacles in applying LLMs to critical domains: token efficiency and faithfulness. To address both constraints jointly, we present TEFM (Token-Efficient Faithful Modeling), a framework designed for structured data analysis in critical domains. TEFM achieves token efficiency by compressing lengthy structured observations into compact Behavioral Code tokens, dramatically reducing token consumption with minimal information loss. Moreover, TEFM enables faithful rationalization through a dual-fidelity objective that jointly optimizes code-level reconstruction and prediction-level fidelity, identifying minimal sufficient feature subsets grounded in input data. Comprehensive experiments across various domain datasets and model backbones (Qwen3, Gemma-2, Phi-4) show that TEFM achieves competitive classification accuracy with dramatic token reduction (approximately 1\% token retention in clinical and 2\% in security domains) while producing faithful rationales.
Chinese Translation
在本文中，我们解决了将大型语言模型（LLMs）应用于关键领域的两个基本障碍：令牌效率和可信性。为共同应对这两个限制，我们提出了TEFM（高效可信建模），这是一个专为关键领域的结构化数据分析设计的框架。TEFM通过将冗长的结构化观察压缩为紧凑的行为代码令牌，从而实现了令牌效率，显著减少了令牌消耗，同时信息损失最小。此外，TEFM通过双重可信性目标实现了可信的推理，该目标共同优化代码级重建和预测级可信性，识别基于输入数据的最小充分特征子集。针对不同领域数据集和模型骨干（Qwen3、Gemma-2、Phi-4）的全面实验表明，TEFM在显著减少令牌的同时（临床领域约保留1%的令牌，安全领域约保留2%），实现了具有竞争力的分类准确性，并生成了可信的推理。
cs.CL / 11 / 2609.09554
BuzzASR: A Swarm of 100+ Monolingual Speech Recognition Models
BuzzASR：一个由100多个单语语音识别模型组成的群体
Abstract
We introduce BuzzASR, a collection of language-specialized fine-tuned Whisper models adapted for automatic speech recognition (ASR) in 102 languages. Large end-to-end Transformer-based ASR models such as Whisper have revolutionized ASR, but most prominent models are highly multilingual. As a result, these models often perform poorly on languages less well-represented in their training set. While it has long been known that effective language adaptation can be achieved through simple fine-tuning on monolingual data, this strategy has only been applied to a small number of languages. We massively scale up this simple approach to 102 languages covered in the FLEURS dataset, while also implementing a more complex language adaptation strategy that integrates monolingual tokenizer replacement and data augmentation using text-only fine-tuning. BuzzASR models outperform Whisper-large-v3 on 77 out of 102 languages, reducing character error rates (CER) by a factor of over 2.8 on average. Our models achieve state-of-the-art CER among open-source systems on 27 of 102 languages on the combined FLEURS and Common Voice test set. Our tokenizer replacement strategy yields an average 3.3x improvement in compression rate (characters per token) over Whisper's multilingual BPE, with gains of up to 21.7x. We release all models, code, and detailed results: https://lemn-lab.github.io/buzz-asr
Chinese Translation
我们介绍了BuzzASR，这是一个针对102种语言的语言专用微调Whisper模型的集合，旨在用于自动语音识别（ASR）。基于大规模端到端Transformer的ASR模型，如Whisper，已经彻底改变了ASR领域，但大多数知名模型都是高度多语种的。因此，这些模型在训练集中代表性不足的语言上通常表现不佳。虽然早已知道通过在单语数据上进行简单微调可以实现有效的语言适应，但这一策略仅应用于少数语言。我们将这种简单的方法大规模扩展到FLEURS数据集中覆盖的102种语言，同时实施一种更复杂的语言适应策略，该策略结合了单语分词器替换和仅文本微调的数据增强。BuzzASR模型在102种语言中有77种超越了Whisper-large-v3，平均减少字符错误率（CER）超过2.8倍。我们的模型在FLEURS和Common Voice测试集的102种语言中，有27种达到了开源系统中的最先进CER。我们的分词器替换策略在压缩率（每个标记的字符数）上比Whisper的多语种BPE平均提高了3.3倍，最高可达21.7倍。我们发布了所有模型、代码和详细结果：https://lemn-lab.github.io/buzz-asr
cs.CL / 12 / 2609.09561
Towards Automatic Evolution Tree Generation from Citation Graphs
基于引用图的自动演化树生成研究
Abstract
Surveys remain the primary way researchers grasp the lineage of methods within an AI subfield, but they scale poorly against the current rate of publication. Existing taxonomy-induction methods are largely leaf-bound and time-agnostic; they tend to force transitional papers into mature leaves and can create topological inversions between ancestors and descendants. We propose EvoTree, a staged framework that decouples conceptual backbone learning from temporal refinement: a graph-aware encoder with distribution-based hierarchical clustering yields a stable taxonomy backbone; temporal fine-tuning then re-attaches marginal papers to internal nodes under monotonic-path constraints; a final LLM pass labels concepts without altering the topology. We release the first annotated benchmark for this task across 11 AI subfields. EvoTree attains the highest NMI and citation-direction accuracy among all baselines and the best concept purity on the annotated benchmark, and is the only method with non-trivial marginal-paper detection on the annotated set.
Chinese Translation
文献综述仍然是研究人员理解人工智能子领域内方法沿革的主要方式，但其在当前出版速度下的扩展性较差。现有的分类法诱导方法主要局限于叶节点，并且对时间不敏感；它们往往将过渡性论文强行归入成熟叶节点，并可能导致祖先与后代之间的拓扑反转。我们提出了EvoTree，一个分阶段的框架，将概念骨架学习与时间细化解耦：一个图感知编码器结合基于分布的层次聚类生成稳定的分类法骨架；随后，时间微调在单调路径约束下将边缘论文重新附加到内部节点；最后，经过一次大型语言模型（LLM）处理，标记概念而不改变拓扑结构。我们发布了该任务在11个人工智能子领域的第一个标注基准。EvoTree在所有基线中获得了最高的归一化互信息（NMI）和引用方向准确率，并在标注基准上实现了最佳概念纯度，是唯一在标注集上具有非平凡边缘论文检测能力的方法。
cs.CL / 13 / 2609.09569
Reproducing Omitted Temporal Expressions in Japanese News for Retrieval-Augmented Applications
在日本新闻中再现省略的时间表达以用于检索增强应用
Abstract
News articles often contain omitted temporal expressions, such as day-only or month-only mentions, which must be interpreted with reference to the publication date. When such articles are indexed or processed as standalone text in search and retrieval-augmented generation (RAG) systems, these omissions can cause temporal mismatches and unstable interpretation by large language models. We focus on reproducing omitted temporal expressions as concrete dates or intervals using the publication date as external context before the articles are indexed for search and RAG applications. Specifically, building on established temporal-expression extraction and normalization techniques and informed by a manual analysis of Japanese news articles, we propose jaROTE, a rule-based pipeline for Japanese news. Experiments on two news corpora demonstrate that jaROTE achieves high performance, and remains competitive with LLMs while providing a fast, low-cost pipeline. We further show that temporal reproduction improves time-constrained lexical retrieval, demonstrating the practical value of publication-date-grounded normalization for Japanese news retrieval.
Chinese Translation
新闻文章中常常包含省略的时间表达，例如仅提及日期或月份，这些表达必须参考发布日期进行解释。当这些文章被索引或作为独立文本处理在搜索和检索增强生成（RAG）系统中时，这些省略可能导致时间不匹配和大型语言模型的不稳定解释。我们专注于在文章被索引用于搜索和RAG应用之前，利用发布日期作为外部上下文，将省略的时间表达再现为具体的日期或时间区间。具体而言，基于已建立的时间表达提取和规范化技术，并结合对日本新闻文章的手动分析，我们提出了jaROTE，一个针对日本新闻的基于规则的处理管道。对两个新闻语料库的实验表明，jaROTE实现了高性能，并在提供快速、低成本管道的同时，与大型语言模型保持竞争力。我们进一步展示了时间再现改善了时间限制的词汇检索，证明了基于发布日期的规范化在日本新闻检索中的实际价值。
cs.CL / 14 / 2609.09575
Beyond Top Words: MonoTM for Topic Modeling with Interpretable Monosemantic Features
超越关键词：具有可解释单义特征的主题建模框架MonoTM
Abstract
Topic models summarize large text corpora, but top-ranked words often provide only a limited representation of topic semantics. Sparse autoencoders (SAEs) offer a way to move beyond word-level descriptors by extracting interpretable features from dense representations, yet how feature interpretability relates to topic-inference quality remains unclear. We introduce \textbf{MonoTM}, an interpretable topic modeling framework that decouples these roles. Across three benchmark corpora, we show that document--topic mixture estimation and semantic interpretation favor different SAE configurations and feature subsets. MonoTM estimates mixtures from the full SAE bag-of-features representation and, with them fixed, learns topic descriptors over a separate vocabulary of corpus-grounded semantic features. This design preserves global topic structure while representing topics with semantic units more meaningful than individual words, making them more useful for downstream corpus analysis.
Chinese Translation
主题模型用于总结大规模文本语料库，但排名靠前的词汇往往只能提供有限的主题语义表示。稀疏自编码器（Sparse Autoencoders, SAEs）通过从密集表示中提取可解释特征，为超越基于词汇的描述提供了一种方法，但特征可解释性与主题推断质量之间的关系仍不明确。我们提出了 extbf{MonoTM}，一个可解释的主题建模框架，解耦了这两种角色。在三个基准语料库上，我们展示了文档-主题混合估计和语义解释偏好不同的SAE配置和特征子集。MonoTM从完整的SAE特征包表示中估计混合，并在固定这些特征的情况下，学习基于语料库的语义特征的单独词汇上的主题描述。该设计在保留全局主题结构的同时，用比单个词更有意义的语义单元表示主题，使其在后续的语料库分析中更具实用性。
cs.CL / 15 / 2609.09672
SEA-SpeechBench: A Large-Scale Multitask Benchmark for Speech Understanding Across Southeast Asia
SEA-SpeechBench：一个针对东南亚语音理解的大规模多任务基准
Abstract
The rapid advancement of audio and multimodal large language models has unlocked transformative speech understanding capabilities, yet evaluation frameworks remain predominantly English-centric, leaving Southeast Asian (SEA) languages critically underrepresented. We introduce SEA-SpeechBench, to the best of our knowledge, the first large-scale multitask benchmark that evaluates speech understanding in 11 SEA languages through 97,194 samples across 99 evaluation sets and 597 hours of curated audio data. Our benchmark comprises 9 diverse tasks across 3 categories: speech processing (automatic speech recognition, speech translation, spoken question answering), paralinguistic analysis (emotion, gender, age, speaker recognition), and temporal understanding, a novel dimension featuring timestamped content queries and temporal localization within extended audio sequences up to 3 minutes. We implement multilingual prompting in both native SEA languages and English to reflect user interactions with audio-language models. Evaluation of leading open-source and proprietary systems reveals marked performance gaps. Across all models, performance remains underwhelming on temporal understanding, emotion recognition, and speech translation. Prompting in low-resource languages such as Burmese and Tamil lags behind English by up to 41 percentage points. Our findings expose critical model limitations and underscore the need for inclusive model development. The SEA-SpeechBench benchmark is available at https://zwenyu.github.io/SEA-SpeechBench/.
Chinese Translation
音频和多模态大语言模型的快速发展解锁了变革性的语音理解能力，但评估框架仍以英语为中心，导致东南亚（SEA）语言严重缺乏代表性。我们介绍了SEA-SpeechBench，尽我们所知，这是第一个大规模多任务基准，评估11种东南亚语言的语音理解，涵盖97,194个样本、99个评估集和597小时的精心策划音频数据。我们的基准包含3个类别的9个多样化任务：语音处理（自动语音识别、语音翻译、口语问答）、副语言分析（情感、性别、年龄、说话人识别）和时间理解，这是一个新颖的维度，涉及带时间戳的内容查询和在最长达3分钟的扩展音频序列中的时间定位。我们在本地东南亚语言和英语中实施多语言提示，以反映用户与音频语言模型的互动。对领先的开源和专有系统的评估揭示了明显的性能差距。在所有模型中，时间理解、情感识别和语音翻译的表现均不尽如人意。在低资源语言如缅甸语和泰米尔语中，提示的表现比英语低41个百分点。我们的研究揭示了模型的关键局限性，并强调了包容性模型开发的必要性。SEA-SpeechBench基准可在https://zwenyu.github.io/SEA-SpeechBench/获取。
cs.CL / 16 / 2609.09677
X2-NativeCursor: Native-Token Text Progress Tracking for Incremental-Text Streaming Codec TTS
X2-NativeCursor：增量文本流编码器 TTS 的原生标记文本进度跟踪
Abstract
Incremental-text streaming text-to-speech (TTS) needs online text progress tracking for synchronized highlighting, interruption handling, and dialogue-history updates. Input text arrives before it is spoken, so text arrival alone cannot indicate speech progress. Existing waveform-based alignment requires complete audio or adds acoustic processing during streaming. We propose X2-NativeCursor, a lightweight observer that tracks progress from native speech tokens before waveform decoding without changing the TTS generator. Its normalization plan links spoken labels to their original-text spans. Text and native-token encoders feed a local matcher that estimates the current label position. A separate output rule converts revisable position estimates into a cursor that never moves backward. Mean absolute error against an automatic reference is 0.151 Chinese characters with 80-ms lookahead, versus 1.253 characters with 320-ms lookahead for an online waveform baseline. Alignment real-time factor also decreases from 0.3598 to 0.0180 relative to this baseline. Lower tracking error is retained under a second automatic alignment reference. We evaluate X2-NativeCursor on Qwen3-TTS and validate its adaptation to CosyVoice2 by training a separate observer for each backbone. Code is publicly available at https://github.com/X-Square-Robot/X2Streaming-TTS.
Chinese Translation
增量文本流文本到语音（TTS）需要在线文本进度跟踪，以实现同步高亮、处理中断和更新对话历史。输入文本在被朗读之前到达，因此仅凭文本到达无法指示语音进度。现有的基于波形的对齐方法需要完整的音频或在流媒体过程中增加声学处理。我们提出了 X2-NativeCursor，这是一种轻量级观察者，在波形解码之前跟踪原生语音标记的进度，而无需更改 TTS 生成器。其归一化方案将朗读标签与其原始文本范围关联。文本和原生标记编码器为本地匹配器提供输入，该匹配器估计当前标签位置。一个单独的输出规则将可修正的位置估计转换为一个永不向后移动的光标。与自动参考的平均绝对误差为 0.151 个汉字，80 毫秒的前瞻时间，而在线波形基线的前瞻时间为 320 毫秒时为 1.253 个汉字。相对于该基线，对齐实时因子也从 0.3598 降至 0.0180。在第二个自动对齐参考下，较低的跟踪误差仍然保持。我们在 Qwen3-TTS 上评估了 X2-NativeCursor，并通过为每个主干训练一个单独的观察者验证其对 CosyVoice2 的适应性。代码可在 https://github.com/X-Square-Robot/X2Streaming-TTS 上公开获取。
cs.CL / 17 / 2609.09684
Which Medical Questions Deserve Rationales? Perturbation-Sensitive Selection for Robust QA
哪些医学问题值得提供理由？针对稳健问答的扰动敏感选择
Abstract
Medical question-answering datasets often contain answer labels, whereas high-quality rationales remain scarce, noisy, or costly to validate. This changes the acquisition question: rather than asking which questions should be labeled, we ask which already-labeled questions should receive rationale supervision under a fixed token budget. We study an offline version of this problem in which candidate rationales are visible to the selector but withheld from downstream training unless selected. We propose root-mean-square Robustness-based Sample Prioritization (RMS-RSP), which perturbs hidden states only at rationale tokens and measures the resulting shift in the gold-versus-best-distractor margin. Across five medical QA datasets, MedGemma-4B-IT, three training seeds, ten budgeted non-RSP selectors, and an unbudgeted full-supervision reference, RMS-RSP provides a deliberately qualified result. Its locked-budget accuracy is 60.61% on average versus 60.08% for Random, with a statistically resolved gain only on AfriMed-QA (+1.44 points). Its full-budget accuracy area is not better than Random. However, after three answer-option reorderings, RMS-RSP improves robust accuracy and semantic consistency by 1.91 and 2.85 points on average, respectively, with the same direction on all five datasets. Training on every pool rationale raises macro accuracy to 63.74%, but consumes 29--254 times more rationale tokens and does not uniformly improve robustness. These findings do not establish universal accuracy gains; they instead suggest that rationale-local boundary sensitivity can identify supervision that improves invariance to semantically equivalent formatting changes.
