arXiv Daily Digest

406

Papers

Inverse kinematic solution for generic 3R positional robots using Conformal Geometric Algebra

使用共形几何代数的通用3R位置机器人逆运动学解

Nayak, Abhilash, Salunkhe, Durgesh Haribhau

Abstract

The inverse kinematics of generic 3R robots has been investigated through multiple approaches, mainly algebraic methods involving the solution of certain equation sets. Previous geometric interpretations of the solution, characterized as the intersection of a pair of conics have been confined to the joint-space domain. In this article, we study the Inverse Kinematic Model (IKM) of 3R robots, using the advantages of Conformal Geometric Algebra (CGA) to provide further insights on its kinematic properties. Our approach directly yields a univariate polynomial in terms of theta_2 without the need to eliminate theta_1 and theta_3 by reframing the problem as the intersection of two circles, which are fundamental elements within this algebraic framework.

Chinese Translation

通用3R机器人的逆运动学已通过多种方法进行研究，主要是涉及某些方程组解的代数方法。先前对解的几何解释，特征为一对圆锥曲线的交点，局限于关节空间域。在本文中，我们研究了3R机器人的逆运动学模型（Inverse Kinematic Model, IKM），利用共形几何代数（Conformal Geometric Algebra, CGA）的优势，进一步深入探讨其运动学特性。我们的方法直接得到一个关于theta_2的单变量多项式，而无需通过将问题重新构建为两个圆的交点来消去theta_1和theta_3，这在这个代数框架中是基本元素。

View on arXiv Download PDF AI Translation

cs.RO / 2 / 2609.00316

Geometric analysis of generic 3R robots, and necessary and sufficient conditions for a class of orthogonal robots to have four IKS

通用3R机器人的几何分析及一类正交机器人具有四个逆运动学解的必要和充分条件

Salunkhe, Durgesh Haribhau, Nayak, Abhilash

Abstract

The kinematic analysis of a generic 3R robot has been investigated with multiple approaches in the past. The algebraic approaches have established concrete results but are unfortunately limited to special classes or architectural simplifications. Geometric approaches on the other hand have extended the analysis to generic robots while also providing an intuitive understanding of their kinematic properties. We use the best of both approaches to present the kinematic analysis of a generic 3R robot, using the inverse kinematic model inherited from a method based on conformal geometric algebra. The paper discusses a generic framework to study the conditions for a 3R robot to have four inverse kinematic solutions (IKS) and allows to study the distribution of IKS as seen in workspace. The necessary and sufficient conditions for a class of orthogonal robots are presented using the proposed approach.

Chinese Translation

过去对通用3R机器人的运动学分析采用了多种方法。代数方法虽然建立了具体的结果，但不幸的是仅限于特殊类别或结构简化。另一方面，几何方法扩展了对通用机器人的分析，同时提供了对其运动学特性的直观理解。我们结合两种方法的优点，利用基于共形几何代数的方法继承的逆运动学模型，呈现通用3R机器人的运动学分析。本文讨论了研究3R机器人具有四个逆运动学解（IKS）条件的通用框架，并允许研究在工作空间中观察到的IKS的分布。使用所提出的方法，给出了正交机器人一类的必要和充分条件。

View on arXiv Download PDF AI Translation

cs.RO / 3 / 2609.00333

Scene Graph-based Driving Scenario Extraction for Automotive Egocentric Datasets

基于场景图的汽车自我中心数据集驾驶场景提取

Ramdhan, Stefan, Dagenais, Kyanna, Pantelic, Vera, Bandur, Victor, Lawford, Mark

Abstract

Extracting scenarios from unlabelled real-world sensor data streams is a critical but challenging task in the development process of automated driving systems (ADS). Automatically sifting through large datasets to spatially and temporally locate critical scenarios can enable scenario-based coverage analysis of ADS datasets. In this paper, we present a method for extracting scenarios from egocentric datasets using scene graphs and Linear Temporal Logic (LTL). We first process egocentric sensor data and HD maps to generate a sequence of scene graphs representing a driving scenario. Next, we use LTL to formally specify driving scenarios of interest, then extract all instances of the scenarios from the dataset using an off-the-shelf model checker, which evaluates the LTL formula against the sequence of scene graphs. Our approach can be used on both simulated and real world datasets. We evaluate the method on the training and validation datasets from Argoverse 2 consisting of 850 15-second real-world driving logs, and several videos of dashcam footage. We demonstrate the effectiveness of our approach for extracting and querying scenarios by evaluating against a rule-based benchmark based on track annotations and HD maps.

Chinese Translation

从未标记的真实世界传感器数据流中提取场景是自动驾驶系统（ADS）开发过程中的一项关键但具有挑战性的任务。自动筛选大型数据集以空间和时间上定位关键场景，可以实现基于场景的ADS数据集覆盖分析。本文提出了一种使用场景图和线性时序逻辑（LTL）从自我中心数据集中提取场景的方法。我们首先处理自我中心传感器数据和高清地图，以生成表示驾驶场景的场景图序列。接下来，我们使用LTL正式指定感兴趣的驾驶场景，然后利用现成的模型检查器从数据集中提取所有场景实例，该检查器根据场景图序列评估LTL公式。我们的方法可以应用于模拟和真实世界数据集。我们在包含850个15秒真实驾驶日志和若干行车记录仪视频的Argoverse 2训练和验证数据集上评估该方法。通过与基于轨迹注释和高清地图的规则基准进行评估，我们展示了我们的方法在提取和查询场景方面的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 4 / 2609.00385

Risk-Aware Decision-Making for Autonomous Overtaking: A World Model-Based Mixture-of-Experts Framework

基于世界模型的风险感知自主超车决策框架

Liu, Yongzhi, Zhang, Sunan, Xu, Jinchang, Wang, Jiawei, Qiu, Yushu, Lv, Chen, Zhuang, Weichao

Abstract

Autonomous highway overtaking demands foresighted decision-making to handle complex interactions, stochastic traffic evolution, and temporal risk accumulation. However, standard safe reinforcement learning approaches typically rely on implicit value-based risk estimations rather than explicit dynamics modeling, thereby struggling to accurately capture complex risk propagation over multi-step horizons. This limitation frequently results in behaviors that are locally safe but induce substantial latent risks in the long term. To address this, a World Model-based Risk-aware Mixture-of-Experts (WM-RMoE) framework is proposed. First, a learned latent dynamics model facilitates parallel multi-step rollouts, elevating safety assessment from the action level to the trajectory level via cumulative risk evaluation. Second, to enhance robustness under varying interaction intensities, a hierarchical gating mechanism dynamically coordinates experts across long-horizon, short-horizon, and rule-based safety modules. Furthermore, a Gaussian Mixture Model is integrated to preserve multimodal maneuvering branches, thereby mitigating the issue of behavioral mode averaging. Experimental results demonstrate that WM-RMoE significantly outperforms representative baselines in terms of safety compliance, decision stability, and generalization capability. Furthermore, benefiting from the risk-aware formulation, the proposed framework uniquely exhibits the ability to generate foresighted and semantically distinct overtaking maneuvers across diverse traffic densities.

Chinese Translation

自主高速公路超车要求前瞻性的决策，以应对复杂的交互、随机的交通演变和时间风险的累积。然而，标准的安全强化学习方法通常依赖于隐式的基于价值的风险估计，而非显式的动态建模，从而难以准确捕捉多步视野下复杂的风险传播。这一局限性常常导致局部安全的行为，但在长期内引发显著的潜在风险。为此，提出了一种基于世界模型的风险感知专家混合框架（WM-RMoE）。首先，学习的潜在动态模型促进了并行的多步展开，将安全评估从动作层面提升到轨迹层面，通过累积风险评估实现。其次，为了增强在不同交互强度下的鲁棒性，层次化的门控机制动态协调长时间视野、短时间视野和基于规则的安全模块中的专家。此外，集成了高斯混合模型以保留多模态的机动分支，从而减轻行为模式平均化的问题。实验结果表明，WM-RMoE在安全合规性、决策稳定性和泛化能力方面显著优于代表性基线。此外，得益于风险感知的框架，所提出的方法独特地展现出在不同交通密度下生成前瞻性和语义上独特的超车机动的能力。

View on arXiv Download PDF AI Translation

cs.RO / 5 / 2609.00478

Beyond Object Selection:Markerless Gaze-based Robot Placement at Arbitrary Position

超越物体选择：无标记的基于注视的机器人在任意位置的放置

Lai, Yuzhi, Marx, William, Yuan, Shenghai, Li, Peizheng, Ran, Zhuoyu, Zell, Andreas

Abstract

Gaze-based assistive manipulation typically supports object selection, while arbitrary-position placement requires accurate spatial alignment between the headset and robot. However, for gaze-based manipulation, pose accuracy does not necessarily translate into task accuracy: translational and rotational errors jointly affect the transformed gaze ray and may compensate for each other. To study cross-device alignment from this task-oriented perspective, we present a markerless interaction framework and a dedicated cross-device dataset. We propose Graph-based Reference Selection to address sparse robot references. We further develop and benchmark multiple task-specific alignment pipelines under a unified protocol. Specifically, we introduce Gaze--Surface Intersection Error (GSIE), which directly measures the spatial error of the gaze-specified target. Experiments show that alignment methods ranked highly by conventional pose metrics are not always optimal in GSIE, demonstrating the importance of evaluating gaze-based manipulation at the task level.

Chinese Translation

基于注视的辅助操作通常支持物体选择，而任意位置的放置则需要头戴设备与机器人之间的精确空间对齐。然而，对于基于注视的操作，姿态精度并不一定转化为任务精度：平移和旋转误差共同影响变换后的注视光线，并可能相互补偿。为了从任务导向的角度研究跨设备对齐，我们提出了一种无标记的交互框架和一个专门的跨设备数据集。我们提出了基于图的参考选择方法以解决稀疏的机器人参考问题。我们进一步开发并基准测试了多个任务特定的对齐管道，采用统一的协议。具体而言，我们引入了注视-表面交点误差（Gaze--Surface Intersection Error, GSIE），该指标直接测量注视指定目标的空间误差。实验表明，传统姿态指标排名较高的对齐方法在GSIE中并不总是最优，展示了在任务层面评估基于注视的操作的重要性。

View on arXiv Download PDF AI Translation

cs.RO / 6 / 2609.00539

Exploring Nonlinear Body Oscillations for Natural Quadruped Gaits

探索自然四足步态的非线性身体振荡

Schmidt, Annika, Calzolari, Davide, Loeffl, Florian, Sachtler, Arne, Seidel, Daniel, Herrmann, Milan, Burger, Robert, Gumpert, Thomas, Raffin, Antonin, Ehlert, Tristan, Pries, Maximilian, Wandinger, David, Schmidt, Florian, Keppler, Manuel, Lee, Jinoh, Albu-Schäffer, Alin

Abstract

Animals' body morphology shapes the gait patterns they can perform, where mechanical resonance reduces the need for active control. By tuning posture and muscle stiffness, they leverage their embodied intelligence to achieve effective gaits for different speeds. In contrast, most quadruped robots are not specifically designed to exploit mechanical resonance due to the complexity of nonlinear dynamics and require dedicated locomotion controllers. To provide an alternative, we present a proof of concept framework making the nonlinear dynamics of a robot predictable in the design process and show how this knowledge can be leveraged such that multi-gait locomotion can emerge from nonlinear resonances, shaped by gravity, inertia, and elasticity. We present the highly compliant quadruped robot eBert, on which we identify six nonlinear normal modes (NNMs) using our new theoretical tools and validate their existence in simulation and hardware. With black-box optimization to determine step length, simulations show how each NNM naturally develops into a distinct gait, manifesting different speeds, which also largely transfers to the robotic hardware. Our experiments show that eBert can exploit its mechanics to generate task-specific movements which may serve as foundation for designing a new generation of agile and efficient robots leveraging embodied intelligence.