Chinese Translation
医学问答数据集通常包含答案标签，而高质量的理由却稀缺、嘈杂或验证成本高。这改变了获取问题的方式：我们不再询问哪些问题应该被标记，而是询问在固定的标记预算下，哪些已经标记的问题应该接受理由监督。我们研究了这个问题的离线版本，其中候选理由对选择者可见，但在未被选择的情况下不会用于下游训练。我们提出了基于均方根稳健性样本优先选择（RMS-RSP），该方法仅对理由标记扰动隐藏状态，并测量金标准与最佳干扰项边际的变化。在五个医学问答数据集（MedGemma-4B-IT）、三个训练种子、十个预算非RSP选择器和一个无预算全监督参考中，RMS-RSP提供了一个经过深思熟虑的结果。其锁定预算的准确率平均为60.61%，而随机选择为60.08%，在AfriMed-QA上仅有统计学上显著的增益（+1.44点）。其全预算准确率区域不优于随机选择。然而，在经过三次答案选项重新排序后，RMS-RSP在稳健准确率和语义一致性上分别平均提高了1.91和2.85点，在所有五个数据集上均呈现相同的方向。对每个池中理由进行训练将宏观准确率提高至63.74%，但消耗的理由标记数量是29至254倍，并且并未均匀改善稳健性。这些发现并未确立普遍的准确率提升；相反，它们表明理由局部边界敏感性可以识别出改善对语义等效格式变化的不变性的监督。
cs.CL / 18 / 2609.09691
Looped GPT-BERT: Trading Parameters for Computation in Small Language Modeling
循环 GPT-BERT：在小型语言建模中权衡参数与计算
Abstract
When training data are limited, increasing parameter count is not the only way to improve language-model performance. A small parameter set, when repeatedly applied, can also deliver comparable performance. We study Looped GPT-BERT in the BabyLM 2026 Strict-small setting, combining GPT-BERT's masked next-token and causal language-modeling objectives with depth-wise parameter sharing. We train on a preprocessed 7.48M-word English corpus and compare objective ratios, non-looped and looped architectures, and loop counts. Our final $4\times12$ model uses four physical layers for twelve recurrent traversals and contains 12.18M parameters. The BabyLM 2026 leaderboard reports an Overall Average of 35.42 and an NLP Average of 48.48. Compared with public BabyLM 10M Strict-small GPT-2 and GPT-BERT baselines, it achieves comparable performance on selected linguistic and downstream metrics, including BLiMP and GLUE, with fewer parameters. The loop ablations show that additional recurrent computation can improve training and preserve strong performance on selected linguistic tasks, whereas poorer performance on other tasks may reveal an inherent limitation of the looped design: using only a few physical layers restricts the model's representational space.
Chinese Translation
在训练数据有限的情况下，增加参数数量并不是提高语言模型性能的唯一方法。通过重复应用小参数集，也可以实现可比的性能。我们在 BabyLM 2026 严格小型设置中研究了循环 GPT-BERT，将 GPT-BERT 的掩码下一个标记和因果语言建模目标与深度参数共享相结合。我们在一个经过预处理的 748 万词的英语语料库上进行训练，并比较了目标比率、非循环和循环架构以及循环次数。我们的最终 $4 imes12$ 模型使用四个物理层进行十二次递归遍历，包含 1218 万个参数。BabyLM 2026 排行榜报告了整体平均值 35.42 和 NLP 平均值 48.48。与公共的 BabyLM 1000 万严格小型 GPT-2 和 GPT-BERT 基准相比，它在选定的语言学和下游指标（包括 BLiMP 和 GLUE）上实现了可比的性能，同时参数更少。循环消融实验表明，额外的递归计算可以改善训练并保持在选定语言任务上的强性能，而在其他任务上的较差表现可能揭示了循环设计的固有限制：仅使用少量物理层限制了模型的表征空间。
cs.CL / 19 / 2609.09696
When Auditors Fabricate: Batch-Size Degradation and Confident Hallucination in LLM Detection of Planted Document Contamination
审计师伪造时：批量规模退化与大型语言模型在植入文档污染检测中的自信幻觉
Abstract
Large language models are increasingly proposed as automated auditors of document quality, yet their reliability as detectors of planted errors is poorly characterised. We construct a contaminated corpus of 150 academic papers spanning supply chain management and medical research, injecting 450 known contaminants of three types: typographical corruption, semantic reversal, and absurd out-of-context insertion. We then evaluate Google Gemini 3.0 Pro's ability to recover a 180-contaminant answer-key subset across 60 documents under three prompting regimes of increasing scale: single document, small batch, and large batch. Detection holds at small scale and then collapses: 50% recovery on single documents, 60% on small batches, and 2.8% on large batches. The failure mode at scale is not abstention but fabrication. Rather than reporting incomplete processing, the model produced confident findings including invented contaminants of its own, absurdities such as "telepathic squirrel" and "quantum-powered toaster" that mimic the style of the planted material but do not appear in any document. Detection also varies by contamination type: absurd insertions were recovered at 75% in completed evaluations, while semantic reversals and typographical corruptions were each recovered at only 50%. The corruptions most likely to occur in the wild, plausible ones, are the ones most often missed. We conclude that LLM document auditing degrades not gracefully but deceptively, and outline the harness such systems require: bounded batch sizes, direct content injection, and mechanical verification of every reported finding against source text.
Chinese Translation
大型语言模型越来越多地被提议作为文档质量的自动审计工具，但它们作为植入错误检测器的可靠性尚未得到充分表征。我们构建了一个包含150篇学术论文的污染语料库，涵盖供应链管理和医学研究，注入了450个已知的三种类型污染物：排版错误、语义颠倒和荒谬的上下文插入。然后，我们评估了Google Gemini 3.0 Pro在三种逐步扩大的提示模式下，从60篇文档中恢复180个污染物答案关键子集的能力：单文档、小批量和大批量。在小规模下检测保持有效，但随后崩溃：单文档的恢复率为50%，小批量为60%，而大批量仅为2.8%。在大规模下的失败模式不是放弃而是伪造。模型并未报告处理不完整，而是产生了自信的结果，包括自己虚构的污染物，以及诸如“心灵感应松鼠”和“量子驱动烤面包机”等荒谬内容，这些内容模仿了植入材料的风格，但并未出现在任何文档中。检测的效果也因污染类型而异：在完成的评估中，荒谬插入的恢复率为75%，而语义颠倒和排版错误的恢复率各为50%。在现实中最可能发生的、看似合理的污染物，往往是被遗漏的。我们得出结论，LLM文档审计的退化并非优雅，而是具有欺骗性，并概述了此类系统所需的约束：有限的批量规模、直接内容注入，以及对每个报告结果与源文本进行机械验证。
cs.CL / 20 / 2609.09716
Scaling E-Commerce Attribute Extraction with Parallel Decoding
通过并行解码扩展电子商务属性提取
Abstract
Customers rely on specific product attributes to compare products and make purchasing decisions, but e-commerce catalogs are messy and unstructured, making it difficult to identify which attributes matter most and extract them at scale. Standard Attribute Value Extraction (AVE) systems treat all attributes equally, producing large, inconsistent attribute sets that do not reflect the factors consumers use to differentiate products. We introduce a two-stage LLM pipeline that first discovers a compact, ranked schema of purchase-discriminative attributes for each product category, then extracts their values from catalog text using a fine-tuned compact LLM (Qwen3-4B) with Hyper-Parallel Decoding (HPD). This pipeline achieves 85% extraction accuracy, on par with the foundational LLM it was distilled from, while reducing inference costs by 92% over foundational LLMs, enabling production-scale use for product discovery and catalog enrichment. The resulting category-level structured representations effectively constitute automatically constructed product knowledge bases, providing consistent, comparable attributes across varied product categories that can ground downstream knowledge-intensive applications.
Chinese Translation
客户依赖特定的产品属性来比较产品并做出购买决策，但电子商务目录杂乱无章，难以识别哪些属性最为重要并进行大规模提取。标准的属性值提取（Attribute Value Extraction, AVE）系统对所有属性一视同仁，产生的大量不一致的属性集合并未反映消费者用于区分产品的因素。我们提出了一种两阶段的大型语言模型（LLM）管道，首先为每个产品类别发现一个紧凑的、按重要性排序的购买区分属性架构，然后利用经过微调的紧凑型 LLM（Qwen3-4B）和超并行解码（Hyper-Parallel Decoding, HPD）从目录文本中提取这些属性的值。该管道实现了85%的提取准确率，与其提炼的基础 LLM 相当，同时将推理成本降低了92%，使其能够在生产规模上用于产品发现和目录丰富。最终生成的类别级结构化表示有效构成了自动构建的产品知识库，提供了一致、可比较的属性，适用于不同产品类别，从而为下游知识密集型应用奠定基础。
cs.CL / 21 / 2609.09719
StreamAlign: Streaming Text-Aligned Speech Tokenization
StreamAlign：流式文本对齐语音标记化
Abstract
Text-aligned speech tokenization methods have emerged to better align speech tokens with LLM token spaces, enabling more effective utilization of pretrained LLMs. However, they rely on offline automatic speech recognition (ASR), leading to two key limitations: (i) the need for complete utterances before tokenization, precluding real-time streaming, and (ii) vocabulary mismatch between ASR and LLMs, which reduces acoustic granularity from the subword to the word level. We introduce StreamAlign, a text-aligned speech tokenization framework that enables streaming tokenization for real-time speech-text joint modeling. StreamAlign performs online speech-text alignment by combining character-level RNN-Transducer alignment with word-level ASR guidance, mitigating ASR-LLM vocabulary mismatch while preserving recognition accuracy. A proactive word boundary classifier anticipates word completion at chunk boundaries, reducing tokenization latency from 560 ms to 270 ms. On LibriSpeech, StreamAlign achieves the lowest WER and highest UTMOS among evaluated tokenizers. Furthermore, StreamAlign-SLM, a spoken language model trained on StreamAlign units, outperforms other end-to-end spoken language models in speech continuation while achieving the strongest overall consistency on SALMon and spoken StoryCloze.
Chinese Translation
文本对齐的语音标记化方法已经出现，以更好地将语音标记与大规模语言模型（LLM）标记空间对齐，从而更有效地利用预训练的LLM。然而，这些方法依赖于离线自动语音识别（ASR），导致两个主要限制：（i）在标记化之前需要完整的发声，这阻碍了实时流式处理；（ii）ASR与LLM之间的词汇不匹配，降低了从子词到词级的声学粒度。我们提出了StreamAlign，一个文本对齐的语音标记化框架，能够实现实时语音-文本联合建模的流式标记化。StreamAlign通过结合字符级RNN-Transducer对齐与词级ASR指导，在线进行语音-文本对齐，减轻了ASR与LLM之间的词汇不匹配，同时保持识别准确性。一个主动的词边界分类器在块边界处预测词的完成，从而将标记化延迟从560毫秒减少到270毫秒。在LibriSpeech数据集上，StreamAlign在评估的标记化器中实现了最低的词错误率（WER）和最高的用户感知音质（UTMOS）。此外，基于StreamAlign单元训练的口语语言模型StreamAlign-SLM在语音续写任务中优于其他端到端口语语言模型，同时在SALMon和口语StoryCloze上实现了最强的一致性。
cs.CL / 22 / 2609.09764
SocialRL: Refining LLMs' Social Intelligence through Multi-turn Reinforcement Learning and Reward Design
SocialRL：通过多轮强化学习和奖励设计提升大型语言模型的社会智能
Abstract
Social intelligence enables agents to read social context, infer intent, and adapt over sustained dialogue. As language models become autonomous collaborators, it is central to building effective and trustworthy human-AI interaction. Existing reinforcement learning methods optimize single-turn utterances and sparse outcome rewards, producing short-sighted policies that struggle to manage goal-relationship tensions across multi-turn interactions. We propose SocialRL, a multi-turn reinforcement learning framework addressing both challenges. First, we apply multi-turn reinforcement learning using PPO that propagates delayed outcome rewards back to each turn, enabling long-horizon planning. Second, we design six process reward dimensions capturing the goal-relationship trade-off, including goal advancement, relational attunement, contextual coherence, etc. A reward model dynamically generates fine-grained scoring criteria for each dimension, while a stage-aware weight schedule prioritizes relationship-building in early turns, goal advancement mid-way, and balanced closure late. Across multiple social-dialogue benchmarks, SocialRL improves Goal Achievement by an average of 9.2 percentage points over the corresponding Base models. These results demonstrate the effectiveness of SocialRL across synthetic and real social scenes, as well as standard and challenging social scenarios.
Chinese Translation
社会智能使代理能够理解社会背景、推断意图并在持续对话中适应。当语言模型成为自主协作伙伴时，构建有效且值得信赖的人机互动变得至关重要。现有的强化学习方法优化单轮发言和稀疏的结果奖励，导致短视的策略，难以在多轮互动中管理目标与关系之间的紧张关系。我们提出了SocialRL，一个针对这两个挑战的多轮强化学习框架。首先，我们应用使用PPO的多轮强化学习，将延迟的结果奖励传播回每一轮，从而实现长远规划。其次，我们设计了六个过程奖励维度，以捕捉目标与关系之间的权衡，包括目标推进、关系调适、上下文一致性等。奖励模型动态生成每个维度的细化评分标准，而阶段感知的权重调度则优先考虑早期回合中的关系建立、中期的目标推进以及后期的平衡收尾。在多个社会对话基准测试中，SocialRL在目标达成率上比相应的基线模型平均提高了9.2个百分点。这些结果展示了SocialRL在合成和真实社会场景以及标准和具有挑战性的社会情境中的有效性。
cs.CL / 23 / 2609.09766
CARRE: Counterfactual Action Retrieval and Reason Evaluation for Explainable Churn Prescription
CARRE：可解释流失处方的反事实行动检索与理由评估
Abstract
Churn models typically identify high-risk customers but do not specify which feasible retention action should be considered or why that action is appropriate. We present CARRE (Counterfactual Action Retrieval and Reason Evaluation), a three-stage framework that combines retrieval-augmented candidate generation, cost-aware counterfactual scoring, and large language model (LLM) reasoning. CARRE retrieves a predefined catalog of retention actions, estimates model-predicted churn-risk changes under explicit feature transformations, and generates a structured churn reason and a profile-grounded explanation for the selected action. On the IBM Telco Customer Churn dataset, CARRE achieves 79.8% greater mean model-predicted risk reduction than the plain SHAP baseline and 80.4% greater reduction than the cost-controlled SHAP+Cost baseline across 313 high-risk test cases; its cost-normalized efficiency is 10.5% higher than that of plain SHAP. On a 136-case reason-stratified evaluation sample, diagnosis-driven prompt refinement increases weak-label agreement from 79.4% to 90.4%, with no auxiliary-plan constraint violations; because the same sample was used for error diagnosis and re-evaluation, the post-refinement result is not an independent estimate of generalization. For 135 explanations generated using the pre-refinement v2 reason outputs, two cross-vendor LLM judges assign mean scores ranging from 4.02 to 5.00 out of 5, although one judge saturates on actionability, and a deterministic audit finds no contradictions among 66 verifiable profile claims. Retrieval ablations show that k=5 provides the best evaluated compromise between high candidate coverage and downstream reasoning agreement in this dataset. These results illustrate how retrieval, model-based counterfactual scoring, and language generation can be separated and jointly evaluated in a prototype churn-prescription pipeline.