Chinese Translation

动物的身体形态决定了它们能够执行的步态模式，其中机械共振减少了对主动控制的需求。通过调整姿势和肌肉刚度，它们利用其具身智能在不同速度下实现有效的步态。相比之下，大多数四足机器人并未专门设计用于利用机械共振，原因在于非线性动力学的复杂性，并且需要专门的运动控制器。为提供一种替代方案，我们提出了一个概念验证框架，使机器人的非线性动力学在设计过程中可预测，并展示如何利用这一知识，使多步态运动能够从非线性共振中产生，这些共振受重力、惯性和弹性的影响。我们展示了高度柔顺的四足机器人eBert，并利用我们新的理论工具识别出六种非线性正常模式（NNMs），并在仿真和硬件中验证了它们的存在。通过黑箱优化确定步长，仿真显示每个NNM自然发展为一种独特的步态，表现出不同的速度，这种特性在机器人硬件中也得到了很大程度的转移。我们的实验表明，eBert能够利用其机械特性生成特定任务的运动，这可能为设计新一代灵活高效的机器人奠定基础，充分利用具身智能。

View on arXiv Download PDF AI Translation

cs.RO / 7 / 2609.00555

Potential-Guided Particle Steering for Negation-Constrained Dexterous Grasping

基于潜在引导的粒子引导用于否定约束的灵巧抓取

Kim, Geonho, Kim, SooGon, Lee, Jongmin

Abstract

Language-driven dexterous grasp models, such as DextER, perform well when instructions specify where to grasp, but we find they fail systematically when an instruction also specifies where not to grasp (e.g., "grasp the handle but avoid the body"). Existing training corpora, DexGYSNet among them, contain virtually no avoidance instructions, and collecting examples for every possible constraint is impractical. Moreover, because every part mentioned during training denotes a contact target, models may interpret a forbidden part as another region to grasp rather than one to avoid. We therefore introduce an inference-time framework for negation-constrained dexterous grasping that requires no negation-specific training examples. Combining Sequential Monte Carlo with classifier-free guidance, our method guides sampling toward the instructed part while pruning candidates headed for the forbidden region, without any negation examples during training. A frozen 3D part-grounding model localizes the forbidden region from the language instruction. To evaluate this setting, we construct NegGrasp, a benchmark of paired positive/negative instructions with constraint-aware metrics that credit a grasp only if it both accomplishes the task and respects the stated constraint. On NegGrasp, our method reduces the violation rate of the strongest baseline from 57.9% to 17.2% while improving both constraint-aware and physical success.

Chinese Translation

语言驱动的灵巧抓取模型，如 DextER，在指令明确指定抓取位置时表现良好，但我们发现当指令同时指定避免抓取的位置时（例如，“抓住手柄但避免抓住主体”），它们系统性地失败。现有的训练语料库，包括 DexGYSNet，几乎没有避免指令，收集每个可能约束的示例是不切实际的。此外，由于训练期间提到的每个部分都表示一个接触目标，模型可能将禁止的部分解释为另一个抓取区域，而不是一个需要避免的区域。因此，我们提出了一种用于否定约束灵巧抓取的推理时框架，该框架不需要特定于否定的训练示例。我们的方法结合了序列蒙特卡洛与无分类器引导，指导采样朝向指令部分，同时修剪前往禁止区域的候选项，在训练期间没有任何否定示例。一个冻结的 3D 部分定位模型根据语言指令定位禁止区域。为了评估这一设置，我们构建了 NegGrasp，一个成对正/负指令的基准，并使用约束感知指标，仅在抓取既完成任务又遵循所述约束时给予信用。在 NegGrasp 上，我们的方法将最强基线的违规率从 57.9% 降低到 17.2%，同时提高了约束感知和物理成功率。

View on arXiv Download PDF AI Translation

cs.RO / 8 / 2609.00612

A Wearable Pneumatic Device for Continuous, Closed-Loop, Bidirectional Tactile Interaction

一种可穿戴气动设备用于连续、闭环、双向触觉交互

Pasquier, Cosima du, Smith, Aliyah, Huber, Serin, Phelps, Joshua, Cohen, Ilana A., Chen, Ava, Kennedy III, Monroe, Okamura, Allison M.

Abstract

We present a system of two wearable pneumatic haptic devices that supports continuous, closed-loop, bidirectional tactile interaction at perceptually relevant force and temporal scales. A single device can contain up to twelve pressure sensing channels connected to textile-based pneumatic pouches. Each channel in a device can be used as a sensor, an actuator, or both. As an actuator with integrated sensing, the channel generates stable skin indentation through local closed-loop control. As a sensor, a channel can be mounted (or worn) on any surface, including on a robot gripper or on the human body, and used to measure touch interactions with the environment or a human user. A distributed architecture supports sustained pressure output, rapid dynamic response, and wireless pairing of identical devices in a system to transmit and reproduce tactile pressure signals in real time. Device-level characterization demonstrates force bandwidth exceeding 30 Hz, rapid and well-damped step responses, and extended pressure retention compared to prior compact pneumatic platforms. Human studies show that pressure-based fingertip feedback enables discrimination of force and stiffness, improves teleoperated manipulation by reducing applied pressures by up to 23.1% and task duration by up to 27.4%, and lowers subjective mental workload by 18.8%, particularly under visually constrained conditions. By unifying tactile sensing and haptic feedback within a single pneumatic modality, the device provides a practical foundation for bidirectional touch interaction in teleoperation.

Chinese Translation

我们提出了一种由两个可穿戴气动触觉设备组成的系统，支持在感知相关的力和时间尺度上进行连续、闭环、双向触觉交互。单个设备最多可以包含十二个压力传感通道，这些通道连接到基于纺织的气动袋。设备中的每个通道可以用作传感器、执行器或两者兼具。作为集成传感的执行器，该通道通过局部闭环控制生成稳定的皮肤凹陷。作为传感器，通道可以安装（或佩戴）在任何表面上，包括机器人抓手或人体上，用于测量与环境或人类用户的触摸交互。分布式架构支持持续的压力输出、快速的动态响应，以及在系统中无线配对相同设备以实时传输和再现触觉压力信号。设备级特性表明，力带宽度超过30 Hz，快速且良好阻尼的阶跃响应，以及与先前紧凑型气动平台相比，延长的压力保持时间。人类研究表明，基于压力的指尖反馈能够区分力和刚度，通过将施加压力降低多达23.1%和任务持续时间缩短多达27.4%，改善远程操作的操控，并在视觉受限条件下降低主观心理负担18.8%。通过在单一气动模式中统一触觉感知和触觉反馈，该设备为远程操作中的双向触摸交互提供了实用基础。

View on arXiv Download PDF AI Translation

cs.RO / 9 / 2609.00619

DSG: Dynamic 3D Scene Graph Construction for Embodied Agents in Changing Indoor Environments

DSG：用于变化室内环境中具身智能体的动态三维场景图构建

Liao, Ming, Ye, Chao, Fei, Jianing, Lin, Weiyang

Abstract

In indoor environments, object positions frequently change due to human activities or embodied-agent interactions, causing previously constructed scene graphs to become inconsistent with the current scene. To address this issue, we propose DSG, a dynamic 3D scene graph construction framework that detects object changes and performs spatial relationship reasoning. First, we construct a semantic-aware 3D Gaussian scene representation and develop a dual-view rendering-based object change detection method to enable reliable scene graph node updates. Second, we propose a spatial relationship reasoning method that incorporates multi-granularity visual context, enabling a large language model to identify a richer set of interobject spatial relationships. Furthermore, we introduce DynTHOR, a dynamic indoor scene graph benchmark built on the AI2-THOR simulation platform for evaluating scene graph construction in dynamic environments. Extensive experiments on Dyn-THOR, 3RScan, and real-world scenes demonstrate that DSG consistently outperforms existing methods in both object node construction and spatial relationship reasoning, significantly improving the accuracy of dynamic scene graph construction.

Chinese Translation

在室内环境中，由于人类活动或具身智能体的交互，物体位置经常发生变化，这导致先前构建的场景图与当前场景不一致。为了解决这一问题，我们提出了DSG，一个动态三维场景图构建框架，能够检测物体变化并进行空间关系推理。首先，我们构建了一个语义感知的三维高斯场景表示，并开发了一种基于双视图渲染的物体变化检测方法，以实现可靠的场景图节点更新。其次，我们提出了一种结合多粒度视觉上下文的空间关系推理方法，使大型语言模型能够识别更丰富的物体间空间关系。此外，我们引入了DynTHOR，这是一个基于AI2-THOR仿真平台构建的动态室内场景图基准，用于评估动态环境中的场景图构建。在Dyn-THOR、3RScan和真实场景上的大量实验表明，DSG在物体节点构建和空间关系推理方面始终优于现有方法，显著提高了动态场景图构建的准确性。

View on arXiv Download PDF AI Translation

cs.RO / 10 / 2609.00641

AM-Bench: A Modular Simulation Suite and Benchmark for Aerial Manipulation Policy Learning

AM-Bench：一个用于空中操控策略学习的模块化仿真套件和基准测试

Wang, Yutong, Lee, Dongjae, Guo, Xiaofeng, Zhan, Yuanzhu, Jiang, Yufei, Saravanan, Bavin, Cao, Muqing, Xie, Jia, Mao, Chenyang, Scherer, Sebastian, Geng, Junyi, Shi, Guanya

Abstract

Standardized benchmarks have played a central role in advancing robot manipulation learning, yet most focus on ground-supported manipulation systems, which limits their applicability to dynamics-critical domains such as aerial manipulation (AM). AM presents distinct system-level challenges, including environmental disturbances, coupled dynamics between the manipulator and floating base, and constrained degrees of freedom. Consequently, task performance depends jointly on robot embodiment, low-level control, and high-level policy design. We introduce AM-Bench, a modular simulation suite and benchmark for multirotor-based AM policy learning. AM-Bench includes representative embodiments spanning underactuated, fully actuated, and overactuated systems, 12 tasks across contact, transport, and constrained interaction, configurable aerodynamic disturbances and actuator saturation, standard low-level controllers, and baseline policy-learning algorithms. Unlike prior manipulation benchmarks that primarily emphasize end-to-end policy performance, AM-Bench enables system-level evaluation of how embodiment, control, disturbances, and policy choices interact. We demonstrate its diagnostic value through three simulation studies spanning high-level policies, policy--control interfaces, and embodiments, together with real-world validation of modeled effects and a hardware test of the learning pipeline.

Chinese Translation

标准化基准在推动机器人操控学习方面发挥了重要作用，但大多数基准集中于地面支持的操控系统，这限制了它们在动态关键领域（如空中操控（AM））的适用性。AM面临独特的系统级挑战，包括环境干扰、操控器与浮动基座之间的耦合动态以及受限的自由度。因此，任务性能共同依赖于机器人的体现、低级控制和高级策略设计。我们介绍了AM-Bench，一个用于多旋翼AM策略学习的模块化仿真套件和基准测试。AM-Bench包含代表性的体现，涵盖欠驱动、全驱动和过驱动系统，涉及接触、运输和受限交互的12个任务，可配置的气动干扰和执行器饱和、标准的低级控制器以及基线策略学习算法。与以往主要强调端到端策略性能的操控基准不同，AM-Bench使得能够系统性地评估体现、控制、干扰和策略选择之间的相互作用。我们通过三个仿真研究展示了其诊断价值，这些研究涵盖了高级策略、策略-控制接口和体现，并结合对建模效果的实际验证和学习流程的硬件测试。

View on arXiv Download PDF AI Translation

cs.RO / 11 / 2609.00648

Human-robot conversation with multiple participants in noisy public spaces

嘈杂公共空间中多参与者的人机对话

Lala, Divesh, Muthukumaran, Yogeeswaran, Fernandes, Vincent, Kato, Kazushi, Fujiki, Shota, Chi, Zihao, Iwasaki, Masaya, Shintani, Taiken, Kawata, Megumi, Sakai, Kazuki, Inoue, Koji, Yoshikawa, Yuicihiro, Kawahara, Tatsuya

Abstract

For noisy real-world environments such as those in open public spaces, spoken dialogue systems for both autonomous robots and avatars should be carefully designed to provide enhanced speech signals. These signals can be used either for speech recognition or, in the case of an avatar system, transmitted as clean speech to a remote operator. This work proposes an audio system that can be used for both these scenarios and was demonstrated as a proof-of-concept at the 2025 World Expo in Osaka. The first scenario is an attentive listening system with the android ERICA, and the second is a conversation support system with mobile Teleco robots, with one of them acting as an avatar for a remote operator. Both systems feature multi-party conversation and use a single multi-channel microphone array. We describe how our audio system not only enhances the speech of multiple speakers in a noisy environment, but provides a form of spatial audio which allows for more immersiveness in avatar-based conversational interactions.