Chinese Translation
流失模型通常识别高风险客户，但并未具体说明应考虑哪些可行的保留行动或为何该行动是合适的。我们提出了CARRE（Counterfactual Action Retrieval and Reason Evaluation），这是一个三阶段框架，结合了增强检索的候选生成、成本感知的反事实评分和大型语言模型（LLM）推理。CARRE从预定义的保留行动目录中检索，估计在显式特征变换下模型预测的流失风险变化，并为所选行动生成结构化的流失原因和基于个人资料的解释。在IBM电信客户流失数据集上，CARRE在313个高风险测试案例中实现了比普通SHAP基线高出79.8%的平均模型预测风险降低，且比成本控制的SHAP+Cost基线高出80.4%；其成本标准化效率比普通SHAP高出10.5%。在一个136案例的理由分层评估样本中，基于诊断的提示优化将弱标签一致性从79.4%提高到90.4%，且没有违反辅助计划约束；由于同一样本用于错误诊断和重新评估，因此后优化结果并不是对泛化的独立估计。对于使用优化前的v2理由输出生成的135个解释，两个跨供应商的LLM评审员给出的平均评分在4.02到5.00之间，尽管一位评审员在可操作性上达到了饱和状态，且一个确定性审计在66个可验证的个人资料声明中未发现矛盾。检索消融实验表明，k=5在该数据集中提供了高候选覆盖率与下游推理一致性之间的最佳评估折衷。这些结果展示了如何在原型流失处方管道中分离和联合评估检索、基于模型的反事实评分和语言生成。
cs.CL / 24 / 2609.09772
SymbolicLight V2: Hybrid Neuromorphic Architecture and Sparse Execution for Low-Energy Language Inference
SymbolicLight V2：低能耗语言推理的混合神经形态架构与稀疏执行
Abstract
SymbolicLight V2 combines sparse event computation with continuous-state processing in a hybrid neuromorphic language architecture. Extending V1's spike-gated dual paths, it adds graded signed events at further projections and softmax-free local attention. We implement the 194M-parameter model on an Alveo U50C FPGA using digital fixed-point arithmetic and on an ARM CPU using sparse integer execution. Across three same-checkpoint FPGA implementations at 175 MHz, active-row weight gathering and valid-state KV loading raise decode throughput from 474.6 to 643.2 tokens/s for a 32-token prefix and 128 outputs. Estimated gross card energy falls from 0.06087 to 0.04407 J per generated token, a 27.6% reduction. Complete-request energy, including prefill, falls by 24.4-27.7% across three prefix lengths. An independent idle split attributes 82.8% of gross card energy to loaded idle, explaining the benefit of shorter token latency. Against the recorded RTX 5090 compiled-FP32 baseline, integer FPGA execution uses 89.1% less estimated card energy during short-context decode; arithmetic precisions differ, and the GPU baseline is not the lowest-energy tested configuration. On four Cortex-A76 cores of a ROCK 5T, complete requests reach 65.4 tokens/s at 9.80 W and 0.151 J per generated token at the adapter's AC input. These results connect event sparsity to omitted computation and data movement. The mechanisms also support other dedicated V2 implementations: increasing throughput by a greater factor than active power lowers energy per generated token. Evaluation holds the deployed checkpoint fixed; its quality trails a same-budget dense control, so the results do not establish equal-quality efficiency.
Chinese Translation
SymbolicLight V2 将稀疏事件计算与连续状态处理结合在一个混合神经形态语言架构中。它扩展了 V1 的脉冲门控双路径，增加了在进一步投影中的分级符号事件和无 softmax 的局部注意力。我们在 Alveo U50C FPGA 上实现了 194M 参数模型，使用数字定点算术，并在 ARM CPU 上使用稀疏整数执行。在三个相同检查点的 FPGA 实现中，175 MHz 的活跃行权重收集和有效状态 KV 加载将解码吞吐量从 474.6 提高到 643.2 个标记/秒，适用于 32 个标记前缀和 128 个输出。每生成一个标记的估计总卡能量从 0.06087 降至 0.04407 焦耳，减少了 27.6%。完整请求能量（包括预填充）在三个前缀长度中下降了 24.4-27.7%。独立的空闲分割将 82.8% 的总卡能量归因于加载的空闲状态，解释了较短标记延迟的好处。与记录的 RTX 5090 编译的 FP32 基线相比，整数 FPGA 执行在短上下文解码期间使用了 89.1% 更少的估计卡能量；算术精度不同，且 GPU 基线并不是测试的最低能耗配置。在 ROCK 5T 的四个 Cortex-A76 核心上，完整请求在适配器的交流输入下以 9.80 W 的功耗达到 65.4 个标记/秒，每生成一个标记的能量为 0.151 焦耳。这些结果将事件稀疏性与省略的计算和数据移动联系起来。这些机制还支持其他专用的 V2 实现：通过更大幅度提高吞吐量，活跃功率的降低使每生成一个标记的能量减少。评估保持已部署的检查点不变；其质量低于同预算的密集控制，因此结果并未确立等质量效率。
cs.CL / 25 / 2609.09778
ROAM: Robust Organization of Atomic Memories for Agents through Semantic Relations
ROAM：通过语义关系实现智能体原子记忆的稳健组织
Abstract
Long-term language-model agents rely on external memory across interactions. Atomic memories are particularly useful: their fine-grained semantic boundaries enable precise retrieval and direct comparison between observations. Yet accumulating atoms inevitably become redundant, overlapping, or conflicting. Existing methods often ask an LLM manager to add, update, delete, or rewrite memories directly, coupling semantic interpretation, storage decisions, and content generation in one error-prone operation. We introduce ROAM, a relation-guided framework that uses atomicity for management while allowing richer answer-time representations. ROAM classifies incoming--stored atom pairs as independent, equivalent, directionally subsuming, or conflicting, then organizes observations into active Primary and supporting Evidence roles. Fusion subsequently combines complementary details and temporal changes into compact, potentially non-atomic views. Only Primary views are retrieved for answering, preventing redundant or outdated atoms from competing independently. Across models and evaluation settings, ROAM improves answer accuracy by up to 29.8 percentage points. Ablations show complementary benefits from different relations and consistent gains from fusion beyond role organization. Mechanism analysis further finds 15.6-point higher answer-critical source recall and an 11.5-point lower confounder-token share. ROAM remains robust across manager scales.
Chinese Translation
长期语言模型智能体在交互中依赖外部记忆。原子记忆特别有用：其细粒度的语义边界能够实现精确的检索和观察之间的直接比较。然而，累积的原子不可避免地会变得冗余、重叠或冲突。现有方法通常要求大型语言模型（LLM）管理器直接添加、更新、删除或重写记忆，将语义解释、存储决策和内容生成结合在一个易出错的操作中。我们提出了ROAM，一个以关系为指导的框架，利用原子性进行管理，同时允许更丰富的回答时表示。ROAM将传入的存储原子对分类为独立、等价、方向性包含或冲突，然后将观察组织为主动的主要角色和支持证据角色。融合随后将互补的细节和时间变化组合成紧凑的、潜在的非原子视图。只有主要视图被检索用于回答，防止冗余或过时的原子独立竞争。在不同模型和评估设置中，ROAM将回答准确性提高了多达29.8个百分点。消融实验显示不同关系带来的互补效益，以及融合在角色组织之外的一致增益。机制分析进一步发现，回答关键源的召回率提高了15.6个百分点，而混淆词的比例降低了11.5个百分点。ROAM在管理器规模上保持稳健。
cs.CL / 26 / 2609.09791
MUCnoHARM@GermEval Shared Task 2026: Retrieval-based In-Context Learning for Defamatory Offences, and Where It Falls Short
MUCnoHARM@GermEval 2026共享任务：基于检索的上下文学习用于诽谤罪的检测及其不足之处
Abstract
With hate speech being ubiquitous online, automatic detection is crucial, in particular when it comes to criminally relevant social media posts. We study a variety of retrieval-based in-context learning (RetICL) strategies for detecting defamatory offences under {\S}{\S} 185-187 StGB (the subject of GermEval 2026 Subtask 4). Few-shot prompting beats zero-shot, but retrieval-based approaches offer only marginal gains over random demonstrations, and even fall behind an optimised static set of demonstrations. Providing concrete legal knowledge helps, yet model choice outweighs every other system choice. Models over-predict criminal relevance while still missing 26-57% of criminally relevant posts, suiting them for triage rather than autonomous moderation.
Chinese Translation
随着仇恨言论在网络上的普遍存在，自动检测显得尤为重要，尤其是在涉及刑事相关的社交媒体帖子时。我们研究了多种基于检索的上下文学习（RetICL）策略，用于检测根据德国刑法第185-187条（StGB）规定的诽谤罪（GermEval 2026子任务4的主题）。少量示例提示优于零示例，但基于检索的方法在随机示例上仅提供了边际收益，甚至不及优化的静态示例集。提供具体的法律知识有所帮助，但模型选择的影响超过了其他系统选择。模型过度预测刑事相关性，同时仍然漏掉26-57%的刑事相关帖子，适合用于初步筛查而非自主审核。
cs.CL / 27 / 2609.09835
HyperTrace: Hypothesis-Based Preference Tracing for Online LLM Personalization
HyperTrace：基于假设的在线大语言模型个性化偏好追踪
Abstract
Personalized language models aim to adapt responses to individual users, whose preferences are often latent and revealed gradually through interaction. Existing training-free methods rely on stored histories or retrieved memories, but they often struggle to reconcile long- term preferences with short-term topic-specific needs. To address this issue, we propose HyperTrace, a training-free framework that formulates online personalization as latent preference tracing. HyperTrace maintains interpretable natural-language hypotheses over short-term intent and long-term preferences, and updates them through an SMC-style reweight process using an LLM-based surrogate choice model. By updating these hypotheses across turns and sessions, HyperTrace enables personalization without parameter updates. Experiments on PRISM and PersonaMem-v2 show that HyperTrace improves response alignment, preference prediction, and profile consistency over strong online baselines, demonstrating the effectiveness of tracing latent user preferences for robust personalization. Code and scripts are available in the repository: https://github.com/jiseshen/HyperTrace.
Chinese Translation
个性化语言模型旨在根据个体用户的需求调整响应，而用户的偏好往往是潜在的，并通过互动逐渐显现。现有的无训练方法依赖于存储的历史记录或检索的记忆，但它们常常难以协调长期偏好与短期特定主题需求之间的关系。为了解决这一问题，我们提出了HyperTrace，一个将在线个性化表述为潜在偏好追踪的无训练框架。HyperTrace维护关于短期意图和长期偏好的可解释自然语言假设，并通过基于LLM的替代选择模型使用SMC风格的重加权过程对其进行更新。通过在多个回合和会话中更新这些假设，HyperTrace能够在不更新参数的情况下实现个性化。在PRISM和PersonaMem-v2上的实验表明，HyperTrace在响应对齐、偏好预测和用户画像一致性方面优于强大的在线基线，证明了追踪潜在用户偏好在实现稳健个性化中的有效性。代码和脚本可在以下仓库获取：https://github.com/jiseshen/HyperTrace。
cs.CL / 28 / 2609.09852
$S^3$-Bench: Evaluating Speech Interaction Models as Scientific Voice Assistants
$S^3$-Bench：评估作为科学语音助手的语音交互模型
Abstract
The advance of multimodal large language models (MLLMs) has fundamentally reshaped the paradigm of human-computer interaction, especially speech interaction models capable of seamless conversations. Despite remarkable performance as general voice assistants, their performance in specialized domains remains underexplored, particularly in scientific areas. Scientific interactions introduce formidable challenges, involving rare technical terminology, spoken norms of abbreviations, and the natural verbalization of symbolic special expressions. In this paper, we introduce S$^3$-Bench, a systematic evaluation framework covering 10 major disciplines, consisting of a Knowledge set for speech question-answering and a Dialogue set for multi-turn progressive interactions with simulated user agents. By decomposing a complete atomic turn into stages of speech recognition, perception, knowledge utilization with reasoning, and response pronunciation, we systematically characterize the common challenges and performance tradeoffs of existing approaches. Furthermore, experiments on multi-turn interactions reveal persistent limitations in user adaptation and the generation of accurate, comprehensive, and efficient responses.
Chinese Translation
多模态大型语言模型（MLLMs）的进展从根本上重塑了人机交互的范式，尤其是能够进行无缝对话的语音交互模型。尽管作为通用语音助手表现出色，但它们在专业领域的表现仍然未得到充分探索，特别是在科学领域。科学交互带来了严峻的挑战，包括稀有的技术术语、缩略词的口语规范以及符号特殊表达的自然口头化。在本文中，我们介绍了 S$^3$-Bench，这是一个系统的评估框架，涵盖10个主要学科，包含用于语音问答的知识集和用于与模拟用户代理进行多轮渐进交互的对话集。通过将完整的原子轮次分解为语音识别、感知、知识利用与推理以及响应发音的各个阶段，我们系统地描述了现有方法的共同挑战和性能权衡。此外，关于多轮交互的实验揭示了用户适应性和生成准确、全面、高效响应的持续局限性。
cs.CL / 29 / 2609.09887
When Does Defendant Statement Matter? A Study of Bias and Persuasion in LLM-Simulated Jurors
被告陈述何时重要？LLM模拟陪审员中的偏见与说服研究
Abstract
LLMs have been used to simulate human decision-making in professional settings, yet their behaviors in common-law jury trials remain unexplored. We study when and how a defendant's courtroom statement affects LLM-simulated jurors, focusing on persuasion, ideological bias, and background-based affinity. To support the analysis, we introduce JuryBench, a benchmark containing controversial criminal cases in U.S. criminal law. In each case, a defendant can claim various plausible justifications to support acquittal or reduced liability. We fix the base case and design defendants of different backgrounds, who give courtroom statements with varying emotional appeal or rebuttal. Jurors with diverse ideological profiles across the spectrum are simulated. We examine 20 frontier LLMs, resulting in a total of 432K decisions and rationales, and quantify changes in verdict severity. Our findings show that LLM-jury simulation echoes many human-jury findings. First, emotional persuasion can be detrimental, since jurors may perceive it as evidence of guilt or inconsistency. Next, we show that background fit between jurors and defendants is a stronger and significant factor than other isolated factors, and that jurors are in general harsher toward opposite-background defendants and lenient toward same-background ones. Finally, we find that juror ideology also strongly shapes severity judgments. These findings highlight both the promise and risks of using LLMs to model jury reasoning and call for careful evaluation. The data and code are available at https://github.com/choyingw/JuryBench
Chinese Translation
大型语言模型（LLMs）已被用于模拟专业环境中的人类决策，但它们在普通法陪审团审判中的行为仍未得到探索。我们研究被告在法庭上的陈述何时以及如何影响LLM模拟的陪审员，重点关注说服力、意识形态偏见和基于背景的亲和力。为了支持分析，我们引入了JuryBench，这是一个包含美国刑法中有争议的刑事案件的基准。在每个案件中，被告可以提出各种合理的辩解以支持无罪或减轻责任。我们固定基本案例，并设计不同背景的被告，他们在法庭上发表具有不同情感吸引力或反驳的陈述。模拟了具有多样意识形态特征的陪审员。我们考察了20个前沿的LLM，结果产生了总计432K的决策和理由，并量化了裁决严厉程度的变化。我们的研究结果表明，LLM陪审团模拟反映了许多人类陪审团的发现。首先，情感说服可能是有害的，因为陪审员可能将其视为罪恶或不一致的证据。接下来，我们展示了陪审员与被告之间的背景契合度是比其他孤立因素更强且显著的因素，并且陪审员通常对背景相反的被告更为严厉，而对背景相同的被告则更为宽容。最后，我们发现陪审员的意识形态也强烈影响裁决的严厉程度。这些发现突显了使用LLM建模陪审团推理的潜力与风险，并呼吁进行仔细评估。数据和代码可在 https://github.com/choyingw/JuryBench 获取。
cs.CL / 30 / 2609.09889
Leveraging Fine-grained Error Correction in Korean Speech Recognition for Consultation Services
利用细粒度错误修正提升韩语语音识别在咨询服务中的应用
Abstract
Automatic Speech Recognition (ASR) technology is fundamental to customer service automation and large-scale transcription. However, even advanced ASR models exhibit inevitable errors in complex real-world environments such as call center conversations. When privacy restrictions preclude audio access, error correction must rely on text-based post-editing. Existing text-only approaches face significant challenges in low-resource languages, mainly due to a critical scarcity of annotated corpora and tailored correction methodologies. For Korean, this resource gap is particularly pronounced, as existing resources are predominantly designed for ASR training rather than text-based error correction. To address this, we introduce DasanCallDial, the first large-scale Korean benchmark dataset specifically curated for dialogue-level ASR error correction. Derived from genuine call center interactions, it comprises 1,974 dialogues with 115,460 utterances. Leveraging this resource, we propose Detector-Gated Contextual Span Correction (DCSC), a text-only post-editing framework for error-sparse Korean speech recognition transcripts. DCSC combines an encoder-based detector that first performs token-level error detection, followed by a language model-based corrector trained to rectify fine-grained span-level errors. Additionally, we employ dialogue-level context augmentation to enable the model to leverage discourse history for disambiguation. By employing multi-level granularity, our method achieves state-of-the-art performance, effectively overcoming the limitations of general LLMs in low-resource settings.