Chinese Translation

对于开放公共空间等嘈杂的现实环境，针对自主机器人和虚拟化身的语音对话系统应经过精心设计，以提供增强的语音信号。这些信号可以用于语音识别，或者在虚拟化身系统的情况下，作为清晰的语音传输给远程操作员。本研究提出了一种音频系统，可用于这两种场景，并在2025年大阪世博会上作为概念验证进行了展示。第一个场景是与安卓机器人ERICA的专注听取系统，第二个场景是与移动Teleco机器人进行对话支持系统，其中一个机器人作为远程操作员的虚拟化身。两个系统都具备多方对话功能，并使用单个多通道麦克风阵列。我们描述了我们的音频系统如何不仅增强嘈杂环境中多个发言者的语音，还提供了一种空间音频形式，使基于虚拟化身的对话交互更加沉浸。

View on arXiv Download PDF AI Translation

cs.RO / 12 / 2609.00659

Fleets Need a Context Plane: Rethinking Cooperative Perception for Autonomous Drones

舰队需要一个上下文平面：重新思考自主无人机的协同感知

Liu, Liangkai, Wu, Xiaoxiao

Abstract

Cooperative perception allows a drone fleet to combine observations from multiple viewpoints. However, existing systems typically fix their feature-sharing policies at design time or adapt to only one context signal. This is a poor fit for aerial fleets, whose missions, bandwidth, formation geometry, and scene coverage can change during flight. We quantify the cost of context-blind sharing on UAV3D by controlling feature exchange at evaluation time using a released DiscoNet checkpoint, without retraining. Mission-aware sharing matches full-sharing accuracy while using only 5-10% of the bytes. The best tested peer selection policy changes with the byte budget, and choosing the wrong policy loses up to 7.7 AP. Moreover, under a constrained budget, two policies with the same full-scene accuracy differ by 5.9 AP within the mission region, showing that multiple context axes must be considered jointly. We therefore propose the context plane, a bounded, structured interface for runtime context. Each drone publishes a descriptor of at most 1 KB at 10 Hz, and lightweight, replaceable policies use the fleet context to decide what each drone computes, shares, and fuses. Existing sharing schemes become fixed policies within this interface. In our ROS 2 prototype on a Jetson AGX Orin, the context plane uses approximately 0.01% of the data-plane bandwidth, and each policy decision takes 0.10 ms. These results show that an explicit context interface can support low-overhead runtime adaptation without modifying or retraining the perception model.

Chinese Translation

协同感知允许无人机舰队结合来自多个视角的观察。然而，现有系统通常在设计时固定其特征共享策略，或仅适应一个上下文信号。这对于空中舰队来说并不合适，因为其任务、带宽、编队几何形状和场景覆盖范围在飞行过程中可能会发生变化。我们通过在评估时控制特征交换，使用发布的 DiscoNet 检查点，而不进行重新训练，量化了上下文盲共享在 UAV3D 上的成本。任务感知共享在仅使用 5-10% 的字节的情况下，达到了完全共享的准确性。测试的最佳对等选择策略随着字节预算的变化而变化，选择错误的策略可能导致高达 7.7 的平均精度 (AP) 损失。此外，在预算受限的情况下，两个具有相同全场景准确性的策略在任务区域内的表现相差 5.9 AP，这表明必须共同考虑多个上下文轴。因此，我们提出了上下文平面，这是一个用于运行时上下文的有界、结构化接口。每个无人机以 10 Hz 的频率发布最多 1 KB 的描述符，轻量级、可替换的策略利用舰队上下文来决定每个无人机计算、共享和融合的内容。现有的共享方案在此接口内变为固定策略。在我们基于 Jetson AGX Orin 的 ROS 2 原型中，上下文平面使用大约 0.01% 的数据平面带宽，每个策略决策耗时 0.10 毫秒。这些结果表明，显式的上下文接口可以支持低开销的运行时适应，而无需修改或重新训练感知模型。

View on arXiv Download PDF AI Translation

cs.RO / 13 / 2609.00669

Behavior--Realization Separation for Constrained Physical Human--Robot Interaction

约束物理人机交互中的行为与实现分离

Cao, Yongyan

Abstract

Physical human--robot interaction software often couples desired-behavior specification with constrained realization; we treat these as separate layers. A \emph{behavior layer} supplies a desired contact-port acceleration $a_k^{\mathrm{id}}=f_\theta(e_k,\dot e_k,F_{h,k})$. A \emph{realization layer} converts it into constrained robot commands and reports total desired-versus-realized acceleration error instead of hiding it in saturation. A same-objective unconstrained counterfactual separates regularization from constraint intervention, while plant data expose model error. This paper implements a receding-horizon quadratic program realizing memoryless affine behaviors. Changing the behavior modifies objective coefficients through $(C_\theta,G_\theta)$ while the robot-command variable and feasible set remain unchanged. A planar study instantiates impedance and admittance; the same running layer accepts an impedance--admittance--impedance reassignment without reconstruction, under its existing rate limit. On a torque-controlled 7-DOF Franka FR3 in MuJoCo, the runtime freezes task-space dynamics per solve and enforces torque feasibility across its horizon. Under a sustained 20~N push, it holds a slack-relaxed workspace boundary to within approximately 0.1--0.2~mm, versus 4.4~cm (impedance) and 4.7~cm (admittance) overshoot from instantaneous clipping. A derated actuator budget then activates the torque constraint: horizon-wide enforcement keeps its frozen-model plan feasible to $2.1\times10^{-4}$~N$\cdot$m, whereas a first-step-only ablation plans up to 11.329~N$\cdot$m beyond budget; on the executed nonlinear plant, where both share the same local-model error, the gap is smaller but still favors horizon-wide enforcement (0.161 vs.\ 0.380~N$\cdot$m). These results are a focused proof of behavior--realization separation.

Chinese Translation

物理人机交互软件通常将期望行为规范与约束实现相结合；我们将其视为两个独立的层次。一个 extit{行为层}提供所需的接触端口加速度 $a_k^{ ext{id}}=f_ heta(e_k, ext{dot } e_k,F_{h,k})$。一个 extit{实现层}将其转换为约束的机器人命令，并报告总的期望与实现加速度误差，而不是将其隐藏在饱和状态中。一个具有相同目标的无约束反事实将正则化与约束干预分开，而植物数据则揭示了模型误差。本文实现了一个递归时间范围的二次规划，能够实现无记忆的仿射行为。改变行为通过 $(C_ heta,G_ heta)$ 修改目标系数，而机器人命令变量和可行集保持不变。一个平面研究实例化了阻抗和导纳；相同的运行层在其现有速率限制下接受阻抗-导纳-阻抗的重新分配，而无需重构。在MuJoCo中对一个扭矩控制的7自由度Franka FR3进行的实验中，运行时在每次求解时冻结任务空间动态，并在其时间范围内强制执行扭矩可行性。在持续的20~N推力下，它将松弛的工作空间边界保持在大约0.1~0.2~mm以内，而瞬时剪切导致的超调分别为4.4~cm（阻抗）和4.7~cm（导纳）。降额的执行器预算随后激活了扭矩约束：全范围的强制执行使其冻结模型计划在 $2.1 imes10^{-4}$~N$ullet$m内保持可行，而仅第一步的消融计划则超出预算达到11.329~N$ullet$m；在执行的非线性植物上，二者共享相同的局部模型误差，差距较小但仍然更倾向于全范围的强制执行（0.161 vs. 0.380~N$ullet$m）。这些结果是行为与实现分离的有力证明。

View on arXiv Download PDF AI Translation

cs.RO / 14 / 2609.00677

ADAPT: Agile Diffusion Action Priors for Robust and Steerable Online Text-Driven Humanoid Control

ADAPT：用于稳健和可操控的在线文本驱动类人控制的敏捷扩散动作先验

Wu, Yan, Li, Chenhao, Zhao, Kaifeng, Li, Gen, Hutter, Marco, Tang, Siyu

Abstract

We present ADAPT, an end-to-end framework for interactive, text-conditioned humanoid whole-body control. Unlike dominant text-to-motion pipelines that generate kinematic motions for a separate tracker, ADAPT solves language control with an end-to-end closed-loop control framework, where the robot must continuously respond to changing commands while maintaining balance, natural motion, and smooth transitions. ADAPT learns a diffusion-based action prior from text-labeled humanoid state-action trajectories, enabling diverse motion skills to be directly executed from language commands. To improve long-horizon robustness and smooth prompt switching, we train a lightweight residual reinforcement learning policy on top of the frozen diffusion controller. We further show that the same diffusion policy can be reused as a steerable text-conditioned motion prior for downstream task adaptation. Experiments demonstrate robust language-grounded skill execution, smooth interactive transitions, and style-preserving downstream control.

Chinese Translation

我们提出了ADAPT，一个用于交互式文本条件下类人全身控制的端到端框架。与主流的文本到运动管道不同，后者为单独的跟踪器生成运动学动作，ADAPT通过一个端到端的闭环控制框架解决语言控制问题，在该框架中，机器人必须在保持平衡、自然运动和流畅过渡的同时，持续响应变化的指令。ADAPT从文本标记的类人状态-动作轨迹中学习基于扩散的动作先验，使得可以直接从语言指令执行多样的运动技能。为了提高长时间范围的稳健性和流畅的提示切换，我们在冻结的扩散控制器之上训练了一个轻量级的残差强化学习策略。我们进一步表明，相同的扩散策略可以作为可操控的文本条件运动先验，重复用于下游任务适应。实验结果表明了稳健的语言基础技能执行、流畅的交互过渡以及风格保持的下游控制。

View on arXiv Download PDF AI Translation

cs.RO / 15 / 2609.00682

Context-Aware Intelligent Vehicles

基于情境感知的智能车辆

Liu, Liangkai, Shi, Shuyao, Wang, Mingke, Curran, Noah T., Li, Chuan, Bai, Fan, Shin, Kang G.

Abstract

Intelligent vehicles increasingly support adaptive applications beyond driving themselves, ranging from context-aware ADAS and automated driving to in-cabin monitoring and fleet management, all under tight requirements on accuracy, latency, cost, and reliability. Meeting these requirements is challenging because vehicles operate in complex, uncertain, and rapidly changing environments while running on resource-constrained computing platforms. This paper argues that context-situational factors that give meaning to sensor signals and constrain decisions-should be treated as a first-class principle for next-generation vehicle systems, and operationalized as a unified, shared state for learning, risk assessment, and closed-loop control across the software stack. We systematically review state-of-the- art (SOTA) context-aware methods spanning (i) environment understanding, (ii) planning and control, (iii) safety and security, and (iv) connected vehicles. Based on a trend analysis of context-aware design, we identify four key technical challenges in building a general contextual engine for future intelligent vehicles: multi-modal context fusion, temporal context modeling, handling rare events, and collaborative context sharing. We hope this survey will motivate the development of robust and efficient context-aware vehicle applications.

Chinese Translation

智能车辆越来越多地支持超越自我驾驶的自适应应用，包括情境感知的高级驾驶辅助系统（ADAS）、自动驾驶、车内监控和车队管理，所有这些都在对准确性、延迟、成本和可靠性有严格要求的情况下进行。满足这些要求具有挑战性，因为车辆在复杂、不确定和快速变化的环境中运行，同时依赖于资源受限的计算平台。本文认为，赋予传感器信号意义并限制决策的情境因素应作为下一代车辆系统的首要原则，并在软件堆栈中作为统一的共享状态进行操作，以便于学习、风险评估和闭环控制。我们系统回顾了涵盖（i）环境理解、（ii）规划与控制、（iii）安全与保障、（iv）联网车辆的最新情境感知方法。基于对情境感知设计的趋势分析，我们确定了构建未来智能车辆通用情境引擎的四个关键技术挑战：多模态情境融合、时间情境建模、处理稀有事件和协作情境共享。我们希望这项调查能够激励开发稳健且高效的情境感知车辆应用。

View on arXiv Download PDF AI Translation

cs.RO / 16 / 2609.00751

One Print, Many Moves: Monolithic Origami-inspired Folding Actuator for Composable Soft Multi-DoF Systems

一体化、多功能：单体折纸启发的折叠驱动器用于可组合的软多自由度系统

Jang, Jaehyung, Zhakypov, Zhenish, Palmer, Jasmin Elena, Klein, Melissa, Ryu, Jee-Hwan, Okamura, Allison Mariko

Abstract

Conventional soft robot actuators excel in compliance, but their uncontrolled deformations compromise accuracy and hinder scaling to multi-degree-of-freedom (DoF) systems. We introduce a MONOlithic ORIGAMI-inspired soft folding actuator design (MONORIGAMI) that establishes a design strategy based on spatially programmed stiffness anisotropy to preserve material compliance along desired folding directions while selectively restricting deformation in unwanted directions. The actuator leverages stiffness tiers based on material thickness, patterned in an origami-inspired geometry with facets and creases, converting unconstrained soft deformation into accurate, repeatable, and composable folding motions without additional reinforcements. The design is fully 3D-printable through a single-material, single-print process that requires no assembly. Each actuator serves as a scalable motion primitive, and linking and orienting multiple actuators mechanically programs multi-DoF trajectories. Using the same fundamental module, we demonstrate three 3D-printed soft multi-DoF robotic systems spanning distinct application domains: (1) a compact 4-DoF wearable haptic device for high-fidelity cutaneous feedback in virtual reality (VR), (2) a 3-DoF joystick for kinesthetic feedback in teleoperation, and (3) a modular robotic gripper capable of underwater operation with geometry-encoded grasp trajectories. These systems demonstrate the module's capabilities for compact multi-axis integration, controlled physical interaction, and geometry-programmed operation across different environments. Together, these results show that MONORIGAMI provides a general, composable, accessible, reliable, and scalable platform for high-precision soft multi-DoF robotics, addressing long-standing limitations in both soft actuator design and fabrication.