Chinese Translation
自动语音识别（ASR）技术是客户服务自动化和大规模转录的基础。然而，即使是先进的ASR模型在复杂的现实环境中，如呼叫中心对话中，仍然不可避免地会出现错误。当隐私限制阻止音频访问时，错误修正必须依赖基于文本的后期编辑。现有的仅基于文本的方法在低资源语言中面临重大挑战，主要是由于标注语料库和量身定制的修正方法的严重匮乏。对于韩语而言，这种资源差距尤为明显，因为现有资源主要是为ASR训练而设计，而非文本基础的错误修正。为了解决这一问题，我们引入了DasanCallDial，这是第一个专门为对话级ASR错误修正而精心策划的大规模韩语基准数据集。该数据集来源于真实的呼叫中心互动，包含1,974个对话和115,460个发言。利用这一资源，我们提出了检测器门控上下文跨度修正（Detector-Gated Contextual Span Correction, DCSC），这是一个针对错误稀疏的韩语语音识别转录的文本后期编辑框架。DCSC结合了一个基于编码器的检测器，首先执行基于标记的错误检测，然后是一个基于语言模型的修正器，旨在修正细粒度的跨度级错误。此外，我们采用对话级上下文增强，使模型能够利用话语历史进行消歧。通过采用多级粒度，我们的方法实现了最先进的性能，有效克服了在低资源环境中通用大型语言模型的局限性。
cs.CL / 31 / 2609.09901
Deep and shallow biases in language models
语言模型中的深层与浅层偏见
Abstract
Large language models often repeatedly select the same answer even when many alternatives are plausible. Prior work treats this concentration as bias, but it does not distinguish stable model preferences from responses that depend on a particular prompt wording. We introduce a bias depth score that measures both how strongly a model prefers its top answer under direct prompting and whether that answer survives scenario reframing. Across 4,442 opinion prompts and four large language models, only about a quarter of the concentrated preferences survive reframing. We call these persistent cases Deep biases, and the remaining prompt-dependent cases Shallow biases. Our results show that Deep biases are more often inherited from pretraining and preserved through SFT. Under both continued fine-tuning and prompt-based debiasing for diversity, Deep biases are consistently harder to remove than Shallow biases. Bias depth therefore separates stable learned biases from prompt-wording artifacts that single-prompt metrics conflate. Code, models, and data are available at deepbias.github.io.
Chinese Translation
大型语言模型在面对许多合理的替代答案时，往往会反复选择相同的答案。之前的研究将这种集中现象视为偏见，但并未区分模型的稳定偏好与依赖特定提示措辞的反应。我们引入了一种偏见深度评分，既衡量模型在直接提示下对其首选答案的偏好强度，也评估该答案在情境重构中是否依然有效。在4442个意见提示和四个大型语言模型的研究中，只有约四分之一的集中偏好在重构中得以保留。我们将这些持久的情况称为深层偏见（Deep biases），而将其余依赖提示的情况称为浅层偏见（Shallow biases）。我们的结果表明，深层偏见更常常是从预训练中继承而来，并通过微调（SFT）得以保留。在继续微调和基于提示的去偏见多样性处理下，深层偏见始终比浅层偏见更难以去除。因此，偏见深度能够将稳定的学习偏见与单一提示度量混淆的提示措辞伪影区分开。代码、模型和数据可在 deepbias.github.io 获取。
cs.CL / 32 / 2609.09902
Contrastive Projection: Reading Transformer Internals by Differencing Logit Lenses
对比投影：通过差异化逻辑透镜阅读变换器内部
Abstract
Reading a transformer's internal states in token space is easy to do and hard to trust: a logit lens on a single hidden state is dominated, at intermediate layers, by the generic tokens the model would predict for almost any input. We read the difference instead. Subtracting two closely matched prompts' hidden states and projecting through the unembedding cancels the shared component and surfaces what separates them, an operation equivalent to reading a RepE/ActAdd steering vector through a logit lens. Built into a training-free tracer that reads at every position, sub-layer, and head and averages over designed baselines, it traces a compound- noun MLP->attention chain in Phi-2, confirmed there by activation patching, with the same distinction recovered across three architectures by readout and probe rather than by patching; it reads what retrieval surfaces for real versus fictional entities, and reads metaphor as a set of domain-to-domain mappings rather than a single figurativity feature. A cross-seed control marks the boundary: across five networks differing only in initialization, the same distinction surfaces as almost entirely different tokens (top-10 overlap 0.08). What a computation looks like in token space is network-specific; the distinction it draws is not
Chinese Translation
在标记空间中读取变换器的内部状态既简单又难以信任：在中间层，单个隐藏状态上的逻辑透镜被模型对几乎任何输入预测的通用标记所主导。我们选择读取差异。通过减去两个紧密匹配的提示的隐藏状态并通过去嵌入进行投影，可以消除共享成分并显现出它们之间的差异，这一操作等同于通过逻辑透镜读取 RepE/ActAdd 引导向量。该方法内置于一个无训练的追踪器中，能够在每个位置、子层和头部进行读取，并对设计的基线进行平均，它追踪了 Phi-2 中的复合名词 MLP->注意力链，并通过激活修补在此得到确认；同样的区分通过读取和探测而非修补在三种架构中恢复；它读取真实与虚构实体的检索表面，并将隐喻视为一组领域到领域的映射，而不是单一的比喻特征。跨种子控制标记了边界：在五个仅在初始化上有所不同的网络中，相同的区分表现为几乎完全不同的标记（前10个重叠为0.08）。计算在标记空间中的表现是网络特定的；它所描绘的区分并非如此。
cs.CL / 33 / 2609.09953
Improving Cross-Lingual Token Representations by Adding a Pinch of SALT
通过添加少量SALT来改善跨语言标记表示
Abstract
Cross-lingual sentence encoders enable scalable transfer across hundreds of languages, powering applications such as translation mining and zero-shot learning in low-resource settings. Although trained for sentence-level alignment, they are increasingly also applied to token-level tasks such as hallucination detection and sequence tagging, exposing a mismatch between training and usage. We propose SALT, a lightweight post-training method that improves token representations by injecting span-level supervision into existing sentence encoders. Across five multilingual token-level benchmarks, SALT achieves the best overall results on four of them, outperforming alternative fine-tuning strategies and competitive encoders. It also improves sentence-level performance on cross-lingual retrieval and classification tasks. These results demonstrate that span-level supervision is an effective signal for improving both token and sentence representations.
Chinese Translation
跨语言句子编码器使得在数百种语言之间进行可扩展的迁移成为可能，推动了翻译挖掘和低资源环境下的零样本学习等应用。尽管这些编码器经过训练以实现句子级对齐，但它们越来越多地被应用于标记级任务，如幻觉检测和序列标注，这暴露了训练与使用之间的不匹配。我们提出了SALT（Span-level Supervision for Alignment and Tokenization），一种轻量级的后训练方法，通过将跨度级监督注入现有的句子编码器来改善标记表示。在五个多语言标记级基准测试中，SALT在其中四个上取得了最佳整体结果，超越了其他微调策略和竞争性编码器。它还改善了跨语言检索和分类任务的句子级性能。这些结果表明，跨度级监督是改善标记和句子表示的有效信号。
cs.CL / 34 / 2609.09964
5-Dialects-BN: Unmasking the Impact of Transliteration on Bangla Dialectal LLMs
5-Dialects-BN：揭示音译对孟加拉方言大型语言模型的影响
Abstract
Large Language Models (LLMs) have achieved remarkable progress across natural language processing (NLP) tasks, yet their capabilities degrade sharply for low-resource languages and dialectally diverse settings. Bangla, the world's sixth most spoken language, exemplifies this gap: existing resources overwhelmingly target Standard Bangla, leaving its regional dialects without the benchmarks needed to develop or evaluate dialect-aware systems. We address this gap with 5-Dialects-BN, the first multi-annotation Bangla dialect benchmark to align Romanized transliteration with dialectal text, Standard Bangla, English, and subjectivity labels across five regional varieties. The dataset comprises 6,000 manually annotated entries spanning five major dialects: Chittagong, Barisal, Noakhali, Sylhet, and Rangpur (Chittagong 1,900; Noakhali 1,500; Sylhet 1,200; Barisal 700; Rangpur 700), reflecting natural online availability. Each entry is enriched with five aligned annotations: the original dialectal text, a Romanized transliteration, an English translation, a Standard Bangla translation, and a subjectivity label (subjective vs. objective). Annotations were produced and cross-validated by native speakers and undergraduate linguistics students to ensure dialectal authenticity and semantic fidelity. The resulting resource supports a diverse suite of tasks, including dialect identification, dialect-to-standard normalization, machine translation, subjectivity classification, and parameter-efficient fine-tuning (e.g., LoRA) of multilingual LLMs. By providing a standardized, multi-annotation benchmark, 5-Dialects-BN enables principled evaluation of LLMs on dialectally diverse Bangla and lays a foundation for further research in low-resource, dialect-aware NLP.
Chinese Translation
大型语言模型（LLMs）在自然语言处理（NLP）任务中取得了显著进展，但在低资源语言和方言多样化的环境中，其能力急剧下降。孟加拉语是世界第六大语言，充分体现了这一差距：现有资源主要针对标准孟加拉语，导致其区域方言缺乏开发或评估方言感知系统所需的基准。我们通过5-Dialects-BN解决了这一问题，这是第一个多注释的孟加拉方言基准，旨在将罗马化音译与方言文本、标准孟加拉语、英语及主观性标签对齐，涵盖五种区域变体。该数据集包含6,000个手动注释的条目，涵盖五种主要方言：吉大港、巴里萨尔、诺阿卡利、西里杰和朗布尔（吉大港1,900；诺阿卡利1,500；西里杰1,200；巴里萨尔700；朗布尔700），反映了自然在线可用性。每个条目都配有五个对齐的注释：原始方言文本、罗马化音译、英语翻译、标准孟加拉语翻译和主观性标签（主观与客观）。注释由母语者和本科语言学学生制作并交叉验证，以确保方言的真实性和语义的准确性。该资源支持多样化的任务，包括方言识别、方言到标准化的转换、机器翻译、主观性分类以及多语言LLMs的参数高效微调（例如，LoRA）。通过提供一个标准化的多注释基准，5-Dialects-BN使得对方言多样化的孟加拉语LLMs进行原则性评估成为可能，并为低资源、方言感知的NLP进一步研究奠定了基础。
cs.CL / 35 / 2609.09974
Towards Stress-Aware Sentence-Level Filipino G2P With Weakly-Supervised ByT5 Fine-Tuning
面向压力感知的句子级菲律宾语G2P的弱监督ByT5微调研究
Abstract
Grapheme-to-phoneme conversion (G2P) refers to the task of converting a sequence of graphemes to a corresponding sequence of phonemes. While Filipino G2P is fairly straightforward due to its shallow orthography, the inclusion of prosodic features such as stress adds a layer of complexity that requires sentence-level context instead of single-word inputs. However, sentence-level data for Filipino typically do not include phoneme transcriptions, posing a challenge for training G2P models. As such, we investigate how to obtain sentence-level phoneme data for Filipino using available data and compare the resulting models with multilingual word-level G2P as well as measure how accurately they predict stress marker position for Filipino. We propose fine-tuning a ByT5-based model, pre-trained on multilingual word-level G2P data, on three sentence-level G2P datasets annotated with an LLM-assisted pipeline guided by data from Wiktionary. This approach produces models that perform well on the G2P task, achieving at best around 0.54% PER and 2.50% CER, a significant decrease compared to base model PER at around 19.74%, on a manually-corrected test set. The model is able to correctly classify most of the main stress classes in Filipino, but struggles particularly with malumi words. We show that a ByT5-based model performs well at sentence-level Filipino G2P and offers strong potential for Filipino homograph disambiguation.
Chinese Translation
图形到音素转换（G2P）是指将一系列图形转换为相应的音素序列的任务。尽管菲律宾语的G2P由于其浅显的正字法相对简单，但加入诸如重音等韵律特征则增加了复杂性，这需要句子级的上下文，而非单词输入。然而，菲律宾语的句子级数据通常不包含音素转录，这对训练G2P模型构成了挑战。因此，我们研究如何利用现有数据获取菲律宾语的句子级音素数据，并将所得模型与多语言单词级G2P进行比较，同时测量它们对菲律宾语重音标记位置的预测准确性。我们提出对基于ByT5的模型进行微调，该模型在多语言单词级G2P数据上进行了预训练，并在三个使用LLM辅助管道标注的句子级G2P数据集上进行微调，管道由来自Wiktionary的数据指导。这种方法产生的模型在G2P任务上表现良好，在经过人工修正的测试集上，最佳的音素错误率（PER）约为0.54%，字符错误率（CER）约为2.50%，相比基础模型的PER约为19.74%有显著下降。该模型能够正确分类大多数菲律宾语的主要重音类别，但在处理malumi词时表现较差。我们展示了基于ByT5的模型在句子级菲律宾语G2P任务中表现良好，并为菲律宾语同形异义词消歧提供了强大的潜力。
cs.CL / 36 / 2609.09984
Multi-Functional Embedding Models for Funder Name Disambiguation in Scientific Publication Records
用于科学出版记录中资助者名称消歧义的多功能嵌入模型
Abstract
Understanding the historical allocation and distribution of research funding advances our knowledge of how scientific research is supported across fields, institutions, and regions. However, large-scale analyses are hindered by the lack of comprehensive funder name disambiguation solutions, as funder names often exhibit spelling variations, translations, abbreviations, and inconsistent levels of granularity. In this paper, we present a framework for developing multilingual, multi-functional funder name disambiguation models and demonstrate its application to research publications in biodiversity conservation. To construct a training dataset, we integrated the Research Organization Registry (ROR), which provides unique identifiers for research organizations, with two publication datasets: the Web of Science (WoS) and the Crossref Open Funder Registry (OFR). We used multi-task learning with Contrastive Loss and Multiple Negatives Ranking Loss to fine-tune three open-weight embedding models from the Sentence Transformer, Gemma, and Qwen3 families. The best-performing models achieved accuracy above 0.90 when matching WoS funder names to ROR identifiers, outperforming general-purpose LLMs, including GPT-5.2, Claude-Sonnet-4.6, and Gemini-2.5-Flash, by more than 0.1. For funder names not indexed in ROR, we constructed a similarity network among funder names and identified clusters within it. Finally, we analyzed the disambiguation results and highlighted challenges arising from limited knowledge of smaller funders and funders from non-English-speaking countries. This work provides a reusable framework for funder name disambiguation with potential applicability across different model architectures and datasets, featuring cost-effective training data creation and multi-task learning and disambiguation.
Chinese Translation
理解研究资金的历史分配和分布有助于我们了解科学研究在各个领域、机构和地区的支持方式。然而，由于资助者名称常常存在拼写变体、翻译、缩写和不一致的粒度水平，大规模分析受到限制。本文提出了一种开发多语言、多功能资助者名称消歧义模型的框架，并展示其在生物多样性保护研究出版物中的应用。为了构建训练数据集，我们将提供研究组织唯一标识符的研究组织注册（Research Organization Registry, ROR）与两个出版物数据集整合：科学网（Web of Science, WoS）和Crossref开放资助者注册（Open Funder Registry, OFR）。我们使用对比损失（Contrastive Loss）和多负样本排名损失（Multiple Negatives Ranking Loss）进行多任务学习，以微调来自句子变换器（Sentence Transformer）、Gemma和Qwen3系列的三个开放权重嵌入模型。表现最佳的模型在将WoS资助者名称与ROR标识符匹配时，准确率超过0.90，超越了包括GPT-5.2、Claude-Sonnet-4.6和Gemini-2.5-Flash在内的通用大型语言模型（LLMs）超过0.1。对于未在ROR中索引的资助者名称，我们构建了资助者名称之间的相似性网络，并识别出其中的聚类。最后，我们分析了消歧义结果，并强调了由于对小型资助者和非英语国家资助者的知识有限而带来的挑战。这项工作提供了一个可重复使用的资助者名称消歧义框架，具有在不同模型架构和数据集中的潜在适用性，特点是成本效益高的训练数据创建和多任务学习与消歧义。
cs.CL / 37 / 2609.09985
VLX-VR: An Agentic-Aware Video Reasoning Model
VLX-VR：一种具有代理意识的视频推理模型
Abstract
Real-world video understanding requires integrating visual, audio, textual, and temporal evidence distributed across a video. Yet many pipelines use a fixed video context and single-pass inference, limiting adaptive evidence acquisition when observations are incomplete, ambiguous, or conflicting. We present VLX-VR, an agentic-aware video reasoning model trained within a video reasoning framework defined by a Think--Memory--Observation loop. At each step, VLX-VR determines the needed evidence, invokes read_memory or write_memory, incorporates the returned Observation, and decides whether to continue or produce the task output. We train VLX-VR with multimodal data, including videos and agent trajectories, using reinforcement learning to learn evidence acquisition, memory use, and termination. On MINERVA, VLX-VR achieves state-of-the-art performance among the models included in our comparison, with 78.79% accuracy. Under the original three duration groups, its accuracies are 76.70%, 78.73%, and 80.92%, with a cross-duration accuracy variance of 2.97~$\mathrm{pp}^2$. On correctly answered samples, 96.20% of VLX-VR's reasoning traces are consistent with the MINERVA reference reasoning traces and the evidence described by them, while approximately 75.80% of all evaluated samples satisfy both answer correctness and this evidence-grounded trace criterion. These results show strong performance and broadly stable behavior across durations, while counting, state changes, causal reasoning, and spatial perception remain challenging.
Chinese Translation
现实世界的视频理解需要整合分布在视频中的视觉、音频、文本和时间证据。然而，许多处理流程使用固定的视频上下文和单次推理，这限制了在观察不完整、模糊或冲突时的自适应证据获取。我们提出了VLX-VR，一种在由思维-记忆-观察循环定义的视频推理框架内训练的具有代理意识的视频推理模型。在每一步中，VLX-VR确定所需的证据，调用read_memory或write_memory，整合返回的观察结果，并决定是继续还是产生任务输出。我们使用多模态数据（包括视频和代理轨迹）对VLX-VR进行训练，通过强化学习学习证据获取、记忆使用和终止。在MINERVA数据集上，VLX-VR在我们比较的模型中实现了最先进的性能，准确率达到78.79%。在原始的三个时长组中，其准确率分别为76.70%、78.73%和80.92%，跨时长的准确率方差为2.97~$ ext{pp}^2$。在正确回答的样本中，96.20%的VLX-VR推理轨迹与MINERVA参考推理轨迹及其描述的证据一致，而大约75.80%的所有评估样本同时满足答案正确性和这一基于证据的轨迹标准。这些结果显示出在各个时长上强大的性能和广泛稳定的行为，尽管计数、状态变化、因果推理和空间感知仍然具有挑战性。
cs.CL / 38 / 2609.09989
Stable Answers, Unfinished Reasoning: Why Self-Consensus Is Not a Safe Early-Exit Signal
稳定的答案，未完成的推理：为何自我共识不是安全的早期退出信号
Abstract
A natural way to cut reasoning-model inference cost is to repeatedly probe a single partial trajectory for its current answer and stop once probes agree -- self-consensus. We ask whether any such rule is both safe and token-saving, and whether one can be selected once and reused. A preregistered sweep of 3,520 consensus rules, replayed on frozen trajectories from two models and three benchmarks, clears none of three acceptance gates fixed in advance; the frontier reproduces on a held-out split and on two unseen models -- while a boundary-confidence control (DEER) swept through the same pipeline clears all three. The reason lies in the signal: agreement establishes that the current answer persists under a fixed probing procedure, not that the reasoning has terminated -- a consensus-termination gap. Stopping on it commits non-terminal answers. At a rule still saving 32% of the tokens, one stop in nine fires on an answer the trajectory itself later abandons, and most of those stops cut off a correction it would otherwise have made. Widening the agreement window does not remove them: the share levels off near 7%, and by then the saving has fallen to 8%. Probe re-wording and a hand-labelled error taxonomy show the agreed answer is often a placeholder the model had not settled on. Used on its own as the stop signal, agreement fails not because it is insufficiently strict, but because it repeatedly measures the wrong object.
Chinese Translation
降低推理模型推断成本的一种自然方法是反复探测单个部分轨迹以获取其当前答案，并在探测结果一致时停止——自我共识。我们探讨了是否存在既安全又节省令牌的规则，以及是否可以选择一个规则并重复使用。对3520个共识规则进行的预注册筛查，在来自两个模型和三个基准的冻结轨迹上进行重放，未能通过预先设定的三个接受门；而在持出分割和两个未见模型上，前沿表现出重现性——同时，边界置信度控制（DEER）在相同流程中通过了所有三个门。原因在于信号：一致性表明当前答案在固定探测程序下持续存在，而不是推理已经结束——这形成了共识终止差距。在此基础上停止会导致非终端答案的承诺。在一个仍能节省32%令牌的规则下，每九次停止中就有一次在一个轨迹随后放弃的答案上触发，而这些停止大多数会切断它本可以做出的修正。扩大一致性窗口并未消除这些问题：其比例稳定在约7%，而此时节省已降至8%。探测重措和手动标注的错误分类法显示，达成一致的答案往往是模型尚未确定的占位符。作为停止信号单独使用时，一致性失败的原因不是因为它不够严格，而是因为它反复测量了错误的对象。
cs.CL / 39 / 2609.09999
SalamandraTA at WMT 2026 Terminology Shared Task: Hard Examples Are Better Teachers
SalamandraTA在WMT 2026术语共享任务中的表现：困难示例是更好的教师
Abstract
Terminology-aware translation asks for more than a correct translation: the output must use the exact terms a glossary prescribes. The standard recipe, fine-tuning on glossary-annotated translation pairs, hides an inefficiency: for most examples the glossary prescribes exactly what the model would have produced anyway, so they teach nothing about following a glossary. We therefore keep only the examples where the model's own translation contradicts the glossary. In a controlled study at fixed data volume, this selection alone raises term accuracy from 78.7% to 89.9%. The filtered data, built by a two-way synthetic pipeline on open models, is part of the instruction-tuning mixture of our public release SalamandraTA-7b-instruct v3.0, which, used exactly as released and wrapped in a document-level inference pipeline, forms the BSC submission to the WMT26 Terminology Shared Task Track 1. At the official WMT26 evaluation, our system achieves 94.2% term success at 74.6 chrF++, with only two of the twenty-two submissions outperforming it on both metrics. On last year's benchmark, it also surpasses our GRPO-based system, despite being trained solely with ordinary supervised fine-tuning.
Chinese Translation
术语感知翻译不仅仅要求正确的翻译：输出必须使用词汇表所规定的确切术语。标准做法是对带有术语注释的翻译对进行微调，但这隐藏了一种低效性：对于大多数示例，词汇表所规定的正是模型本来会产生的内容，因此这些示例并没有教会模型如何遵循词汇表。因此，我们仅保留模型自身翻译与词汇表相矛盾的示例。在固定数据量的控制研究中，仅此选择就将术语准确率从78.7%提高到89.9%。经过双向合成管道构建的过滤数据，作为我们公共发布的SalamandraTA-7b-instruct v3.0的指令微调混合的一部分，正是以发布的方式使用，并封装在文档级推理管道中，形成了WMT26术语共享任务第1轨道的BSC提交。在官方WMT26评估中，我们的系统在74.6 chrF++的基础上实现了94.2%的术语成功率，仅有22个提交中有两个在这两个指标上超越了我们。在去年的基准测试中，尽管仅通过普通的监督微调进行训练，它也超越了我们基于GRPO的系统。
cs.CL / 40 / 2609.10049
MedDeID enables locally governed clinical-text de-identification from real or synthetic training data
MedDeID 实现了基于真实或合成训练数据的本地管理临床文本去标识化
Abstract
Clinical notes contain personally identifiable information (PII), restricting reuse for research and medical AI, especially when data cannot leave an institution. We developed MedDeID, an on-premises framework combining in-house annotation and synthetic-note generation with model training, inference, pseudonymisation and evaluation. On an independently annotated, adjudicated 300-note Dutch hospital benchmark, a hospital-trained compact transformer detected 98.9% of identifying text while redacting 0.24% of text outside annotated identifiers; a synthetic-only counterpart detected 96.1%. On 100 primary-care notes, the synthetic-trained model achieved higher recall than the hospital-trained model (90.3% versus 87.0%) and greater robustness to identifier-format perturbations. An English instantiation trained without real text detected 99.7% and 98.9% of annotated identifier characters on two external synthetic benchmarks. These results demonstrate transfer of the workflow to another language, but not clinical English performance. MedDeID provides a route to locally governed de-identification using real or synthetic training data.
Chinese Translation
临床记录包含个人可识别信息（PII），限制了其在研究和医疗人工智能中的再利用，尤其是在数据无法离开机构的情况下。我们开发了 MedDeID，这是一个本地框架，结合了内部注释和合成记录生成，以及模型训练、推理、假名化和评估。在一个独立注释、裁定的 300 份荷兰医院基准数据集上，经过医院训练的紧凑型变换器检测到了 98.9% 的识别文本，同时对注释标识符之外的文本进行了 0.24% 的删减；而仅使用合成数据的对应模型检测到了 96.1%。在 100 份初级护理记录中，合成训练模型的召回率高于医院训练模型（90.3% 对比 87.0%），并且对标识符格式扰动表现出更大的鲁棒性。一个未使用真实文本训练的英文实例在两个外部合成基准上检测到了 99.7% 和 98.9% 的注释标识符字符。这些结果展示了工作流程向另一种语言的迁移，但在临床英语性能上并未取得相应效果。MedDeID 提供了一条使用真实或合成训练数据进行本地管理去标识化的途径。
cs.CL / 41 / 2609.10052
Direct Diversity Optimization for Diverse Successful Trajectories in Preference Post-Training
偏好后训练中多样成功轨迹的直接多样性优化
Abstract
LLM agents for sequential decision tasks are often post-trained with trajectory-level outcome labels, but such labels provide little supervision for preserving multiple successful branches from the same decision state. We study this problem as successful strategy coverage: how broadly a model realizes distinct successful strategies under a fixed rollout budget. We present Direct Diversity Optimization (DDO), an offline post-training method that combines Divergence-Tree Collection (DTC) with the Reference-Relative Target-Odds Objective (RTO). DTC constructs state-aligned branch sets rooted at shared decision states, and RTO trains the model to match reference-relative targets over successful alternatives. DDO achieves the strongest task success and successful strategy coverage among the compared post-training methods across BabyAI, BabaIsAI, and WebShop. It also achieves the highest recovery rate after local action replacement and higher task success and coverage than successful-only imitation and decoding-time diversification controls.
Chinese Translation
用于序列决策任务的大型语言模型（LLM）代理通常会通过轨迹级结果标签进行后训练，但这些标签对保留来自同一决策状态的多个成功分支提供的监督有限。我们将这个问题视为成功策略覆盖：模型在固定的展开预算下实现不同成功策略的广度。我们提出了直接多样性优化（Direct Diversity Optimization, DDO），这是一种离线后训练方法，结合了发散树集合（Divergence-Tree Collection, DTC）和参考相对目标赔率目标（Reference-Relative Target-Odds Objective, RTO）。DTC构建了以共享决策状态为根的状态对齐分支集，而RTO则训练模型以匹配成功替代方案的参考相对目标。在与BabyAI、BabaIsAI和WebShop的比较后训练方法中，DDO在任务成功率和成功策略覆盖方面表现最强。它在局部动作替换后的恢复率也最高，并且在任务成功率和覆盖率上优于仅成功模仿和解码时间多样化控制。
cs.CL / 42 / 2609.10113
Data-Centric Post-Training for Financial Reasoning: Mining, Distillation, and Verifiable Learning
以数据为中心的金融推理后训练：挖掘、蒸馏与可验证学习
Abstract
Financial text, textbooks, and question-answer pairs are abundant, but only a small fraction is directly usable for reasoning-focused post-training. Existing QA pairs often lack explicit reasoning, sufficient context, or reliably verifiable answers, while textbooks must first be transformed into synthetic training examples. We present a data-centric pipeline that constructs complementary corpora by mining open-source reasoning traces, distilling financial instruction data, and generating knowledge-graph-guided question-answer pairs from financial educational material. After semantic deduplication, three lightweight sequence classifiers select finance-relevant examples, reject under-specified questions, and identify tasks suitable for reinforcement learning with compact rule-based verifiers. For model adaptation, we study supervised fine-tuning and reinforcement learning, while self-distilled fine-tuning and post-training model merging are used to prevent the loss of financial capabilities already present in the starting model. We evaluate the adapted language models using FINESSE-Bench, reporting aggregate performance and changes relative to their starting checkpoints. Across the selected comparisons, ordinary SFT reduces FINESSE-Bench accuracy by 3.2-4.0 percentage points, whereas self-distilled SFT improves over the corresponding starting models by 1.0-2.8 points. Equal-weight merging recovers 3.0 points over its SFT parent and finishes 0.9 points above the original model; GRPO on hard tasks adds 0.4 points after self-distilled SFT or 3.0 points when applied directly to verifiable tasks. These results show that retention-aware adaptation can improve financial reasoning without the regressions observed after ordinary SFT.
Chinese Translation
金融文本、教科书和问答对丰富，但只有一小部分可以直接用于以推理为重点的后训练。现有的问答对通常缺乏明确的推理、足够的上下文或可靠的可验证答案，而教科书必须首先转化为合成训练示例。我们提出了一种以数据为中心的流程，通过挖掘开源推理痕迹、蒸馏金融指令数据以及从金融教育材料中生成知识图谱引导的问答对，构建互补的语料库。在语义去重后，三个轻量级序列分类器选择与金融相关的示例，拒绝不明确的问题，并识别适合使用紧凑的基于规则的验证器进行强化学习的任务。对于模型适应性，我们研究了监督微调和强化学习，同时使用自蒸馏微调和后训练模型合并来防止丢失起始模型中已具备的金融能力。我们使用FINESSE-Bench评估适应后的语言模型，报告相对于其起始检查点的整体性能和变化。在所选比较中，普通的监督微调（SFT）使FINESSE-Bench的准确率降低了3.2-4.0个百分点，而自蒸馏的SFT则使相应的起始模型提高了1.0-2.8个百分点。等权重合并相对于其SFT父模型恢复了3.0个百分点，并且比原始模型高出0.9个百分点；在困难任务上，GRPO在自蒸馏SFT后增加了0.4个百分点，或在直接应用于可验证任务时增加了3.0个百分点。这些结果表明，关注保留的适应性可以改善金融推理，而不会出现普通SFT后观察到的回归。
cs.CL / 43 / 2609.10122
ProbPlug: A Plugin Uncertainty Network for Reliable Confidence in LLM Binary Classification
ProbPlug：一种插件不确定性网络，用于提高大型语言模型二分类的可靠性
Abstract
Large language models (LLMs) have achieved strong performance across a broad range of classification settings, yet the reliability of their predictions remains a major obstacle to deployment in high-stakes scenarios. Although confidence estimation for LLMs has been widely studied, confidence calibration for LLM-based classification remains underexplored. We introduce ProbPlug, a lightweight confidence estimation framework for LLM-based binary classification, which predicts whether an output is correct using internal token features extracted from a frozen LLM. ProbPlug employs a self-attention module to aggregate hidden representations and can be integrated into the original inference pipeline without modifying the base model. Experiments across multiple tasks involving both text-based and multimodal large models show that ProbPlug provides more reliable confidence estimates, improves classification performance with negligible additional overhead, and exhibits strong generalization across tasks. These results indicate that ProbPlug serves as a practical solution for confidence estimation in LLM-based classification. Our code is publicly available at Github.
Chinese Translation
大型语言模型（LLMs）在广泛的分类场景中表现出色，但其预测的可靠性仍然是高风险场景部署的主要障碍。尽管对LLMs的置信度估计已经得到了广泛研究，但基于LLM的分类的置信度校准仍然未得到充分探索。我们提出了ProbPlug，一种轻量级的置信度估计框架，旨在用于基于LLM的二分类任务，该框架利用从冻结的LLM中提取的内部标记特征来预测输出是否正确。ProbPlug采用自注意力模块来聚合隐藏表示，并且可以在不修改基础模型的情况下集成到原始推理管道中。针对多个任务的实验，包括基于文本和多模态的大型模型，结果表明ProbPlug提供了更可靠的置信度估计，改善了分类性能且额外开销微乎其微，并在不同任务中展现出强大的泛化能力。这些结果表明，ProbPlug为基于LLM的分类中的置信度估计提供了一个实用的解决方案。我们的代码已在Github上公开。
cs.CL / 44 / 2609.10142
Active Adaptation, Not Static Defense: Temporal Dynamics of Preventative Steering in Adversarial Fine-Tuning
主动适应，而非静态防御：对抗微调中预防性引导的时间动态
Abstract
Large language models remain fragile against malicious fine-tuning, motivating training-time defenses against harmful persona drift. Preventative Steering injects undesirable-trait persona vectors during fine-tuning and removes them at evaluation time, yet the mechanism behind its lasting protection remains unclear. Analyzing its temporal optimization dynamics, we find that the defense emerges from an early compensatory adaptation phase followed by a steady-state phase where the corrective signal decays; in parameter space, attention output projections emerge as the dominant residual-write route for defensive updates. Through Intervention Delta Preservation (IDP) and IDP Continuation experiments, we further show that preserving or reinjecting the weight offset fails to maintain protection, indicating that preventative steering relies on active adaptation rather than a static defense. Motivated by this finding, we propose Progressive Intensity Scheduling (PIS), which starts with a moderate injection strength and increases it after static-strength alignment begins to decay. Across the evaluated Qwen2.5 and Gemma-3 models, PIS improves safety robustness over static-strength steering while reducing harmful trait expression.