Chinese Translation

传统的软机器人驱动器在柔顺性方面表现出色，但其不受控制的变形会影响精度，并妨碍多自由度（DoF）系统的扩展。我们提出了一种单体折纸启发的软折叠驱动器设计（MONORIGAMI），该设计基于空间编程的刚度各向异性建立了一种设计策略，以在所需的折叠方向上保持材料的柔顺性，同时选择性地限制在不希望的方向上的变形。该驱动器利用基于材料厚度的刚度层次，采用折纸启发的几何图形，具有面和折痕，将不受约束的软变形转化为精确、可重复和可组合的折叠运动，而无需额外的加固。该设计完全可通过单一材料、单次打印的过程进行3D打印，无需组装。每个驱动器作为可扩展的运动原语，多个驱动器的连接和定向可以机械编程多自由度轨迹。利用相同的基本模块，我们展示了三种3D打印的软多自由度机器人系统，涵盖不同的应用领域：（1）用于虚拟现实（VR）中高保真皮肤反馈的紧凑型4自由度可穿戴触觉设备，（2）用于远程操作的3自由度操纵杆，和（3）能够在水下操作的模块化机器人抓手，具有几何编码的抓取轨迹。这些系统展示了该模块在紧凑多轴集成、受控物理交互和跨不同环境的几何编程操作方面的能力。综合来看，这些结果表明，MONORIGAMI为高精度软多自由度机器人提供了一个通用、可组合、易于访问、可靠且可扩展的平台，解决了软驱动器设计和制造中的长期局限性。

View on arXiv Download PDF AI Translation

cs.RO / 17 / 2609.00769

A Compact Robotic Finger with 2-DoF MCP Joint Embedding DoF-Selective Passive Continuously Variable Transmission for Wide Force-Speed Operating Range

一种紧凑型机器人手指，具有2自由度MCP关节嵌入选择性被动连续可变传动以实现广泛的力-速度操作范围

Jang, JaeHyung, Ryu, Jee-Hwan

Abstract

This letter presents a compact two-degree-of-freedom (DoF) robotic finger with a flexion-selective passive continuously variable transmission (CVT) to achieve a wide force-speed operating range. Inspired by the functional differentiation of the human metacarpophalangeal (MCP) joint, the proposed mechanism realizes DoF-specific transmission differentiation by selectively assigning passive CVT to the flexion-extension DoF while preserving direct transmission for abduction-adduction. For a wide force-speed operating range, a force-responsive passive CVT is embedded in the flexion pathway, while direct transmission is preserved for the abduction-adduction pathway. To selectively realize transmission adaptation within a multi-DoF MCP mechanism, an output-side passive CVT employing a moving-pulley-inspired wire-routing structure is introduced. The resultant force generated by the wire tensions acting on the pulley that passively increases the flexion moment arm and transmission ratio according to the applied load without additional actuators, sensors, or control. Experimental results demonstrate a maximum output-force amplification of 4.19-fold and a mean amplification of 3.6-fold across the tested flexion angles ranging from 15 degrees to 75 degrees through moment-arm adaptation, thereby substantially expanding the achievable force-speed operating range. Furthermore, dexterous ball-rolling experiments verify that passive transmission adaptation can be achieved while preserving abduction-adduction functionality. These results demonstrate a scalable transmission design strategy for compact multi-DoF robotic hands.

Chinese Translation

本文提出了一种紧凑型的两自由度（DoF）机器人手指，配备了选择性被动连续可变传动（CVT），以实现广泛的力-速度操作范围。受到人类掌指关节（MCP）功能差异化的启发，所提出的机制通过选择性地将被动CVT分配给屈伸自由度，同时为外展-内收自由度保留直接传动，实现了自由度特定的传动差异化。为了实现广泛的力-速度操作范围，在屈伸路径中嵌入了一种力响应的被动CVT，同时为外展-内收路径保留了直接传动。为了在多自由度MCP机制中选择性地实现传动适应，提出了一种采用移动滑轮启发的线缆布置结构的输出侧被动CVT。通过施加的负载，作用在滑轮上的线缆张力产生的结果力被动地增加了屈伸力臂和传动比，而无需额外的执行器、传感器或控制。实验结果表明，在测试的屈伸角度范围从15度到75度内，最大输出力放大倍数为4.19倍，平均放大倍数为3.6倍，通过力臂适应显著扩展了可实现的力-速度操作范围。此外，灵巧的球滚动实验验证了在保留外展-内收功能的同时可以实现被动传动适应。这些结果展示了一种适用于紧凑型多自由度机器人手的可扩展传动设计策略。

View on arXiv Download PDF AI Translation

cs.RO / 18 / 2609.00771

Non-Prehensile Throwing: A Reinforcement Learning Perspective

非抓握投掷：强化学习视角

Mustafa, Abdullah, Hanai, Ryo, Ramirez-Alpizar, Ixchel G., Erich, Floris, Nakajo, Ryoichi, Domae, Yukiyasu, Ogata, Tetsuya

Abstract

Robotic throwing enables fast object transport and extends a robot's reachable workspace beyond traditional pick-and-place. While prehensile (grasp-based) throwing works well for graspable items, non-prehensile (grasp-free) throwing is better suited for large, heavy, and/or deformable objects. Existing approaches rely on model-based optimization with simplified contact models (e.g., dynamic grasping) and low-dimensional trajectory parameterizations, which limit solution quality and reachable workspace. We propose a reinforcement learning approach that additionally leverages sliding and rolling contact modes and directly optimizes joint-space trajectories without analytical contact models or custom parameterizations. The Markov Decision Process (MDP) is formulated as a dynamical system that evolves the robot's joint state conditioned on the throwing target, object model, and initial configuration. Joint-jerk trajectories are planned offline at a low control rate and upsampled into smooth, high-rate velocity commands for deployment. For sim-to-real transfer, we minimize the robot-dynamics gap through minimum-jerk system identification and train uncertainty-aware policies to mitigate object-modeling errors, particularly sensitivity to dynamic friction. In simulation, the policy achieves 99% success across thousands of configurations and generalizes to unseen objects. Sensitivity analysis shows robustness to mass uncertainty but high sensitivity to dynamic friction, consistent with the sliding-based release mechanism. Deployed zero-shot on a UR5e operating near its physical limits (5 m/s end-effector velocity), our method throws diverse objects including heavy (790 g) and large (20x20x28 cm) items to targets up to 350 cm distance or 180 cm elevation, achieving a 97% real-world success rate.

Chinese Translation

机器人投掷能够快速运输物体，并将机器人的可达工作空间扩展到传统的抓取和放置之外。虽然抓握（基于抓取）投掷适用于可抓取物品，但非抓握（无抓取）投掷更适合大型、重型和/或可变形物体。现有方法依赖于基于模型的优化，使用简化的接触模型（例如，动态抓取）和低维轨迹参数化，这限制了解决方案的质量和可达工作空间。我们提出了一种强化学习方法，额外利用滑动和滚动接触模式，并直接优化关节空间轨迹，而无需解析接触模型或自定义参数化。马尔可夫决策过程（MDP）被构建为一个动态系统，该系统根据投掷目标、物体模型和初始配置演化机器人的关节状态。关节加加速度轨迹在低控制率下离线规划，并上采样为平滑的高频率速度指令以供部署。为了实现模拟到现实的转移，我们通过最小加加速度系统识别最小化机器人动力学差距，并训练不确定性感知策略以减轻物体建模误差，特别是对动态摩擦的敏感性。在模拟中，该策略在数千种配置中实现了99%的成功率，并能够推广到未见过的物体。敏感性分析表明对质量不确定性具有鲁棒性，但对动态摩擦高度敏感，这与基于滑动的释放机制一致。在UR5e上零样本部署，该机器人在其物理极限附近操作（末端执行器速度为5 m/s），我们的方法能够将各种物体（包括重达790克和尺寸为20x20x28厘米的大型物品）投掷到距离达350厘米或高度达180厘米的目标，实际成功率达到97%。

View on arXiv Download PDF AI Translation

cs.RO / 19 / 2609.00804

Connectivity-Aware Graph Extension for Decentralized Multi-Robot Exploration

面向连通性的图扩展策略用于去中心化多机器人探索

Cegarra, Béatrice Garcia, Vanneaux, Elena, Picard, Quentin, Filliat, David

Abstract

Exploring unknown environments with multiple UAVs requires coordination under intermittent communication, making decentralized operation a baseline assumption. We propose, within a decentralized framework, a novel exploration graph extension strategy based on frontier connectivity to extend exploration plans and maintain area partitioning among agents stable and robust to disconnections and changes in spatial layout. The proposed extension method is applied to two state-of-the-art area partitioning methods and evaluated in simulation. Experiments show improved performance over existing graph extension approaches with higher exploration efficiency under low communication rate.

Chinese Translation

使用多架无人机（UAV）探索未知环境需要在间歇性通信下进行协调，因此去中心化操作成为基本假设。在一个去中心化框架内，我们提出了一种基于边界连通性的创新探索图扩展策略，以扩展探索计划，并保持代理之间的区域划分在断开连接和空间布局变化时的稳定性和鲁棒性。所提出的扩展方法应用于两种最先进的区域划分方法，并在仿真中进行了评估。实验结果表明，在低通信速率下，该方法在探索效率上优于现有的图扩展方法。

View on arXiv Download PDF AI Translation

cs.RO / 20 / 2609.00906

Peg-in-Bench: A Modular Benchmark for High-Precision Robotic Insertion

Peg-in-Bench：一种用于高精度机器人插入的模块化基准测试

Delgado, Yosel, Buenaventura-Carreón, José G., Erich, Floris, Mykhailyshyn, Roman, Motoda, Tomohiro, Makihara, Koshi, Domae, Yukiyasu

Abstract

High-precision insertion remains a fundamental challenge in robotic manipulation due to the strict alignment requirements and contact-rich interactions involved. Although peg-in-hole tasks are widely used for evaluation, existing bench- marks often rely on fixed task configurations, limiting their ability to assess robustness and generalization across different insertion scenarios. This paper introduces a reconfigurable peg-in-hole benchmark designed to evaluate task generalization in high-precision insertion. The benchmark consists of a set of fully 3D-printable modular components, including multiple peg geometries, tolerance levels, and configurable base structures that can be combined to generate a large variety of insertion and assembly tasks. By varying object layouts, orientations, and task structures while maintaining controlled physical conditions, the benchmark enables systematic evaluation of adaptation to unseen scenarios. To support reproducibility, we additionally provide a scenario generation tool capable of producing standardized task configurations and machine-readable task descriptions. The scenario generation tool and the STL files of the benchmark pieces are available through the project repository: https://github.com/aistairc/peg-in-bench.

Chinese Translation

高精度插入仍然是机器人操作中的一个基本挑战，因为涉及严格的对齐要求和丰富的接触互动。尽管 peg-in-hole 任务被广泛用于评估，现有的基准测试往往依赖于固定的任务配置，限制了它们在不同插入场景中评估鲁棒性和泛化能力的能力。本文介绍了一种可重构的 peg-in-hole 基准测试，旨在评估高精度插入中的任务泛化。该基准测试由一组完全可3D打印的模块化组件组成，包括多种 peg 几何形状、容差水平和可配置的基础结构，这些组件可以组合生成多种插入和组装任务。通过改变物体布局、方向和任务结构，同时保持受控的物理条件，该基准测试能够系统地评估对未见场景的适应能力。为了支持可重复性，我们还提供了一种场景生成工具，能够生成标准化的任务配置和机器可读的任务描述。场景生成工具和基准测试组件的 STL 文件可通过项目仓库获取：https://github.com/aistairc/peg-in-bench。

View on arXiv Download PDF AI Translation

cs.RO / 21 / 2609.00908

Knowing When to Stop: Adaptive Action Chunking via Internal Cross-Attention Dynamics in VLAs

知道何时停止：通过内部交叉注意力动态实现的自适应动作分块在视觉-语言-动作中的应用

Xu, Runze, Shan, Xiaolong, Dai, Shuang, Wang, Yu, Yu, Jincheng

Abstract

Action chunking is a standard execution strategy in modern Vision-Language-Action (VLA) frameworks, but fixed execution horizons impose a trade-off between efficiency and accuracy. Short chunks require frequent inference and may cause oscillatory behavior, whereas long chunks can become misaligned with newly observed states. We address this limitation with an adaptive action chunking approach based on internal cross-attention dynamics in the action expert. We observe that, as the prediction horizon extends, action-to-observation cross-attention becomes increasingly dispersed and its entropy rises toward a plateau. This pattern is associated with higher action prediction error and provides an online signal that the current observation offers limited grounding for further open-loop execution. Based on this observation, we introduce a training-free truncation mechanism that detects sustained high-entropy plateaus and dynamically selects the execution horizon during inference. The method uses attention weights already computed by the policy and introduces negligible additional overhead. Evaluations on $\pi_{0.5}$ and X-VLA across RoboTwin 2.0, LIBERO, and three real-world manipulation tasks show improved average task success over fixed-horizon and adaptive chunking baselines, while preserving efficient closed-loop control. These results show that cross-attention dynamics can provide a practical internal signal for adaptive action execution in VLAs.