Chinese Translation
大型语言模型在恶意微调面前仍然脆弱，这促使我们在训练阶段采取防御措施以应对有害的个性漂移。预防性引导在微调过程中注入不良特征的个性向量，并在评估时将其移除，但其持久保护机制仍不明确。通过分析其时间优化动态，我们发现该防御机制源于早期的补偿适应阶段，随后进入一个稳态阶段，在该阶段中，纠正信号逐渐衰减；在参数空间中，注意力输出投影成为防御更新的主要残差写入路径。通过干预增量保持（Intervention Delta Preservation, IDP）和IDP延续实验，我们进一步表明，保持或重新注入权重偏移无法维持保护，表明预防性引导依赖于主动适应而非静态防御。基于这一发现，我们提出了渐进强度调度（Progressive Intensity Scheduling, PIS），该方法从适度的注入强度开始，并在静态强度对齐开始衰减后增加强度。在评估的Qwen2.5和Gemma-3模型中，PIS在提高安全鲁棒性的同时，减少了有害特征的表达。
cs.CL / 45 / 2609.10153
YallaMorph: A Benchmark for Evaluating Arabic Morphological Generation in Large Language Models
YallaMorph：评估大型语言模型中阿拉伯语形态生成的基准
Abstract
Arabic morphology remains challenging for large language models, since fluent generation does not guarantee accurate morphosyntactic control. Existing Arabic evaluations mainly target downstream tasks and do not directly test controlled morphological generation from explicit lexical and feature-based input. We introduce YallaMorph, a large-scale benchmark for Arabic morphological generation covering verbs, nouns, adjectives, their cliticized forms, and invalid configurations. We evaluate multilingual and Arabic-oriented LLMs under diacritized and undiacritized settings over 600K benchmark entries. Results show that Arabic morphological generation remains difficult, especially for cliticized, unseen, and morphologically rare forms.
Chinese Translation
阿拉伯语形态学对大型语言模型而言仍然具有挑战性，因为流畅的生成并不保证准确的形态句法控制。现有的阿拉伯语评估主要针对下游任务，并未直接测试基于显式词汇和特征输入的受控形态生成。我们推出了YallaMorph，一个针对阿拉伯语形态生成的大规模基准，涵盖动词、名词、形容词及其附加形式和无效配置。我们在带有和不带有变音符号的设置下，对600K基准条目进行了多语言和阿拉伯语导向的大型语言模型（LLMs）评估。结果表明，阿拉伯语形态生成仍然困难，尤其是对于附加形式、未见形式和形态上稀有的形式。
cs.CL / 46 / 2609.10155
From Retrieval to Weights: Parametric Individualization of Small Language Models with Individual Text Corpora
从检索到权重：利用个体文本语料库对小型语言模型进行参数化个性化
Abstract
We approach a cognitive simulation perspective on episodic and semantic memory in multiple-choice question answering by incorporating text from individual text corpora (ITC) into retrieval-augmented generation and DoRA fine-tuning. We web-crawl the search histories of 515 participants who answered 36 multiple-choice knowledge items and analyze a stratified subsample of 150 participants. For each participant, one DoRA adapter consolidates their ITC into a small language model (SLM) whose baseline correctness falls below the participants' lowest quartile. The adapter measurably writes the ITC into the weights: it fits its own participant's held-out text better than other participants' texts (dz =1.27), an individuality effect that increases with ITC size in rank order. On the generalized knowledge test, however, the adapter adds knowledge rather than alignment with the individual: log-loss match improves, whereas match accuracy under a bias-corrected PMI readout does not, and retrieval adds nothing on top. Our results demonstrate that ITCs can be consolidated into the weights of SLMs, an encouraging basis for individualized tutoring agents, and we discuss how to move from there toward a realistic simulation of episodic and semantic memory at the individual level.
Chinese Translation
我们从认知模拟的角度探讨了在多项选择问答中情景记忆和语义记忆的作用，通过将个体文本语料库（ITC）中的文本纳入增强检索生成和DoRA微调中。我们对515名参与者的搜索历史进行了网络爬取，这些参与者回答了36个多项选择知识题，并分析了150名参与者的分层子样本。对于每位参与者，一个DoRA适配器将其ITC整合到一个小型语言模型（SLM）中，该模型的基线正确率低于参与者的最低四分位数。适配器在权重中可测量地写入ITC：它对自身参与者的保留文本的拟合优于其他参与者的文本（dz = 1.27），这种个体性效应随着ITC规模的增加而按等级顺序上升。然而，在广义知识测试中，适配器增加了知识，而不是与个体的对齐：对数损失匹配有所改善，而在经过偏差校正的PMI读取下，匹配准确率没有提高，检索也没有额外贡献。我们的结果表明，ITC可以整合到SLM的权重中，为个性化辅导代理提供了一个令人鼓舞的基础，我们讨论了如何从这里出发，朝着个体层面上对情景记忆和语义记忆的真实模拟迈进。
cs.CL / 47 / 2609.10192
Who Argues What? Joint Argument-Entity Detection and Classification in Political Debates
谁在争论什么？政治辩论中的联合论点-实体检测与分类
Abstract
Political debates are often analyzed through Argument Mining (AM) to investigate the key arguments that drive them. However, political arguments are rarely interpretable from argumentative spans alone, as claims and premises generally depend on the entities (e.g., people, events, locations, parties) they mention. Existing AM resources and methods typically annotate argumentative spans and roles, but do not provide a paired debate-entity layer for asking which Debate Named Entities (DNE), e.g., actors and events, are invoked within debates. In this work, we address these data and methodological gaps by (i) introducing DNE-ElecDeb, an entity-enriched version of the USElecDeb dataset that adds DNEs in both argumentative and non-argumentative spans and defines Debate Named Entity Recognition (DNER) as the task of detecting DNEs, and (ii) proposing Joint Argument and Entity Tagging (JAET), a generative framework that fine-tunes decoder-only LLMs to insert inline argument and entity tags into debate turns while preserving the original transcript. Under BIO-tagging evaluation, JAET improves relative F1 on the joint AM+DNER task by +27.3%, resp. +41.9%, under the untyped, resp. typed setting over the strongest sequential AM-DNER pipelines, demonstrating that such gains cannot be recovered by composing two independent modules. Notably, similar margins replicate on Persuasive Essays (+26.6%, resp. +52.7%), showing effective generalization to domains orthogonal to political debates. By unifying argumentative and entity-level representations within a single view, our contributions pave the way for richer political debates understanding.
Chinese Translation
政治辩论通常通过论点挖掘（Argument Mining, AM）进行分析，以探讨驱动辩论的关键论点。然而，仅从论证跨度中很少能解释政治论点，因为主张和前提通常依赖于它们提及的实体（例如，人物、事件、地点、政党）。现有的 AM 资源和方法通常注释论证跨度和角色，但未提供配对的辩论-实体层，以询问在辩论中提及了哪些辩论命名实体（Debate Named Entities, DNE），例如，参与者和事件。在本研究中，我们通过以下方式解决这些数据和方法论的空白：（i）引入 DNE-ElecDeb，这是 USElecDeb 数据集的一个实体增强版本，增加了论证和非论证跨度中的 DNE，并将辩论命名实体识别（Debate Named Entity Recognition, DNER）定义为检测 DNE 的任务；（ii）提出联合论点和实体标记（Joint Argument and Entity Tagging, JAET），这是一个生成框架，通过微调仅解码器的 LLM，将内联论点和实体标签插入辩论回合，同时保留原始记录。在 BIO 标记评估下，JAET 在联合 AM+DNER 任务上相对于最强的顺序 AM-DNER 流水线提高了 +27.3% 和 +41.9% 的相对 F1，分别在未标记和已标记设置下，证明这种增益无法通过组合两个独立模块来恢复。值得注意的是，类似的增益幅度在说服性论文中也得到了复制（+26.6% 和 +52.7%），显示出对与政治辩论正交领域的有效泛化。通过在单一视图中统一论证和实体级别的表示，我们的贡献为更丰富的政治辩论理解铺平了道路。
cs.CL / 48 / 2609.10198
Politics of Feelings: Emotional Expression and Legislative Effectiveness in the U.S. Congress
情感政治：美国国会中的情感表达与立法有效性
Abstract
Emotions are a pervasive feature of political communication, yet existing research has focused primarily on describing patterns of emotional expression rather than examining whether they are associated with consequential legislative outcomes. We address this gap by investigating the expression and correlates of discrete emotions in more than 1.7 million speeches delivered in the U.S. Congress between 1973 and 2024. Using a transformer-based emotion classifier, we measure eight discrete emotions: anger, fear, disgust, sadness, joy, enthusiasm, pride, and hope. We examine how these emotions vary over time, across policy topics, legislator characteristics, and their relationship with legislative effectiveness. We find that congressional speeches are becoming emotionally expressive over time. Emotional expression also varies systematically across policy domains and ideological positioning of legislators. Notably, the relationship between emotional expression and legislative effectiveness depends on the specific emotions expressed: enthusiasm and pride are positively associated with effectiveness, whereas anger exhibits a negative association. Emotional valence and emotional diversity are positively associated with legislative effectiveness, while emotional intensity is negatively associated with legislative effectiveness. These findings demonstrate that computationally derived measures of discrete emotions can provide insight into affective dimensions of legislative speeches and facilitate our understanding of how legislators communicate, interact, and perform within democratic institutions.
Chinese Translation
情感是政治传播中的普遍特征，但现有研究主要集中在描述情感表达的模式，而非考察这些模式是否与重要的立法结果相关。我们通过研究1973年至2024年间美国国会发表的超过170万次演讲中的情感表达及其相关因素，填补了这一空白。使用基于变换器的情感分类器，我们测量了八种离散情感：愤怒、恐惧、厌恶、悲伤、快乐、热情、自豪和希望。我们考察了这些情感如何随时间、政策主题、立法者特征的变化而变化，以及它们与立法有效性的关系。研究发现，国会演讲的情感表达随着时间的推移而变得更加丰富。情感表达在不同政策领域和立法者的意识形态定位之间也存在系统性的差异。值得注意的是，情感表达与立法有效性之间的关系取决于所表达的具体情感：热情和自豪与有效性呈正相关，而愤怒则表现出负相关。情感的价值倾向和情感多样性与立法有效性呈正相关，而情感强度则与立法有效性呈负相关。这些发现表明，基于计算的方法获得的离散情感测量可以为立法演讲的情感维度提供洞察，并促进我们理解立法者在民主制度中如何沟通、互动和表现。
cs.CL / 49 / 2609.10210
Through the Looking Glass: Directly Reading and Writing Transformers
透过镜子：直接读取和写入变换器
Abstract
How many of a transformer's components decide a token? Counted by the absolute value of each unit's and channel's contribution to the logit, one prediction rests on thousands to hundreds of thousands of them. But contributions are signed, and across eighteen models the mass pushing away from the predicted token is a median of seven times the mass carrying it. Divide by the net and the count is dozens: on the baseline, 53 components carry ninety percent of a prediction, 13 it cannot survive losing, and 8 suffice to produce it alone. Across twelve models trained elsewhere, 124M to 7B parameters, the sufficient set runs from two components to sixteen, and what a prediction draws on, followed all the way back, is one to three percent of the model, a share that does not grow with size. Three quarters of a layer's update is a fixed linear map of the state it received. Everything is read from the model's own parameters and activations, with nothing trained or fitted, and it names a component on both sides: what it writes, from the predictions it drives, reaching close to half of every model; what it reads, from its weights in the frame of its own layer, at 58.9 percent above chance over its eight strongest inputs. Sorting the remainder by upstream source yields grammatical categories the embedding cannot see. A name can be acted on. An association the model does not hold installs into one spare unit, key and value read from the weights, for a quarter of a percent of held-out loss, a fortieth of what a rank-one update costs. An installed attention head and a unit two layers above it make an edit fire only where a token occurred earlier in the context, and a unit the model trained for itself is driven from two layers upstream, 86 percent of the effect passing through it. An order-preserving activation puts a unit's inputs at the instrument's ceiling, at the price of a two-part install.
Chinese Translation
变换器的多少个组件决定一个标记？通过计算每个单元和通道对对数几率的绝对贡献值，一个预测依赖于成千上万到数十万的组件。然而，贡献是有符号的，在十八个模型中，推向预测标记的质量中位数是推离它的质量的七倍。将其除以净值，结果是几十个：在基线模型中，53个组件承载了90%的预测，13个是其无法承受失去的，8个足以单独产生预测。在其他地方训练的十二个模型中，参数从1.24亿到70亿，足够的组件集从两个组件到十六个不等，而一个预测所依赖的，追溯到模型的1%到3%，这一比例并不随模型规模的增加而增长。一个层的更新中有四分之三是其接收到的状态的固定线性映射。所有信息均来自模型自身的参数和激活，没有经过训练或拟合，并且在两侧都命名了一个组件：它所写的，从它驱动的预测中，接近每个模型的一半；它所读取的，来自其自身层框架中的权重，超过其八个最强输入的58.9%的机会水平。按上游来源对其余部分进行排序，得到了嵌入无法看到的语法类别。一个名称可以被操作。模型不持有的关联被安装到一个备用单元中，键和值从权重中读取，导致四分之一的持出损失，相当于一次秩为一的更新成本的四十分之一。一个安装的注意力头和其上方两层的单元使得编辑仅在上下文中早期出现的标记处触发，而模型为自己训练的单元则由上游两层驱动，86%的效果通过它传递。一个保持顺序的激活将单元的输入置于工具的上限，代价是一个两部分的安装。
cs.CL / 50 / 2609.10226
$\Phi$-Bench: Can Large Language Models Engineer the Infrastructure That Powers Them?
$ ext{Φ}$-Bench：大型语言模型能否设计支撑它们的基础设施？
Abstract
Large language models (LLMs) have demonstrated remarkable capabilities in reasoning and code generation, raising the prospect that they could assist in developing and optimizing the very infrastructure that powers them. However, existing benchmarks mainly focus on isolated kernels, predefined operators, or pre-specified optimization targets, and therefore fail to evaluate the ability of LLMs to perform open-ended, long-horizon LLM infrastructure engineering. To address this gap, we present $\Phi$-Bench, a benchmark for systematically evaluating LLMs on engineering the LLM infrastructure stack. Derived from optimization problems studied in frontier research and grounded in real-world code repositories, $\Phi$-Bench provides broad coverage of the LLM infrastructure stack and spans tasks of varying complexity, ranging from localized kernel-level function completion to long-horizon implementation and end-to-end system optimization. Extensive experiments on frontier LLMs reveal their current capabilities and limitations in engineering complex LLM infrastructure, offering insights into the challenges that remain on the path toward autonomous optimization of future AI infrastructure.
Chinese Translation
大型语言模型（LLMs）在推理和代码生成方面展现了显著的能力，这引发了它们可能协助开发和优化支撑自身的基础设施的前景。然而，现有的基准测试主要集中在孤立的内核、预定义的操作符或预设的优化目标上，因此未能评估LLMs在开放式、长期的LLM基础设施工程方面的能力。为了解决这一问题，我们提出了$ ext{Φ}$-Bench，这是一个系统评估LLMs在工程LLM基础设施栈方面的基准测试。该基准源于前沿研究中研究的优化问题，并基于真实的代码库，提供了对LLM基础设施栈的广泛覆盖，涵盖了从局部内核级函数完成到长期实现和端到端系统优化的各种复杂任务。对前沿LLMs的广泛实验揭示了它们在工程复杂LLM基础设施方面的当前能力和局限性，并提供了对未来AI基础设施自主优化过程中仍然存在的挑战的见解。
cs.CL / 51 / 2609.10237
The Answer Path and the Grounding Instruction in LLM Question Answering over Knowledge Graphs
大型语言模型在知识图谱问答中的答案路径与基础指令
Abstract
A graph retrieval-augmented generation pipeline chooses which triples to put in the prompt, a syntax to write them in, an order to write them in, and a sentence telling the model what to do with them. We vary all four over six large language models and two knowledge-graph question answering benchmarks. Two of the four choices move the answer and the other two are flat. The first is whether the answer path, the triples needed to reach the answer, is in the prompt at all. Holding the number of triples fixed and replacing every triple that is not on the chain with material from an unrelated entity changes answer accuracy by +0.003 F1, while removing the chain costs most of what the graph was worth. Retrieval budget belongs on recall, and precision in the range we can test buys nothing. There is no retriever here: subgraphs come from gold SPARQL, so precision describes the context we build, not a system setting. The second is the grounding instruction. With no facts in the prompt, telling a model to answer using only the provided facts drops F1 from 0.299 to 0.035, a factor of 8.63. That figure describes an evaluation with an empty context arm rather than a working pipeline, and an experiment that applies the instruction to its context arm but not to its no-context baseline manufactures a spurious finding that graph context hurts at depth. We found one in our own results and retract it. Syntax, triple order and subgraph size produce no effect we can measure at multi-hop depth. The comparison that would price the grounding instruction against correct context is not measurable with a format-sensitive scorer, because the instruction determines the response format; we report it as an open contrast rather than a number.
Chinese Translation
图检索增强生成管道选择在提示中放入哪些三元组、以何种语法书写它们、以何种顺序书写以及告诉模型如何处理它们的句子。我们在六个大型语言模型和两个知识图谱问答基准上变更这四个选择。四个选择中的两个是移动答案的，而另外两个是平坦的。第一个选择是答案路径，即到达答案所需的三元组是否在提示中。固定三元组数量并用来自不相关实体的材料替换链上每个不在链上的三元组，会使答案准确率提高 +0.003 F1，而移除链则几乎损失了图的全部价值。检索预算应该关注召回率，而我们可以测试的精确度范围内并没有任何收益。这里没有检索器：子图来自黄金 SPARQL，因此精确度描述我们构建的上下文，而不是系统设置。第二个选择是基础指令。在提示中没有事实的情况下，告诉模型仅使用提供的事实回答会使 F1 从 0.299 降至 0.035，下降幅度为 8.63 倍。该数字描述的是一个空上下文臂的评估，而不是一个有效的管道，且将指令应用于其上下文臂而不应用于其无上下文基线的实验制造了一个虚假的发现，即图上下文在深度上会造成损害。我们在自己的结果中发现了这一点并撤回。语法、三元组顺序和子图大小在多跳深度上没有产生可测量的效果。将基础指令与正确上下文进行比较的方式无法通过格式敏感的评分器进行测量，因为该指令决定了响应格式；我们将其报告为开放对比，而不是一个数字。
cs.CL / 52 / 2609.10244
Two-Token Features and Small-Large Ensembles for VLM Hallucination Detection
双标记特征与小大规模集成用于VLM幻觉检测
Abstract
We present our system for the SHROOM-Visions 2026 shared task on character-level VLM hallucination detection. A small ($4$B-parameter) VLM is fine-tuned as a per-token classifier reading a two-token feature from its own hidden states, and is ensembled with a $\sim$400B zero-shot VLM judge at prediction time. Both components see off-the-shelf OCR of any visible in-image text. We use synthetic hallucination data generated by the large model as a source of ensemble diversity, and use validation to select feature layer, training data and OCR grounding. Our official entry reaches mean Cor $0.487$ / Cor-lbl $0.387$ on the hidden test set, placing $6$th/$28$ (EN), $6$th/$21$ (FR), $8$th/$21$ (IT) and $7$th/$22$ (ZH) on the task's primary Cor-lbl metric.
Chinese Translation
我们提出了用于SHROOM-Visions 2026共享任务的系统，该任务涉及字符级VLM幻觉检测。一个小型（$4$B参数）VLM被微调为逐标记分类器，从其自身的隐藏状态中读取双标记特征，并在预测时与一个约$400B$的零-shot VLM评判器进行集成。两个组件均使用现成的光学字符识别（OCR）技术来处理任何可见的图像文本。我们使用由大型模型生成的合成幻觉数据作为集成多样性的来源，并利用验证选择特征层、训练数据和OCR基础。我们的官方参赛作品在隐藏测试集上达到了平均相关系数Cor $0.487$ / Cor-lbl $0.387$，在任务的主要Cor-lbl指标中排名第$6$/$28$（英语），第$6$/$21$（法语），第$8$/$21$（意大利语）和第$7$/$22$（中文）。
cs.CL / 53 / 2609.10253
DiSCo: A Distribution-First Steering and Cultural Prior Evaluation Framework for Measuring Cultural Preference Bias in LLMs
DiSCo：一种以分布为先的引导和文化偏好评估框架，用于测量大型语言模型中的文化偏好偏差
Abstract
Large language models (LLMs) are increasingly deployed in globally used assistants, yet their default choices in culturally grounded everyday situations can systematically favour some cultures over others, affecting localisation, user trust, and equitable behaviour. Existing cultural benchmarks evaluate accuracy against a single "correct" answer, making it difficult to characterise an LLM's cultural preference prior when multiple culturally grounded responses are all valid; they also conflate default preferences with context-driven adaptation. We propose DiSCo, a distribution-first forced-choice evaluation framework that isolates default cultural priors and tests steerability via a four-level context gradient (C0--C3). Using DiSCo-Bench (304 items) derived from BLEnD spanning 12 cultures, we evaluate six diverse instruction-tuned LLMs. Default priors are heavily concentrated, with UK and US together absorbing approximately 35\% of all selections despite representing only 2 of 12 cultures. Most critically, prompt-based steering consistently widens the selection gap between high- and low-resource cultures, and injecting explicit cultural facts produces negligible distributional disruption, confirming that cultural preference bias cannot be resolved through prompt-based personalisation alone.
Chinese Translation
大型语言模型（LLMs）越来越多地被应用于全球使用的助手中，但它们在文化背景的日常情境中的默认选择可能系统性地偏向某些文化，从而影响本地化、用户信任和公平行为。现有的文化基准通过与单一“正确”答案的比较来评估准确性，这使得在多个文化背景的有效响应中，难以表征LLM的文化偏好先验；它们还将默认偏好与基于上下文的适应混为一谈。我们提出了DiSCo，这是一种以分布为先的强制选择评估框架，旨在隔离默认文化先验并通过四级上下文梯度（C0--C3）测试引导性。使用从BLEnD衍生的DiSCo-Bench（304个项目），涵盖12种文化，我们评估了六种不同的指令调优LLM。默认先验高度集中，尽管仅代表12种文化中的2种，英国和美国的选择总共占约35%。最关键的是，基于提示的引导始终扩大了高资源文化与低资源文化之间的选择差距，而注入明确的文化事实几乎没有产生分布性干扰，确认了文化偏好偏差不能仅通过基于提示的个性化来解决。
cs.CL / 54 / 2609.10266
KVShareArena: KV-Cache Reuse Across Contexts and Model Checkpoints
KVShareArena：跨上下文和模型检查点的KV缓存重用
Abstract
LLM serving systems already reuse KV caches, but only when the reused text sits at the very start of the prompt. Two growing workloads break this condition: a retrieval-augmented generation server assembles a different set of retrieved chunks for every query, and a multi-agent coordinator reads reports written by other agents. Reused inside a new prompt, a cache carries the wrong positions and never attended to the other sources. The cache may also have been written by a different checkpoint of the same model family, which changes the stored values. Repair methods for such caches have appeared in three separate communities, each measured on its own terms, and existing benchmarks test only exact-prefix reuse, where nothing is lost. KVShareArena benchmarks KV-cache reuse across prompt contexts and model checkpoints on retrieved chunks and agent reports. It scores every method by the fraction of the gap it recovers between no cache and full recomputation, and charges compute, memory, and per-request latency with the cache in hand, reporting the one-time cost of building a cache separately. We find that correcting positions, which needs no recomputation, is enough until a question needs several sources at once. There, only methods that pay, by re-encoding part of the cache or by training, recover half to two thirds of the gap; unrepaired caches can be worse than no cache. Cache-compression methods that are harmless on a single prompt fall significantly behind position correction on freshly written agent reports. These patterns hold across three model boards. When a different checkpoint wrote the cache, training-free methods are barely affected, while an adapter trained on one checkpoint's caches loses quality. Harness, frozen querysets, and cost accounting ship as a pip package with an automated submission workflow and a public leaderboard.
Chinese Translation
大型语言模型（LLM）服务系统已经能够重用KV缓存，但仅在重用的文本位于提示的最开始时才有效。两种日益增长的工作负载打破了这一条件：一个检索增强生成服务器为每个查询组装一组不同的检索块，而一个多智能体协调器读取其他智能体撰写的报告。在新的提示中重用的缓存携带了错误的位置，并且从未关注其他来源。缓存也可能是由同一模型系列的不同检查点写入的，这会改变存储的值。针对这种缓存的修复方法已在三个不同的社区中出现，每个社区都以自己的标准进行评估，而现有基准测试仅测试精确前缀重用的情况，在这种情况下没有任何信息丢失。KVShareArena基准测试在检索块和智能体报告的提示上下文和模型检查点之间的KV缓存重用。它通过恢复无缓存与完全重新计算之间的差距的比例来对每种方法进行评分，并计算使用缓存时的计算、内存和每请求延迟，同时单独报告构建缓存的一次性成本。我们发现，纠正位置（不需要重新计算）足以应对，直到一个问题需要同时多个来源。在这种情况下，只有通过重新编码部分缓存或进行训练来付费的方法才能恢复一半到三分之二的差距；未修复的缓存可能比没有缓存更糟。对于单个提示无害的缓存压缩方法在新写的智能体报告上显著落后于位置纠正。这些模式在三个模型板上都成立。当不同的检查点写入缓存时，无需训练的方法几乎不受影响，而在一个检查点的缓存上训练的适配器则会失去质量。Harness、冻结查询集和成本核算作为一个pip包与自动提交工作流和公共排行榜一起发布。
cs.CL / 55 / 2609.10293
GANDR: Claim Auditing for Verifiable Legal Answer Generation
GANDR：可验证法律答案生成的声明审计
Abstract
In high-stakes domains such as legal practice, a language-model answer is only useful to the extent that a reader can verify each claim against the source the system cites. Current grounded-generation pipelines score the answer as a whole, so a correct conclusion can rest on fabricated or loosely matched citations and still score well. Closing this gap requires both a system built for per-claim verification and an evaluation that measures it. We introduce GANDR (Grounded ANswer DRafter), a two-agent system in which a Drafter writes an answer in a structured legal-reasoning format and a separate Critic, with the same view as a human verifier, audits each claim against its cited source and emits a per-claim audit trace on every round. We pair it with a strict correctness criterion requiring every citation to resolve to a passage the retriever returned. On a 185-item legal benchmark where all six systems share one backbone, one retrieval surface, and one citation instruction, GANDR ranks first on every primary metric, reaching 70.8% strict accuracy and leading the strongest baseline by 11.3 points (p<0.01). Reverting the protocol-anchored commit rule lowers strict accuracy by 22.7 points, and the strict lead stays positive on three further backbones, at +3.2 to +6.5 points. This lead traces to the Drafter configuration and the protocol-anchored commit, not to rewriting. Against two law-trained annotators the audit flags under-supported claims at F1 0.84 as a binary detector, while its four-way verdict labels agree only weakly and are advisory. Code is available upon request.
Chinese Translation
在法律实践等高风险领域，语言模型生成的答案只有在读者能够根据系统引用的来源验证每个声明的情况下才有用。目前的基于事实生成流程对答案整体进行评分，因此一个正确的结论可能依赖于虚构或松散匹配的引用，仍然能够获得良好的评分。弥补这一差距需要一个专门为每个声明验证而构建的系统，以及一个衡量该验证的评估方法。我们提出了GANDR（Grounded ANswer DRafter），这是一个双代理系统，其中一个撰写者以结构化的法律推理格式撰写答案，而另一个批评者则以人类验证者的视角审计每个声明与其引用来源的匹配情况，并在每一轮中发出逐条审计记录。我们将其与严格的正确性标准相结合，要求每个引用都必须对应于检索器返回的段落。在一个包含185个项目的法律基准测试中，所有六个系统共享一个基础架构、一个检索表面和一个引用指令，GANDR在每个主要指标上均排名第一，严格准确率达到70.8%，领先最强基线11.3个百分点（p<0.01）。撤销基于协议的提交规则使严格准确率下降22.7个百分点，而在另外三个基础架构上，严格领先仍保持正值，范围为+3.2到+6.5个百分点。这一领先优势源于撰写者配置和基于协议的提交，而非重写。与两位法律训练的注释者相比，审计在F1 0.84的二元检测器下标记了支持不足的声明，而其四向裁决标签仅弱相关，并具有建议性质。代码可根据请求提供。
cs.CL / 56 / 2609.10296
The Semantic Bottleneck: Leveraging Semantic Representations for Non-Invasive Speech Decoding
语义瓶颈：利用语义表示进行非侵入式语音解码
Abstract
Non-invasive speech decoding remains constrained by the low signal-to-noise ratio of neural recordings, which makes fine-grained reconstruction of phonemes or individual words difficult. Motivated by neuroscientific evidence that high-level semantic representations are distributed across cortical regions and evolve over slower temporal scales, we hypothesize that semantic content may provide a more suitable target for non-invasive decoding than low-level acoustic or lexical features. We introduce Brain2Semantics2Text, a method that reconstructs text through an intermediate semantic embedding space. Our model maps sentence-level MEG responses into a semantic manifold and then inverts the predicted embeddings into natural language. This semantic bottleneck enables recovery of high-level meaning without word-level alignment. We describe the core principles of the approach, its implementation, and the strategies used to mitigate the challenges of learning a reliable neural-to-semantic mapping. Finally, we compare against prior non-invasive Brain2Text methods and show improved sentence-level results.
Chinese Translation
非侵入式语音解码受到神经记录的低信噪比的限制，这使得对音素或单个词的细粒度重建变得困难。受到神经科学证据的启发，即高级语义表示分布在皮层区域并在较慢的时间尺度上演变，我们假设语义内容可能比低级声学或词汇特征更适合用于非侵入式解码。我们提出了Brain2Semantics2Text，一种通过中间语义嵌入空间重建文本的方法。我们的模型将句子级别的MEG（脑磁图）反应映射到语义流形中，然后将预测的嵌入反转为自然语言。这个语义瓶颈使得在没有词级对齐的情况下恢复高级意义成为可能。我们描述了该方法的核心原理、实现方式以及用于减轻学习可靠的神经到语义映射挑战的策略。最后，我们与之前的非侵入式Brain2Text方法进行了比较，并展示了改进的句子级结果。
cs.CL / 57 / 2609.10305
RiLM: Parameter-Efficient Language Modeling via Geodesic Decoding
RiLM：通过测地解码实现参数高效的语言建模
Abstract
Language models under one million parameters matter for edge deployment, domain adaptation, and reproducible research, yet a two-layer LSTM or Transformer at embedding width d = 128 still spends roughly one third of its capacity on the output matrix W_out in R^(d x |V|). We propose Riemannian Language Models (RiLM), which remove that layer entirely: context unfolds as a trajectory on a Riemannian manifold, and next-token probabilities arise from squared geodesic distance between the current state and vocabulary embeddings. The same embedding map serves input and output -- decoding is geometry. We instantiate the framework on flat R^d (Flat RiLM) and the Poincare ball H^d (HypRiLM) with a shared MLP composition map phi (~290k parameters, d = 128, |V| = 2000). Across five seeds on WikiText-2, HypRiLM reaches 54.2 +/- 0.2 validation perplexity versus 87.6 +/- 0.6 for Flat RiLM; tied and matched LSTM, Transformer, and SSM controls remain at 113-147 PPL on WT-2 -- HypRiLM leads by roughly 2x over the strongest tied recurrent baseline (SSM, 113.0 +/- 3.8). Penn Treebank and a 10k-vocabulary stress test confirm that geodesic decoding transfers across corpora and larger |V|, while hyperbolic curvature helps selectively. We also characterize boundary collapse in naive hyperbolic recurrence and show how Mobius stabilization restores trainability. Claims are scoped to controlled small-model comparisons, not full-vocabulary state of the art.