Chinese Translation

动作分块是现代视觉-语言-动作（VLA）框架中的一种标准执行策略，但固定的执行时间范围在效率和准确性之间存在权衡。短的分块需要频繁推理，可能导致振荡行为，而长的分块可能与新观察到的状态失去一致性。我们通过基于动作专家内部交叉注意力动态的自适应动作分块方法来解决这一限制。我们观察到，随着预测时间范围的延长，动作与观察之间的交叉注意力变得越来越分散，其熵值上升并趋向于平稳。这一模式与较高的动作预测误差相关，并提供了一个在线信号，表明当前观察对进一步的开环执行提供了有限的基础。基于这一观察，我们引入了一种无训练的截断机制，能够检测持续的高熵平稳状态，并在推理过程中动态选择执行时间范围。该方法利用政策已经计算的注意力权重，增加的额外开销微乎其微。在RoboTwin 2.0、LIBERO和三个真实世界操作任务上的$ ext{π}_{0.5}$和X-VLA评估显示，与固定时间范围和自适应分块基线相比，平均任务成功率有所提高，同时保持了高效的闭环控制。这些结果表明，交叉注意力动态可以为VLA中的自适应动作执行提供实用的内部信号。

View on arXiv Download PDF AI Translation

cs.RO / 22 / 2609.00920

VerNav: Verifier-First Low-Latency Vision-and-Language Navigation

VerNav：以验证器为先的低延迟视觉与语言导航

Wang, Zhixin, Yao, Chengzheyi, Liu, Leyuan, Zhang, Xiaosong, Zhang, Yongzhao

Abstract

Vision-and-Language Navigation (VLN) requires an agent to navigate through unseen 3D environments according to natural-language instructions. Explicit reasoning can improve instruction understanding and semantic grounding, but autoregressive generation at every step accumulates large decision-stage latency over multi-step navigation. We propose VerNav, a verifier-first framework for low-latency LLM-based VLN. The verifier reduces decision-stage latency by replacing per-step autoregressive generation with batched action verification, while an entropy-based adaptive generator is invoked only for uncertain decisions to produce compact state evidence. To further improve navigation performance with the verifier, we introduce a two-stage alignment scheme: (i) VPO improves local action-preference alignment in static verifier training, and (ii) step-level reinforcement fine-tuning provides dense progress rewards over multi-step navigation rollouts during dynamic task execution. Experiments on the Room-to-Room (R2R) benchmark show that the verifier-only decision path of VerNav achieves competitive navigation performance among representative LLM-based VLN agents while reducing average decision-stage LLM latency per step by more than $10\times$ compared with autoregressive methods.

Chinese Translation

视觉与语言导航（VLN）要求智能体根据自然语言指令在未见过的三维环境中进行导航。显式推理可以改善指令理解和语义基础，但在每一步的自回归生成会在多步导航中累积较大的决策阶段延迟。我们提出了VerNav，一种基于低延迟大型语言模型（LLM）的验证器优先框架。该验证器通过用批量动作验证替代每一步的自回归生成，从而减少决策阶段的延迟，而基于熵的自适应生成器仅在不确定决策时被调用，以产生紧凑的状态证据。为了进一步提高与验证器的导航性能，我们引入了一个两阶段对齐方案：（i）VPO在静态验证器训练中改善局部动作偏好对齐，(ii) 步骤级强化微调在动态任务执行期间提供多步导航回放的密集进展奖励。针对Room-to-Room（R2R）基准的实验表明，VerNav的仅验证器决策路径在代表性的基于LLM的VLN智能体中实现了具有竞争力的导航性能，同时与自回归方法相比，每步的平均决策阶段LLM延迟减少了超过10倍。

View on arXiv Download PDF AI Translation

cs.RO / 23 / 2609.00941

ProxPI: Proximal Prior Injection for Sampling-Based MPC under Learned-Prior Mismatch

ProxPI：针对学习先验不匹配的基于采样的模型预测控制中的近端先验注入

Im, Euncheol, Lim, Myotaeg, Lee, Yisoo

Abstract

Combining learned policies with model predictive control can leverage learned task priors while retaining online adaptation to new objectives and constraints, but performance degrades when the policy is out of distribution. In policy-guided model predictive path integral (MPPI) control, a policy-centered warm-start approach centers the sampling distribution on the policy output. When the prior is mismatched, centering the sampling distribution on the policy output restricts exploration around an unsuitable solution and prevents recovery toward the task optimum. We propose Proximal Prior Injection (ProxPI), which retains nominal-centered MPPI sampling and incorporates the policy through a soft proximity cost. This matches the in-distribution performance of existing prior-injection schemes while enabling the optimizer to escape an inaccurate policy and recover vanilla MPPI-level performance. We theoretically show that re-centering on the prior discards the optimizer's correction at every update, whereas nominal-centered sampling retains it and converges to a solution set by both the task cost and the prior, and that this failure is not removed by a larger rollout budget. Simulations and real-robot experiments demonstrate robust performance under both in-distribution and out-of-distribution tasks.

Chinese Translation

将学习到的策略与模型预测控制相结合，可以利用学习到的任务先验，同时保持对新目标和约束的在线适应性，但当策略超出分布时，性能会下降。在策略引导的模型预测路径积分（MPPI）控制中，基于策略的热启动方法将采样分布集中在策略输出上。当先验不匹配时，将采样分布集中在策略输出上会限制对不合适解的探索，并阻碍向任务最优解的恢复。我们提出了近端先验注入（Proximal Prior Injection, ProxPI），该方法保留了名义中心的MPPI采样，并通过软接近成本将策略纳入其中。这使得其在分布内的性能与现有的先验注入方案相匹配，同时使优化器能够摆脱不准确的策略，恢复到普通MPPI水平的性能。我们理论上证明了在每次更新时重新集中在先验上会丢弃优化器的修正，而名义中心采样则保留了这一修正，并收敛到由任务成本和先验共同设定的解集，而这一失败并不会通过更大的回滚预算来消除。仿真和真实机器人实验展示了在分布内和分布外任务下的稳健性能。

View on arXiv Download PDF AI Translation

cs.RO / 24 / 2609.00950

HitMem: Hierarchical Temporal 3D Memory with Multi-Modal Context-Aware Retrieval for Dynamic Environments

HitMem：具有多模态上下文感知检索的层次时间3D记忆框架用于动态环境

Tang, Ruijie, Zou, Chenye, Wu, Guoquan, Wei, Jun, Chen, Wei, Zhu, Jiaxin

Abstract

Executing long-term tasks in dynamic environments requires embodied agents to maintain robust and adaptive 3D scene representations. However, most existing 3D memory frameworks rely on static world assumptions. When objects are displaced by human activities or unobserved events, agents encounter memory-observation conflicts and often require costly geometric recomputations or inefficient global re-exploration. To address this, we propose HitMem, a hierarchical temporal 3D memory framework with a multi-modal context-aware retrieval mechanism. Through continuous perception, HitMem unifies semantic and spatial information into a lightweight topological graph that captures support relationships, while a temporal decay mechanism dynamically regulates memory activeness to mitigate the impact of stale representations. In addition, the multi-modal context-aware retrieval mechanism defaults to filtering candidates using integrated semantic, spatial, and temporal memory features, and activates a specialized two-stage retrieval process when object displacement is detected. This process combines spatial constraints inferred from external agent trajectories with semantic common sense grounded in class affinities, efficiently identifying high-probability candidate regions. Extensive evaluations on our constructed Dyna-THOR benchmark demonstrate that HitMem significantly improves object relocation accuracy, reduces exploration costs, and enhances task execution performance in dynamic environments.

Chinese Translation

在动态环境中执行长期任务要求具身代理保持稳健且适应性强的3D场景表示。然而，大多数现有的3D记忆框架依赖于静态世界假设。当物体因人类活动或未观察到的事件而被移动时，代理会遇到记忆与观察之间的冲突，通常需要昂贵的几何重新计算或低效的全局重新探索。为了解决这个问题，我们提出了HitMem，一个具有多模态上下文感知检索机制的层次时间3D记忆框架。通过持续感知，HitMem将语义和空间信息统一为一个轻量级拓扑图，以捕捉支持关系，同时，时间衰减机制动态调节记忆的活跃性，以减轻陈旧表示的影响。此外，多模态上下文感知检索机制默认使用集成的语义、空间和时间记忆特征过滤候选项，并在检测到物体位移时激活专门的两阶段检索过程。该过程结合了从外部代理轨迹推断的空间约束与基于类别亲和力的语义常识，有效识别高概率候选区域。在我们构建的Dyna-THOR基准上的广泛评估表明，HitMem显著提高了物体重定位的准确性，降低了探索成本，并增强了在动态环境中的任务执行性能。

View on arXiv Download PDF AI Translation

cs.RO / 25 / 2609.01061

Accelerating Reinforcement Learning via MPC Solver-Gradient Guidance for Weights-varying MPC

通过MPC求解器梯度引导加速权重变化的MPC强化学习

Zarrouki, Baha, Thobani, Arslan, Hoffmann, Jasper, Piccinini, Mattia, Reiter, Rudolf, Jahncke, Felix, Gros, Sébastien, Scaramuzza, Davide, Betz, Johannes

Abstract

In Model Predictive Control (MPC), cost-function weights shape closed-loop behavior, yet changing conditions often make fixed parametrizations suboptimal and motivate context-dependent online adaptation. Learning such policies is difficult because behavior depends implicitly on numerical MPC solutions, producing nonlinear, potentially nonsmooth, long-horizon dependencies on policy parameters. This creates a bias-variance tradeoff: Reinforcement Learning (RL) optimizes realized closed-loop return from environment samples but is sample-inefficient, whereas Gradient-Based Policy Learning (GB-PL) uses low-variance solver gradients from differentiable MPC to optimize surrogate losses on predicted trajectories but can be biased under model mismatch. We propose Solver-Gradient Guided Reinforcement Learning (SG-RL), a solver-sensitivity augmentation for RL-based online MPC cost-weight adaptation. SG-RL keeps sampled closed-loop return as the objective and uses bounded solver-derived gradients as auxiliary guidance to improve stability and sample efficiency. We instantiate SG-RL in Proximal Policy Optimization (PPO) with four modular algorithms that inject solver-gradient guidance into actor-update scaling, policy loss, advantage estimation, and value-function learning. On two full-scale autonomous racing platforms with intentional model mismatch, SG-RL reaches PPO's best closed-loop return with up to 70.6% fewer samples, outperforms GB-PL baselines by at least 54% in closed-loop return, and generalizes zero-shot to unseen environments.

Chinese Translation

在模型预测控制（MPC）中，成本函数权重塑造了闭环行为，但变化的条件常常使固定参数化变得次优，从而促使基于上下文的在线适应。学习这样的策略是困难的，因为行为隐含地依赖于数值MPC解决方案，产生对策略参数的非线性、潜在不光滑的长期依赖。这造成了偏差-方差权衡：强化学习（RL）优化来自环境样本的实际闭环回报，但样本效率低，而基于梯度的策略学习（GB-PL）利用来自可微分MPC的低方差求解器梯度来优化预测轨迹上的替代损失，但在模型不匹配时可能存在偏差。我们提出了解决器梯度引导的强化学习（SG-RL），这是一种基于RL的在线MPC成本权重适应的求解器敏感性增强方法。SG-RL将采样的闭环回报作为目标，并使用有界的求解器导出梯度作为辅助指导，以提高稳定性和样本效率。我们在近端策略优化（PPO）中实例化SG-RL，采用四种模块化算法将求解器梯度引导注入到演员更新缩放、策略损失、优势估计和价值函数学习中。在两个具有故意模型不匹配的全尺度自主赛车平台上，SG-RL以最多减少70.6%的样本达到了PPO的最佳闭环回报，闭环回报至少比GB-PL基线提高了54%，并且在未见环境中实现了零样本泛化。

View on arXiv Download PDF AI Translation

cs.RO / 26 / 2609.01089

Adaptive Depth-Map-Guided Bundle Adjustment for Correspondence-Free Multi-View Point Cloud Registration

自适应深度图引导的无对应多视角点云配准束调整

Zhou, Yiran, Wang, Yingyu, Huang, Shoudong, Zhao, Liang

Abstract

Robotic processing of irregular steel scrap requires dense 3-D measurement to replace manual visual assessment in hazardous cutting workcells. The reconstructed map is used to estimate piece dimensions, boundary geometry, feasible preheating and cutting regions, and collision-aware torch paths. The reconstruction errors therefore propagate directly to downstream measurement and planning. Existing multi-view registration methods commonly rely on feature extraction and data association to establish correspondences between views. In workcells with smooth metallic surfaces, repeated structures, occlusions, and partial overlaps, however, wrong correspondences may be established, leading to inaccurate pose estimation and distorted reconstruction. This paper presents an adaptive layered depth-map-guided bundle adjustment framework for correspondence-free multi-view point cloud registration. The scene is represented by a global 2.5-D grid, where each cell can adaptively maintain multiple depth hypotheses. Raw depth observations are directly projected into the global map to form depth constraints without explicit feature correspondences. At grid cells where multiple surfaces produce conflicting depths, a softmax-based layer assignment links each observation to compatible depth hypotheses. The resulting nonlinear least-squares formulation jointly refines sensor poses and the layered depth map, with correspondences implicitly induced by the depth-map representation and projection model. Experiments on self-collected industrial datasets show that the proposed method achieves consistently competitive reconstruction accuracy while maintaining robustness and low computational cost in challenging industrial scenarios. We release the open-source code implementation at: https://github.com/YiranZhou-Robotics/ADM-BA.git

Chinese Translation

机器人处理不规则钢铁废料需要密集的三维测量，以替代在危险切割工作单元中的人工视觉评估。重建的地图用于估计工件尺寸、边界几何形状、可行的预热和切割区域，以及考虑碰撞的切割路径。因此，重建误差会直接传播到下游测量和规划。现有的多视角配准方法通常依赖于特征提取和数据关联来建立视图之间的对应关系。然而，在具有光滑金属表面、重复结构、遮挡和部分重叠的工作单元中，可能会建立错误的对应关系，导致姿态估计不准确和重建失真。本文提出了一种自适应分层深度图引导的无对应多视角点云配准束调整框架。场景通过一个全局2.5-D网格表示，其中每个单元可以自适应地维护多个深度假设。原始深度观测被直接投影到全局地图中，以形成深度约束，而无需显式的特征对应关系。在多个表面产生冲突深度的网格单元中，基于softmax的层分配将每个观测与兼容的深度假设关联。最终的非线性最小二乘公式共同优化传感器姿态和分层深度图，通过深度图表示和投影模型隐式引导对应关系。在自收集的工业数据集上的实验表明，所提出的方法在具有挑战性的工业场景中实现了始终如一的竞争性重建精度，同时保持了鲁棒性和低计算成本。我们在以下网址发布了开源代码实现：https://github.com/YiranZhou-Robotics/ADM-BA.git

View on arXiv Download PDF AI Translation

cs.RO / 27 / 2609.01120

DNC-IMM: Early Lane-Change Intention Recognition via Neural Calibration Based on Driving Context Information

DNC-IMM：基于驾驶上下文信息的早期变道意图识别的神经校准方法

Byun, Woong-Chan, Kong, Seung-Hyun

Abstract

Early recognition of lane-change intention is essential for proactive decision-making in autonomous driving and advanced driver assistance systems. This paper proposes a Dual Neural-Calibrated Interacting Multiple Model (DNC-IMM) that improves adaptability to driving context while preserving the probabilistic structure and interpretability of a conventional IMM. The proposed method encodes driving-context information, including target-vehicle motion, gaps to surrounding vehicles, and relative velocities, with a neural network that calibrates both the transition-probability matrix and measurement likelihoods. The final intention is determined from the calibrated IMM mode posterior rather than from a separate direct classifier. Experiments on the highD dataset demonstrate that the proposed method reliably recognizes lane-change intentions before lane crossing and provides particularly strong performance at the earlier 2-3 s prediction horizons.

Chinese Translation

早期识别变道意图对于自主驾驶和高级驾驶辅助系统中的主动决策至关重要。本文提出了一种双神经校准交互多模型（DNC-IMM），该模型在保持传统交互多模型（IMM）概率结构和可解释性的同时，提高了对驾驶上下文的适应性。所提出的方法通过神经网络编码驾驶上下文信息，包括目标车辆运动、与周围车辆的间隙和相对速度，并校准转移概率矩阵和测量似然性。最终的意图是从校准后的IMM模式后验中确定，而不是通过单独的直接分类器。对highD数据集的实验表明，所提方法能够可靠地在变道前识别变道意图，并在早期的2-3秒预测范围内表现出特别强的性能。

View on arXiv Download PDF AI Translation

cs.RO / 28 / 2609.01207

On Global Regulatability of Robot Manipulators by Classical PID

经典PID对机器人操纵器的全局可调性研究

Zhao, Cheng, Zhu, Jingru, Guo, Lei

Abstract

This paper studies a class of uncertain multi-input multi-output (MIMO) nonlinear systems using extended PID (EPID) control. We focus on systems possessing a well-defined vector relative degree whose components may vary across channels, a setting that received limited attention in the existing literature on PID-type control. We develop a channel-wise EPID controller, where each control input is constructed from the proportional, integral, and derivative terms of the corresponding tracking error, and the highest derivative order is selected according to the relative degree of that channel. Under suitable growth conditions on the unknown nonlinearities, we construct an admissible set of EPID parameters specified by prescribed initial-state bounds, the uncertainties and the reference signal. We show that any parameter choice from this set guarantees semiglobal stability of the closed-loop system and achieves the desired setpoint regulation. These results indicate that the EPID parameters can be designed independently for each channel, justifying the common engineering practice of channel-by-channel tuning in strongly coupled and uncertain MIMO systems, including flight control systems.

Chinese Translation

本文研究了一类不确定的多输入多输出（MIMO）非线性系统，采用扩展PID（EPID）控制。我们关注于具有明确向量相对度的系统，其各个分量可能在不同通道间变化，这一设置在现有的PID类型控制文献中受到的关注有限。我们开发了一种通道级EPID控制器，其中每个控制输入由相应跟踪误差的比例、积分和导数项构成，且最高导数阶数根据该通道的相对度进行选择。在对未知非线性进行适当增长条件的假设下，我们构建了一组可接受的EPID参数集，该参数集由规定的初始状态界限、不确定性和参考信号指定。我们证明，从该参数集中选择的任何参数都能保证闭环系统的半全局稳定性，并实现所需的设定点调节。这些结果表明，EPID参数可以针对每个通道独立设计，从而证明在强耦合和不确定的MIMO系统（包括飞行控制系统）中逐通道调节的常见工程实践是合理的。

View on arXiv Download PDF AI Translation

cs.RO / 29 / 2609.01281

EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents

具身技能：一个统一框架用于协调、训练和部署 VLA 代理

Wang, Wei, Zhang, Wenqiao, Lin, Yutong, Yuan, Yuqian, Lin, Tianwei, Mao, Jinhao, Fan, Zhenxuan, Gao, Mingjian, Dai, Yang, Li, Wentong, Lv, Zheqi, Dong, Zheng, Niu, Yingjie, Zhu, Jiaqi, Xiao, Jun, Li, Chao, Zhuang, Yueting

Abstract

Vision-language-action (VLA) models map visual observations and language instructions directly to robot actions, but long-horizon tasks require more than action prediction. An agent must coordinate perception, planning, execution, progress verification, and recovery as the physical state evolves. An action prediction or a model-generated skill decision does not, by itself, guarantee that the proposed operation is valid in the current state or that its outcome will be verified. We propose EmbodiedSkills, a unified framework that treats each skill decision as an execution proposal: the runtime checks its prerequisites before execution and verifies the outcome afterward. A shared executable-skill interface connects high-level skill selection, bounded low-level VLA execution, and post-action verification within a single agent loop. Because this interface remains fixed, low-level VLA policies can be replaced or adapted without changing the agent loop. The interface also records planning, execution, verification, and recovery events as structured trajectories, which provide supervision for individual components and can support optional online adaptation when interactive feedback is available. We instantiate EmbodiedSkills with Qwen3-VL and OpenPI/pi0.5 on RoboTwin 2.0 and LIBERO. Task-adapted low-level VLA policies achieve an average success rate of 86.20% across 50 RoboTwin 2.0 tasks and 97.40% across the four LIBERO suites. These results establish the execution performance of the task-adapted low-level VLA policies used in EmbodiedSkills. On four memory-dependent RMBench tasks, the same task-adapted execution approach achieves 12.5% average success. The framework provides a trainable and inspectable agent layer for turning these policies into closed-loop embodied systems.

Chinese Translation

视觉-语言-动作（VLA）模型将视觉观察和语言指令直接映射到机器人动作，但长时间任务需要的不仅仅是动作预测。代理必须在物理状态演变的过程中协调感知、规划、执行、进度验证和恢复。单靠动作预测或模型生成的技能决策并不能保证所提议的操作在当前状态下是有效的，或其结果会被验证。我们提出了具身技能（EmbodiedSkills），一个统一框架，将每个技能决策视为执行提案：运行时在执行前检查其前提条件，并在执行后验证结果。一个共享的可执行技能接口将高层技能选择、有限低层 VLA 执行和后续动作验证连接在单一代理循环中。由于该接口保持固定，低层 VLA 策略可以在不改变代理循环的情况下被替换或调整。该接口还将规划、执行、验证和恢复事件记录为结构化轨迹，为各个组件提供监督，并在可用的互动反馈下支持可选的在线适应。我们在 RoboTwin 2.0 和 LIBERO 上用 Qwen3-VL 和 OpenPI/pi0.5 实现了具身技能。经过任务适应的低层 VLA 策略在 50 个 RoboTwin 2.0 任务中实现了 86.20% 的平均成功率，在四个 LIBERO 套件中实现了 97.40% 的成功率。这些结果确立了在具身技能中使用的任务适应低层 VLA 策略的执行性能。在四个依赖记忆的 RMBench 任务中，相同的任务适应执行方法实现了 12.5% 的平均成功率。该框架提供了一个可训练和可检查的代理层，将这些策略转变为闭环的具身系统。

View on arXiv Download PDF AI Translation

cs.RO / 30 / 2609.01339

Integrating Traffic Noise Emission Modelling into Variable Speed Limit Control

将交通噪声排放建模纳入可变限速控制

Meng, Jiawen, Arockiasamy, John Pravin, Vinel, Alexey

Abstract

Road traffic noise remains a major environmental challenge, yet most speed management strategies are static and do not respond to short-term variations in traffic noise emissions. Although variable speed limit (VSL) systems are widely deployed for safety and congestion mitigation, traffic noise is rarely treated as an explicit operational control objective. This paper proposes a noise-aware VSL framework that integrates aggregated traffic-state estimation with a simplified CNOSSOS-EU-based emission indicator. A stage-based controller with time-varying reference thresholds dynamically adjusts discrete speed-limit levels in response to estimated emission conditions. The framework is evaluated using microscopic traffic simulation calibrated with empirical motorway data and replicated across multiple stochastic realisations. Over a 24-hour evaluation period, the adaptive strategy reduces the receiver-based equivalent sound level by 2.9 dB(A) relative to unrestricted traffic conditions, while maintaining an average vehicle speed approximately 11.3 km/h higher than a permanently imposed low-speed regime. Period-wise analysis shows that speed reductions are activated selectively when emission levels approach calibrated targets, rather than enforcing a constant intermediate limit. Traffic stability indicators reveal moderate increases in speed variability compared with unrestricted operation, but substantially lower braking intensity than under uniform low-speed enforcement. These results demonstrate the feasibility of integrating environmental performance indicators into operational speed control, providing a practical complement to conventional infrastructure-based noise mitigation measures.