Chinese Translation
对于边缘部署、领域适应和可重复研究，百万参数以下的语言模型至关重要，但在嵌入宽度 d = 128 的双层 LSTM 或 Transformer 中，输出矩阵 W_out 在 R^(d x |V|) 上仍然大约消耗了三分之一的容量。我们提出了黎曼语言模型（RiLM），完全去除了这一层：上下文在黎曼流形上展开为一条轨迹，下一词的概率源于当前状态与词汇嵌入之间的平方测地距离。同一嵌入映射同时用于输入和输出——解码是几何学。我们在平坦的 R^d（Flat RiLM）和庞加莱球 H^d（HypRiLM）上实例化该框架，使用共享的 MLP 组合映射 phi（约 29 万参数，d = 128，|V| = 2000）。在 WikiText-2 上的五个种子实验中，HypRiLM 达到了 54.2 +/- 0.2 的验证困惑度，而 Flat RiLM 为 87.6 +/- 0.6；与 LSTM、Transformer 和 SSM 控制组的结果相匹配，后者在 WT-2 上的困惑度保持在 113-147 之间——HypRiLM 的表现大约是最强的绑定递归基线（SSM，113.0 +/- 3.8）的两倍。Penn Treebank 和一个 10k 词汇的压力测试确认了测地解码在不同语料和更大 |V| 上的迁移能力，同时超曲率有助于选择性地优化。我们还描述了天真的超曲率递归中的边界崩溃，并展示了 Mobius 稳定化如何恢复可训练性。声明的范围限于受控的小模型比较，而非全词汇的最新技术。
cs.CL / 58 / 2609.10321
On-Policy Distillation for Vision-Language Model Adaptation, an Effective Paradigm on Low-Quality Multimodal Data
基于策略的视觉-语言模型适应蒸馏：低质量多模态数据上的有效范式
Abstract
Knowledge distillation offers an efficient route to transfer a task-adapted vision-language teacher to a compact student. The training target in current vision-language distillation methods is typically constructed from the teacher prediction and applied uniformly to all training samples, making it unreliable under class and domain shifts. In this paper, we argue that distillation target construction should be treated as a dynamic training decision rather than a fixed recipe. To this end, we propose OnPoKD, an on-policy distillation framework for vision-language model adaptation. To the best of our knowledge, OnPoKD is the first framework that applies on-policy distillation to vision-language model adaptation by learning target construction as a policy decision. OnPoKD learns a lightweight controller that constructs sample-wise adaptive targets using reliability and disagreement cues from the teacher model, student model, and zero-shot prior. Instead of relying on a fixed teacher prediction, the controller dynamically balances teacher supervision, zero-shot prior guidance, and hard-label anchoring through bounded policy actions, allowing the distillation target to adapt to varying sample reliability and training stages. The policy controller is updated with validation feedback, encouraging target construction to optimize transferability rather than merely fitting the training distribution. Since the controller is only used during training, OnPoKD can be seamlessly integrated into existing vision-language distillation pipelines while preserving the original inference architecture and test-time cost. Extensive experiments on Base-to-novel generalization and Cross-dataset transfer benchmarks show that OnPoKD consistently improves over strong vision-language distillation baselines.
Chinese Translation
知识蒸馏为将任务适应的视觉-语言教师模型转移到紧凑的学生模型提供了一条高效的途径。目前的视觉-语言蒸馏方法中的训练目标通常是基于教师预测构建，并均匀应用于所有训练样本，这在类别和领域转移下显得不可靠。本文认为，蒸馏目标的构建应被视为一种动态的训练决策，而非固定的配方。为此，我们提出了OnPoKD，一个用于视觉-语言模型适应的基于策略的蒸馏框架。据我们所知，OnPoKD是第一个将基于策略的蒸馏应用于视觉-语言模型适应的框架，通过将目标构建学习作为策略决策。OnPoKD学习一个轻量级控制器，利用教师模型、学生模型和零样本先验中的可靠性和不一致性线索构建样本自适应目标。控制器动态平衡教师监督、零样本先验指导和硬标签锚定，通过有界策略动作使蒸馏目标适应不同样本的可靠性和训练阶段。策略控制器通过验证反馈进行更新，鼓励目标构建优化可转移性，而不仅仅是适应训练分布。由于控制器仅在训练期间使用，OnPoKD可以无缝集成到现有的视觉-语言蒸馏流程中，同时保留原始推理架构和测试时间成本。在基础到新颖的泛化和跨数据集转移基准上的大量实验表明，OnPoKD在强大的视觉-语言蒸馏基线之上始终表现出改进。
cs.CL / 59 / 2609.10395
Rosetta at AlexandriaX-2026: LoRA-Adapted NileChat for Context-Aware Dialectal Arabic Dialogue Translation
亚历山大X-2026的罗塞塔：针对上下文感知的方言阿拉伯语对话翻译的LoRA适配NileChat
Abstract
This paper describes the Rosetta system for Subtask 1 (Context-Aware English-to-Dialectal Arabic Dialogue Translation) of the AlexandriaX shared task, participating in both constrained and unconstrained tracks. The approach fine-tunes a LoRA adapter on NileChat-3B using structured system/user prompts that condition generation on dialect and dialogue context. For the unconstrained track, the adapter is additionally pretrained on MADAR and PADIC. Rosetta ranked 4th in the constrained track (spBLEU 26.10) and 5th in the unconstrained track (spBLEU 25.09). The experimental results demonstrate that external pretraining helps only two of thirteen dialects while slightly hurting overall performance, suggesting negative transfer.
Chinese Translation
本文描述了罗塞塔系统在亚历山大X共享任务的子任务1（上下文感知的英语到方言阿拉伯语对话翻译）中的应用，参与了受限和不受限两个赛道。该方法在NileChat-3B上微调了LoRA适配器，使用结构化的系统/用户提示来根据方言和对话上下文进行生成。对于不受限赛道，适配器还在MADAR和PADIC上进行了额外的预训练。罗塞塔在受限赛道中排名第4（spBLEU 26.10），在不受限赛道中排名第5（spBLEU 25.09）。实验结果表明，外部预训练仅对十三种方言中的两种有所帮助，同时略微影响整体性能，暗示了负迁移的现象。
cs.CL / 60 / 2609.10410
Can Foundation Models Moderate Online Content? Evaluating Instruction- vs. Example-Driven Policy Operationalization
基础模型能否调节在线内容？评估基于指令与基于示例的政策实施方式
Abstract
The growing complexity of content moderation policies presents a critical challenge for their consistent operationalization. While foundation models possess the basic capabilities needed to confront this challenge, whether they can reliably moderate online content remains an unanswered question. In this paper, we systematically compare two competing paradigms for Vision-Language Model (VLM) guidance: an instruction-driven approach where models reason from policy precepts, and an example-driven approach where they generalize from prior precedents. We ground this investigation in ModerationBench, a new benchmark of 4,000 manually annotated, in-the-wild posts from the Bluesky platform. Our experiments reveal that foundation models can substantially outperform Bluesky's deployed moderation system, nearly tripling its $F_1$ score (0.60 vs. 0.22) on Random Posts in the benchmark, with both instruction- and example-driven paradigms achieving comparable peak effectiveness. Our findings thus chart a path toward reliable and adaptable policy operationalization at scale.
Chinese Translation
内容调节政策日益复杂，给其一致性实施带来了重大挑战。尽管基础模型具备应对这一挑战所需的基本能力，但它们是否能够可靠地调节在线内容仍然是一个未解之谜。本文系统地比较了两种竞争性范式的视觉-语言模型（Vision-Language Model, VLM）指导方法：一种是基于指令的方法，模型根据政策原则进行推理；另一种是基于示例的方法，模型从先前的案例中进行概括。我们将这一研究建立在ModerationBench上，这是一个包含来自Bluesky平台的4,000个手动标注的真实帖子的新基准。实验结果表明，基础模型在性能上显著优于Bluesky现有的内容调节系统，在基准测试中的随机帖子上，其$F_1$分数几乎是后者的三倍（0.60对比0.22），且基于指令和基于示例的范式都达到了相似的最佳效果。因此，我们的研究为大规模可靠且可适应的政策实施指明了方向。
cs.CL / 61 / 2609.10434
Do speech foundation models really learn words?
语音基础模型真的学习单词吗？
Abstract
Self-supervised speech foundation models are now used in a wide array of downstream applications, including traditional speech recognition and as the basis for tokens in speech-aware language models. Attempts to understand their usefulness have largely focused on probing their representations' ability to discriminate phonemes and words. However, discriminative ability for words need not imply specialized representation of words per se. Good discrimination of words may be explained by good encoding of word form (phonemes) rather than form-independent word representations encoding identity or syntactic/semantic properties. By partialling out phoneme information using residualization, we show that, in later layers, HuBERT and wav2vec 2.0 do in general learn representations which encode words with reasonable fidelity independently of local phonetic content. We show that this simple approach to disentanglement can enhance higher-order linguistic information in word discovery tasks.
Chinese Translation
自监督语音基础模型现在广泛应用于各种下游任务，包括传统的语音识别以及作为语音感知语言模型中的标记基础。对其有效性的理解尝试主要集中在探究其表示在区分音素和单词方面的能力。然而，单词的区分能力并不一定意味着对单词本身的专门表示。单词的良好区分可能是通过对单词形式（音素）的良好编码来解释的，而不是通过编码身份或句法/语义属性的形式无关的单词表示。通过使用残差化方法部分剔除音素信息，我们展示了在后期层中，HuBERT 和 wav2vec 2.0 确实一般学习到能够合理忠实地编码单词的表示，而不依赖于局部音韵内容。我们表明，这种简单的解耦方法可以增强单词发现任务中的高阶语言信息。
cs.CL / 62 / 2609.10445
Building Multilingual Bridges: Data Mixing as the Pillar of Generalization for In-Language Reasoning
构建多语言桥梁：数据混合作为语言内推理泛化的支柱
Abstract
Reasoning language models have made substantial advances on a variety of complex tasks, yet their capabilities remain overwhelmingly English-centric: models primarily reason in English regardless of the language they are prompted in. This is inaccessible for non-English-speaking users, risks losing the intent of the original question, and forgoes knowledge more readily expressed in the target language. In this work, we advance L2 reasoning, the ability of a model to reason consistently in the language of the user's prompt, thus building an in-language bridge between the prompt and the answer. We approach this problem from a data-centric angle, investigating how to optimize data composition and scheduling in SFT for reasoning generalization. Building Tiny Aya L2-Thinker at 3.35B scale, we achieve an L2 reasoning rate above 93% across 60 languages on 6 benchmarks spanning math, commonsense reasoning, instruction following, open-ended generation, and cultural reasoning while keeping performance strong. We show the path to generalizing L2 reasoning to held-out languages goes through broader language coverage, readily available multilingual non-reasoning data, and a sufficient English reasoning backbone. These findings indicate that reasoning is a language-agnostic behavior that can be transferred across typologically diverse languages through careful data mixing and without requiring reasoning supervision in every target language. We release our model weights and multilingual reasoning data to support further research on accessible, in-language reasoning.
Chinese Translation
推理语言模型在多种复杂任务上取得了显著进展，但其能力仍然以英语为中心：模型主要在英语中进行推理，而不论其提示的语言是什么。这对非英语用户来说是不可及的，可能导致原始问题意图的丧失，并放弃了在目标语言中更容易表达的知识。在本研究中，我们推进了L2推理，即模型在用户提示的语言中一致推理的能力，从而在提示和答案之间建立了语言内的桥梁。我们从以数据为中心的角度出发，研究如何优化数据组成和调度，以实现推理的泛化。通过构建规模为3.35B的Tiny Aya L2-Thinker，我们在涵盖数学、常识推理、指令遵循、开放式生成和文化推理的6个基准测试中，实现了60种语言的L2推理率超过93%，同时保持了强劲的性能。我们表明，将L2推理推广到未见语言的路径是通过更广泛的语言覆盖、 readily available 的多语言非推理数据以及足够的英语推理基础。这些发现表明，推理是一种语言无关的行为，可以通过仔细的数据混合在类型学上多样的语言之间转移，而不需要在每种目标语言中进行推理监督。我们发布了我们的模型权重和多语言推理数据，以支持对可访问的语言内推理的进一步研究。
cs.CL / 63 / 2609.10494
IBIB: A Protocol for Measuring Enterprise AI Systems by Serving Route, Not Model Identifier
IBIB：通过服务路径而非模型标识符测量企业人工智能系统的协议
Abstract
Enterprises deploy systems, not checkpoints. Usable capability depends jointly on weights, serving route, precision, output contract, and harness, yet all 18 audited benchmarks score advertised model identifiers. We treat this as measurement error and give a protocol that makes it reportable. It has three parts. A gold-blind capability-binding preflight verifies that a route can execute the evaluation contract before any task reaches it; a reliability-inclusive first-pass scoring rule keeps failure in the score while keeping unsupported capability out; and adjudication is structurally score-blind. We call the protocol IB2 and release its algorithms, classification tables, request contract, and manifest schemas. Its reference instantiation, 128 locked tasks and 987 assertions over document, spreadsheet, chart, tool and database work, stays sealed: the procedure is the artifact, not the corpus. Across eleven systems, four results. Capability availability is measurable: two complete single-route runs on identical weights later failed distinct predicates of the finalized binding gate, while a third passed that gate before a fresh run. The advertised identifier exposed neither limit. Discrimination is not uniform: four of seven suites saturate under a six-system band, with the spread almost entirely from governed database work and multi-tab joins, so we report interval-backed resolution groups, not ranks; two of the nominal five-label output's four cuts fail multiplicity adjustment. Serving-arm choice moved one declared revision and precision from 77.38 to 82.54, paired interval [0.11,10.60], though the arms differ in access mode, harness generation, and the serving tool-call parser, and harness generation is a property of our evaluator, not any endpoint. Excluding failed responses from denominators changes the point ordering, so reliability inclusion changes a conclusion, not its wording.
Chinese Translation
企业部署的是系统，而非检查点。可用能力共同依赖于权重、服务路径、精度、输出合同和工具，但所有18个审核基准均以广告的模型标识符进行评分。我们将其视为测量误差，并提供一个可报告的协议。该协议分为三个部分。金盲能力绑定的预检验证在任何任务到达之前，确保路径能够执行评估合同；包含可靠性的首次评分规则在保持失败计入分数的同时，排除不支持的能力；裁定结构上是盲评分的。我们将该协议称为IB2，并发布其算法、分类表、请求合同和清单模式。其参考实例化包括128个锁定任务和987个关于文档、电子表格、图表、工具和数据库工作的断言，保持封闭：程序是工件，而非语料库。在十一种系统中，得出了四个结果。能力可用性是可测量的：在相同权重下，两个完整的单路径运行后，分别未能满足最终绑定门的不同谓词，而第三个在新运行之前通过了该门。广告标识符未暴露任何限制。区分性并不均匀：七个套件中的四个在六个系统的带宽下饱和，差异几乎完全来自受控数据库工作和多标签连接，因此我们报告区间支持的分辨组，而非排名；名义上五个标签输出的四个切割中，有两个未能进行多重性调整。服务臂的选择使得一个声明的修订和精度从77.38提升至82.54，配对区间为[0.11,10.60]，尽管这些臂在访问模式、工具生成和服务工具调用解析器上存在差异，而工具生成是我们评估者的属性，而非任何端点。将失败响应排除在分母之外会改变点的排序，因此可靠性包含改变了结论，而非其措辞。
cs.CL / 64 / 2609.10539
IdeaAMBIG: Benchmarking Implementation-Critical Gaps in Research-Idea Specifications
IdeaAMBIG：研究理念规范中实施关键差距的基准测试
Abstract
A research idea may be novel, coherent, and scientifically plausible, yet its proposed method may remain insufficiently specified for faithful implementation. We study the codification readiness of implementation-facing research-method specifications, defined by whether they provide sufficient methodological information for a competent implementer or coding agent to construct the intended method without unsupported assumptions. We construct evidence-grounded specifications and their supported resolutions from papers, codebases, issue threads, and reproduction artifacts. We introduce IdeaAMBIG, a benchmark of 660 evidence-grounded instances: 163 real-world gaps from reproducibility reports and GitHub issues, and 497 controlled synthetic gaps injected into codification-ready references. IdeaAMBIG evaluates three capabilities: codification-readiness assessment, defect localization, and clarification action generation. Defect localization receives only the specification, whereas clarification additionally receives the annotated defect. Across 13 LLMs, the best model achieves 9.6% Macro Defect Recovery Rate on real-world instances but 80.6% Macro Clarification Action Success Rate when given the defect. In an oracle study, supplying the gold resolution raises the downstream codification-ready rate from 14% to 98%. Across all evaluated models, defect localization is the main bottleneck, with stronger clarification given the defect.
Chinese Translation
一个研究理念可能是新颖的、一致的，并且在科学上是合理的，但其提出的方法可能仍然不足以明确规定以实现忠实实施。我们研究了面向实施的研究方法规范的编码准备性，定义为它们是否提供足够的方法信息，使得合格的实施者或编码代理能够在没有不支持假设的情况下构建预期的方法。我们从论文、代码库、问题线程和重现文物中构建了基于证据的规范及其支持的解决方案。我们引入了IdeaAMBIG，这是一个包含660个基于证据的实例的基准：163个来自可重现性报告和GitHub问题的真实世界差距，以及497个注入到编码准备参考中的受控合成差距。IdeaAMBIG评估三种能力：编码准备性评估、缺陷定位和澄清行动生成。缺陷定位仅接收规范，而澄清则额外接收注释缺陷。在13个大型语言模型（LLMs）中，最佳模型在真实世界实例上实现了9.6%的宏观缺陷恢复率，但在给定缺陷时，宏观澄清行动成功率达到了80.6%。在一次预言者研究中，提供黄金解决方案将下游编码准备率从14%提高到98%。在所有评估的模型中，缺陷定位是主要瓶颈，而在给定缺陷的情况下，澄清能力更强。