Chinese Translation

道路交通噪声仍然是一个主要的环境挑战，但大多数速度管理策略是静态的，无法应对交通噪声排放的短期变化。尽管可变限速（VSL）系统被广泛应用于安全和拥堵缓解，但交通噪声很少被视为明确的操作控制目标。本文提出了一种噪声感知的VSL框架，该框架将聚合的交通状态估计与简化的基于CNOSSOS-EU的排放指标相结合。该框架采用基于阶段的控制器，具有时间变化的参考阈值，能够根据估计的排放条件动态调整离散的限速水平。通过使用经过实证高速公路数据校准的微观交通模拟对该框架进行评估，并在多个随机实现中进行复制。在24小时的评估期间，该自适应策略相较于无限制交通条件下，降低了接收器基于的等效声级2.9 dB(A)，同时保持的平均车速约为11.3 km/h，高于永久施加的低速限制。周期性分析显示，当排放水平接近校准目标时，速度降低是选择性激活的，而不是强制施加一个恒定的中间限制。交通稳定性指标表明，与无限制操作相比，速度变异性适度增加，但刹车强度显著低于统一的低速执法。这些结果证明了将环境绩效指标纳入操作速度控制的可行性，为传统基础设施噪声缓解措施提供了实际的补充。

View on arXiv Download PDF AI Translation

cs.RO / 31 / 2609.01351

Scalable Rao-Blackwellized Online Planning for High-Dimensional POMDPs

可扩展的 Rao-Blackwell 化在线规划用于高维 POMDPs

Lee, Jiho, Ahmed, Nisar, Wray, Kyle Hollins, Sunberg, Zachary

Abstract

Online planning under uncertainty remains a fundamental challenge for robotic systems operating in partially observable environments with high-dimensional state spaces. While sampling-based POMDP solvers enable approximate decision-making in large or continuous domains, their performance degrades as belief dimensionality increases due to the high variance inherent in Monte Carlo-based estimation. In this work, we extend the Rao-Blackwellized online POMDP (RB-POMDP) framework to improve its generalizability in high-dimensional settings through hybrid continuous-discrete belief representations. By analytically propagating uncertainty associated with marginalized state components during tree-based planning, the proposed approach reduces sampling-induced variance in value estimation. We demonstrate the effectiveness of this framework in a robotic search-and-rescue task by integrating it with FastSLAM 2.0. Experimental results show that the proposed planner achieves higher cumulative rewards using significantly fewer particles and planning simulations than purely sampling-based methods under equivalent computational budgets. These results suggest that structured high-dimensional robotic problems admitting tractable sufficient statistics can be effectively leveraged within the RB-POMDP framework for computationally feasible online decision-making.

Chinese Translation

在不确定性下的在线规划仍然是机器人系统在部分可观测环境中操作时面临的一个基本挑战，尤其是在高维状态空间中。虽然基于采样的 POMDP 求解器能够在大规模或连续域中实现近似决策，但随着信念维度的增加，由于蒙特卡洛估计固有的高方差，其性能会下降。在本研究中，我们扩展了 Rao-Blackwell 化在线 POMDP (RB-POMDP) 框架，通过混合连续-离散信念表示来提高其在高维环境中的可推广性。通过在基于树的规划过程中分析传播与边际状态组件相关的不确定性，所提出的方法减少了价值估计中由采样引起的方差。我们通过将该框架与 FastSLAM 2.0 集成，展示了其在机器人搜索与救援任务中的有效性。实验结果表明，在相同的计算预算下，所提出的规划器使用显著更少的粒子和规划模拟，获得了更高的累积奖励，优于纯粹基于采样的方法。这些结果表明，允许可处理充分统计量的结构化高维机器人问题可以在 RB-POMDP 框架内有效利用，以实现计算上可行的在线决策。

View on arXiv Download PDF AI Translation

cs.RO / 32 / 2609.01384

Obstacle-Aware Autonomous Coverage and Navigation for Outdoor Robots

面向户外机器人的障碍物感知自主覆盖与导航

Gargani, Leonardo, Frosi, Matteo, Matteucci, Matteo

Abstract

Long-duration outdoor coverage with autonomous platforms remains challenging beyond classical planning: deployments face localization drift in open spaces, obstacles in cluttered sites, controller feasibility in turn-heavy maneuvers, and persistent autonomy with energy management. We propose a unified ROS 2 architecture for outdoor coverage that combines coverage planning, robust localization, and Nav2-based execution. A dual-antenna RTK-GNSS fused in an EKF keeps the robot pose, both position and heading, accurate across long missions; three controller-aware refinements are added to a mature coverage planner; a Behavior-Tree mission manager coordinates multi-goal execution, layered recovery, cost-aware goal management, and autonomous docking for return-to-charge. We validate the stack through simulation and real-world trials across multiple outdoor areas with varying geometries and obstacle densities. Overall, these results show that the proposed stack can reliably complete outdoor coverage missions across varied areas, sweeping 93.1% to 96.1% of the planned coverage area.

Chinese Translation

在经典规划之外，长时间的户外覆盖仍然是一个挑战：部署面临开放空间中的定位漂移、杂乱场地中的障碍物、在转弯重的机动中控制器的可行性，以及能源管理下的持久自主性。我们提出了一种统一的 ROS 2 架构，用于户外覆盖，结合了覆盖规划、稳健的定位和基于 Nav2 的执行。通过在扩展卡尔曼滤波器（EKF）中融合的双天线 RTK-GNSS，保持机器人在长时间任务中的姿态（位置和航向）准确；在成熟的覆盖规划器中增加了三个控制器感知的改进；行为树（Behavior-Tree）任务管理器协调多目标执行、分层恢复、成本感知的目标管理以及自主对接以实现充电返回。我们通过模拟和在多个具有不同几何形状和障碍物密度的户外区域进行的实际试验验证了该系统。总体而言，这些结果表明，所提出的系统能够可靠地完成不同区域的户外覆盖任务，覆盖计划区域的93.1%至96.1%。

View on arXiv Download PDF AI Translation

cs.RO / 33 / 2609.01394

Autonomous robotic bridging using distributed swarm control without inter-agent communication

无代理间通信的分布式群体控制下的自主机器人桥接

Thamaraiselvan, Vishwaak C., Lundberg, Cody L., Theofandis, Michail, Chelian, Suhas, Gans, Nicholas R.

Abstract

We describe SCARAB--Swarm-Capable Autonomous Robotic Aquatic Bridging. Using distributed swarm control and multi-model sensing of agents and docking targets, agents can localize themselves, join into formations and proceed to desired target locations. Our technologies would eventually allow the Army to perform unpredictable, dispersed river crossings, enhance crew survivability, and minimize the logistics footprint compared to the current Improved Ribbon Bridge. Our methods operate without GPS or RF communications, though these can be used in non-contested environments (e.g., civilian disaster relief for flooding, etc.). We demonstrate our system via physics-engine-based simulation of several agents using unmanned surface vehicles (USVs) in the presence of currents and wind.

Chinese Translation

我们描述了SCARAB——具备群体能力的自主机器人水上桥接。通过分布式群体控制和多模型感知，代理能够自我定位，形成编队并前往目标位置。我们的技术最终将使军队能够进行不可预测的、分散的河流渡过，增强乘员生存能力，并相较于当前的改进型带状桥减少后勤足迹。我们的方法在没有GPS或射频通信的情况下运行，尽管在非竞争环境中（例如，民用洪灾救援等）可以使用这些技术。我们通过基于物理引擎的模拟展示了我们的系统，该模拟涉及在水流和风的影响下使用无人水面车辆（USVs）的多个代理。

View on arXiv Download PDF AI Translation

cs.RO / 34 / 2609.01404

Evaluating Multimodal LLMs as Generalist Vision-Language-Action Agents for Drone Control: Commanding, Approaching, Tracking and Searching

评估多模态大型语言模型作为无人机控制的通用视觉-语言-动作代理：指挥、接近、跟踪和搜索

Park, Jaewoo, Lee, Minyoung, Seo, Sukmin, Yim, Moonbin, Yoon, Hyunwook, Ryu, Dohoon, Kim, Daehee, Song, Myungseo, Byun, Jihyuk, Chang, Seunggyu, Kil, Taeho, Kim, Jiseob, Lee, Bado, Kim, Geewook

Abstract

Multimodal Large Language Models (MLLMs) are strong perceivers of images and video. We ask how far that reach extends into acting: dropping an MLLM directly into a drone's control loop, with its entire action space declared solely in the prompt. Recent systems approach this setting but increasingly narrow the model's decision-making. We widen it back. We introduce DroneCATS-Agent, an architecture where the MLLM is a swappable component, and DroneCATS, a benchmark treating the model as the independent variable. Beyond merely flying toward a pixel, our agent entrusts the model to yaw and search, deliberate when unsure, and self-declare arrival---all without fine-tuning or function-calling schemas. Evaluating frontier and open models across four core capabilities---approaching a visible target, tracking a moving one, searching outside the initial view, and commanding a multi-drone fleet---reveals that even the simplest embodied settings are far from solved. Crucially, to identify what breaks first at the edge, our roster scales down to 2B parameters. The findings expose a stark paradox: it is not the flying that fails. Small open models often navigate into the success radius more reliably than frontier models, yet lose the episode by declaring arrival prematurely or not at all. Multi-drone commanding amplifies this divide, with small models failing by blindly copying a single coordinate across distinct views. Viewed as vision-language-action agents, the models' spatial perception holds up, but their action protocol does not. What separates a deployable edge model from a frontier model is not navigation, but the discipline to sustain a declared protocol and emit the correct terminating action. The open problem is closing this gap at onboard compute costs---yielding a fast model that plans persistently and knows exactly when it is done---and DroneCATS is built to measure that distance.

Chinese Translation

多模态大型语言模型（MLLMs）在图像和视频感知方面表现出色。我们探讨这种能力在行动中的延伸程度：将MLLM直接嵌入无人机的控制循环中，其整个动作空间仅在提示中声明。近期的系统接近这一设置，但越来越多地限制了模型的决策能力。我们将其重新扩展。我们引入了DroneCATS-Agent，这是一种架构，其中MLLM是可替换的组件，以及DroneCATS，一个将模型视为自变量的基准测试。我们的代理不仅仅是朝着一个像素飞行，而是信任模型进行偏航和搜索，在不确定时进行深思，并自我声明到达——所有这些都无需微调或功能调用方案。在四个核心能力上评估前沿和开放模型——接近可见目标、跟踪移动目标、在初始视野之外搜索以及指挥多无人机编队——揭示即使是最简单的具身设置也远未解决。关键是，为了识别边缘处首先破裂的因素，我们的模型规模缩小到20亿参数。研究结果揭示了一个鲜明的悖论：失败的并不是飞行。小型开放模型往往比前沿模型更可靠地导航进入成功半径，但却因过早或根本不声明到达而失去这一回合。多无人机指挥加剧了这种差距，小型模型在不同视角间盲目复制单一坐标而失败。作为视觉-语言-动作代理来看，模型的空间感知能力表现良好，但其行动协议却不尽如人意。可部署的边缘模型与前沿模型的区别不在于导航，而在于维持声明协议并发出正确终止动作的能力。当前的开放问题是缩小这一差距，同时控制机载计算成本——实现一个快速的模型，能够持续规划并准确知道何时完成——而DroneCATS旨在测量这一距离。

View on arXiv Download PDF AI Translation

cs.RO / 35 / 2609.01420

Vision-Based Leader-Follower Formation Control for Cooperative UAVs in GPS-Degraded Environments

基于视觉的GPS退化环境下合作无人机的领-follow者编队控制

Angadi, Deekshitha, Budda, Naveena, Agarwal, Vikas, Mulasa, Rojesh Arunkumar, Killamsetty, Ravi, Samshad, Mohamed, Kemsaram, Narsimlu

Abstract

Cooperation in multi-UAV systems requires reliable relative perception so that follower vehicles can maintain formation and continue their mission safely even when absolute positioning sensors degrade or fail. This paper presents a vision-based cooperative formation framework running on a follower UAV that uses a front-facing RGB-D camera to detect, track, and localize a leader UAV in real-time. A lightweight YOLO-based detector is trained on a dedicated drone dataset and deployed onboard to predict leader bounding boxes, which are then fused with depth information via a pinhole camera model to estimate the leader's relative pose. These estimates provide a leader-follower position controller and can also be used as a backup when GPS or external localization is unavailable. This framework is implemented as a set of ROS nodes and evaluated in a physics-based multi-UAV simulation built on XTDrone, with sensor noise and communication dropouts. We evaluate detection accuracy, runtime, and formation-keeping error under nominal conditions and under simulated failures of the positioning sensors. The results show that the proposed framework maintains stable leader-follower formations with reasonable computational cost and provides a practical basis for extending vision-based cooperative formation control to real-world multi-UAV systems.

Chinese Translation

多无人机系统中的合作需要可靠的相对感知，以便跟随无人机能够维持编队并在绝对定位传感器退化或失效时安全地继续其任务。本文提出了一种基于视觉的合作编队框架，该框架在跟随无人机上运行，利用前置RGB-D相机实时检测、跟踪和定位领航无人机。一个轻量级的基于YOLO的检测器在专用无人机数据集上进行训练，并部署在机载系统中，以预测领航无人机的边界框，这些边界框随后通过针孔相机模型与深度信息融合，以估计领航无人机的相对姿态。这些估计结果为领-follow者位置控制器提供支持，并且在GPS或外部定位不可用时也可作为备份。该框架作为一组ROS节点实现，并在基于物理的多无人机仿真环境XTDrone中进行评估，考虑了传感器噪声和通信中断。我们在正常条件下以及在模拟定位传感器故障情况下评估了检测精度、运行时间和维持编队的误差。结果表明，所提出的框架能够以合理的计算成本维持稳定的领-follow者编队，并为将基于视觉的合作编队控制扩展到现实世界的多无人机系统提供了实用基础。

View on arXiv Download PDF AI Translation

cs.RO / 36 / 2609.01453

Does Imitation Learning Preserve Temporal Robustness in Dexterous Manipulation? An Expert-Learner Comparison Across Task Execution Speeds

模仿学习是否保持灵巧操作中的时间鲁棒性？专家与学习者在任务执行速度下的比较

Enwerem, Clinton, Baras, John S., Belta, Calin

Abstract

Dexterous manipulation policies learned by imitation are typically evaluated for robustness to variation in scenes, objects, or instructions, but their performance across task execution speeds is less often examined. This leaves open how much temporal robustness a learner retains relative to the expert it imitates. We compare an expert and learner under the same task conditions, initial-condition draws, and speedup factors. We instantiate the evaluation in ParcelStow, a contact-rich task in which the robot acquires, reorients, and inserts a parcel. The demonstrations span the speedup range for the manipulation phases after parcel acquisition. A scripted expert and an Action Chunking with Transformers (ACT) policy trained from the expert's demonstrations both achieve 100 percent task success at nominal speed. Their success rates diverge within the demonstrated range: at its maximum, expert success is 84 percent and ACT success is 53 percent. Two ACT policies with different parameter initializations show similar degradation, decreasing by 34 and 48 percentage points from nominal speed to the maximum demonstrated speed, compared with 16 points for the expert. Stage-level analysis shows that 35 of ACT's 47 failures at the maximum demonstrated speed are insertion misalignments. Under the relative-motion handoff, every ACT acquisition retains the parcel through reorientation and transfer in free space, but only 64 percent complete the overall task, compared with 95 percent after expert acquisition. Across all evaluated policies and speeds, none of the 414 acquisitions without force closure completes the task. Equal nominal task success therefore does not imply preservation of expert performance across execution speeds. Code, data, and evaluation scripts are available at https://github.com/coenwerem/parcelstow.

Chinese Translation

通过模仿学习的灵巧操作策略通常在场景、物体或指令变化的鲁棒性方面进行评估，但其在任务执行速度下的表现较少被研究。这使得我们不清楚学习者相对于其模仿的专家保留了多少时间鲁棒性。我们在相同的任务条件、初始条件抽取和加速因子下比较了专家和学习者。我们在ParcelStow中进行评估，这是一个接触丰富的任务，机器人需要获取、重新定向并插入一个包裹。演示涵盖了包裹获取后操作阶段的加速范围。一个脚本专家和一个基于专家演示训练的Action Chunking with Transformers (ACT) 策略在名义速度下均实现了100%的任务成功率。在演示的范围内，它们的成功率出现了分歧：在最大值时，专家的成功率为84%，而ACT的成功率为53%。两个具有不同参数初始化的ACT策略显示出类似的性能下降，从名义速度到最大演示速度分别下降了34和48个百分点，而专家仅下降了16个百分点。阶段级分析显示，在最大演示速度下，ACT的47次失败中有35次是由于插入对齐错误。在相对运动交接下，每次ACT获取都能通过重新定向和在自由空间中的转移保持包裹，但仅有64%完成了整体任务，而专家获取后则为95%。在所有评估的策略和速度中，414次没有力闭合的获取均未能完成任务。因此，相同的名义任务成功率并不意味着在执行速度上保持了专家的表现。代码、数据和评估脚本可在 https://github.com/coenwerem/parcelstow 获取。

View on arXiv Download PDF AI Translation

cs.RO / 37 / 2609.01518

A System for Fast, Resilient, and Adaptable Loco-Manipulation Behaviors on Humanoid Robots

一种快速、弹性和适应性强的人形机器人运动操控行为系统

Calvert, Duncan, Penco, Luigi, Anderson, Dexton, Bialek, Tomasz, Chatterjee, Arghya, Park, Beomyeong, Griffin, Robert

Abstract

There is tremendous value in humanoid robots taking on physically demanding, hazardous, and repetitive work in spaces built for humans. However, a useful robot for these spaces must coordinate locomotion, whole-body motion, perception, contact, and operator supervision. We present a robot-local, runtime-editable behavior authoring and runtime system that addresses these challenges. We argue that behavior architecture can be a primary enabler of capability, speed, and reliability, and that runtime editability enables fast behavior creation, adaptation, extension, and combination. Our behavior architecture combines object-centric Affordance Templates, a tree structure that provides organization and logic, and runtime-editable perception through a behavior scene and primitive scene actions. Our operator interface remains continuously synchronized to the robot for runtime authoring, monitoring, and repair. Action primitives execute through a whole-body controller that supports concurrent body motions and walking. Demonstrations of our system cover six task variants on Unitree H1-2 and Alex. We execute a push door traversal in 34 seconds and sort six balls by color in 45 seconds under human disturbance. Timed authoring sessions show scratch creation of new loco-manipulation behaviors and adaptation of existing ones in hours. Comparison against the literature finds our approach to be competitive with recent learned systems.

Chinese Translation

人形机器人在为人类建造的空间中承担体力要求高、危险和重复的工作具有巨大的价值。然而，适用于这些空间的机器人必须协调运动、全身动作、感知、接触和操作员监督。我们提出了一种机器人本地的、可在运行时编辑的行为创作和运行系统，以应对这些挑战。我们认为，行为架构可以成为能力、速度和可靠性的主要推动因素，而运行时可编辑性则使快速行为创建、适应、扩展和组合成为可能。我们的行为架构结合了以对象为中心的可供性模板（Affordance Templates）、提供组织和逻辑的树状结构，以及通过行为场景和基本场景动作实现的运行时可编辑感知。我们的操作员界面与机器人保持持续同步，以便进行运行时创作、监控和修复。动作原语通过一个支持并发身体动作和行走的全身控制器执行。我们系统的演示涵盖了Unitree H1-2和Alex上的六种任务变体。在人类干扰下，我们在34秒内完成了推门穿越，并在45秒内按颜色对六个球进行分类。定时创作会显示出在数小时内快速创建新的运动操控行为和适应现有行为的能力。与文献的比较发现，我们的方法在与最近的学习系统的竞争中表现良好。

View on arXiv Download PDF AI Translation

cs.RO / 38 / 2609.01579

SG-AMP: Scene-Graph-Guided Active Perception and Semantics-Aware Motion Planning for Pepper Plants

SG-AMP：基于场景图的主动感知与语义感知运动规划用于辣椒植物

Menon, Rohit, Lolla, Shiva Rudra, Mueller-Goldingen, Niklas, Chenchani, Gokul, Roscher, Ribana, Bennewitz, Maren

Abstract

We present SG-AMP, integrating robust depth completion with input-conditioned uncertainty, persistent panoptic mapping, plant scene-graph reasoning, and semantics-aware active view-motion planning. Beyond inspecting uncertain observed regions, the scene graph explicitly hypothesizes unobserved pepper--peduncle attachments and directs close-range sensing toward them. Candidate views are selected according to expected information gain, while class-dependent motion costs distinguish protected peppers, peduncles, and stems from conditionally traversable foliage. On pepper data, the perception network achieves $55.27\%$ semantic mIoU, $38.67\%$ PQ, and $40.62\,\mathrm{mm}$ depth RMSE, while input-conditioned uncertainty improves NYUv2 NLL from $-1.6518$ to $-1.6925$ and AUSE from $0.0102$ to $0.0087$.

Chinese Translation

我们提出了SG-AMP，整合了稳健的深度补全与输入条件的不确定性、持久的全景映射、植物场景图推理以及语义感知的主动视角运动规划。除了检查不确定的观察区域外，场景图明确假设未观察到的辣椒-果梗连接，并将近距离传感引导至这些区域。候选视角根据预期信息增益进行选择，而依赖于类别的运动成本则将受保护的辣椒、果梗和茎与条件可穿越的叶片区分开。在辣椒数据上，感知网络实现了$55.27\%$的语义mIoU，$38.67\\%$的PQ，以及$40.62\, ext{mm}$的深度均方根误差（RMSE），同时输入条件的不确定性将NYUv2的负对数似然（NLL）从$-1.6518$提升至$-1.6925$，AUSE从$0.0102$降低至$0.0087$。

View on arXiv Download PDF AI Translation

cs.RO / 39 / 2609.01596

Facet-0: A Robotic Foundation Model for Contact-Rich Precise Manipulation

Facet-0：一种用于接触丰富精确操作的机器人基础模型

Deng, Haoyuan, Liu, Haichao, Guo, Wenkai, Ling, Yuan, Yang, Zaijia, Xue, Yuanjiang, Sun, Haosheng, Wang, Liangzi, Wang, Ziwei

Abstract

Real-world robotic assembly at sub-millimeter tolerances demands spatial precision, compliant interaction, and robustness to contact failures. We present Facet-0, a robotic foundation model that predicts and values the contact consequences of its actions. Facet-0 unifies multimodal representation learning and reinforcement learning (RL) post-training around a joint action-wrench proposal: a causal wrench history is aligned with vision-language semantics and kinematic state, and flow matching generates each action chunk together with the future wrist-wrench profile it is expected to induce. Deployment rollouts train a distributional Action-Wrench Critic to distinguish motions with similar task progress but different contact outcomes, while phase-aware rewards and contact-selective credit concentrate policy improvement on decisive interactions. To accommodate part-specific dynamics, a lightweight bounded actor reuses the frozen representation for on-robot adaptation; RL remains defined over executable Cartesian actions, while an auxiliary wrench head preserves predictive, non-commanded action-contact coupling. Trained on ManuFacet-1K, a 1,000-hour force-synchronized corpus spanning three embodiments and multiple manufacturing cells, the bounded task-adapted system reaches 82% mean success on five sub-millimeter computer-assembly tasks, compared with 15% for the strongest baseline, with 0.5 mm placement accuracy and 50 ms command latency.

Chinese Translation

现实世界中的机器人组装在亚毫米公差下需要空间精度、顺应性互动和对接触失败的鲁棒性。我们提出了Facet-0，这是一种预测和评估其动作接触后果的机器人基础模型。Facet-0将多模态表示学习与强化学习（RL）后训练统一在一个联合动作-扭矩提议上：因果扭矩历史与视觉-语言语义和运动学状态对齐，流匹配生成每个动作块以及预期引发的未来腕部扭矩特征。部署回滚训练一个分布式的动作-扭矩评论员，以区分具有相似任务进展但不同接触结果的动作，同时阶段感知奖励和接触选择性信用将策略改进集中在决定性互动上。为了适应特定部件的动态，轻量级有界演员重用冻结表示进行机器人适应；强化学习仍然定义在可执行的笛卡尔动作上，而辅助扭矩头则保持预测的、非指令的动作-接触耦合。在ManuFacet-1K上进行训练，这是一个跨越三种实施和多个制造单元的1,000小时力同步语料库，该有界任务适应系统在五个亚毫米计算机组装任务上达到82%的平均成功率，而最强基线仅为15%，具有0.5毫米的放置精度和50毫秒的指令延迟。

View on arXiv Download PDF AI Translation

计算机视觉 (Computer Vision)

116

cs.CV / 1 / 2609.00036

A Cone-Constrained Bilinear Decomposition for Total Scaled-Gradient Variation Models

一种锥约束双线性分解的全尺度梯度变差模型

Su, Haibin, Wu, Chunlin, Chang, Huibin, Liu, Zhifang

Abstract

The total scaled-gradient variation (TSGV) regularizer, derived from sparse modeling of piecewise-linear structures, has been shown to preserve edges and corners in image restoration. However, its highly nonconvex and nonlinear nature poses severe computational challenges, as existing methods often suffer from parameter sensitivity or lack convergence guarantees. To overcome this, we propose a tailored bilinear decomposition that decouples the nonlinear weighted gradient in the TSGV regularizer. This approach yields an equivalent optimization problem governed by cone or sphere constraints, depending on the chosen scaling function. In particular, the cone constraint plays a central role in characterizing edge- and corner-preserving behavior. We solve this reformulation using the alternating minimization method (AMM) equipped with a majorization--minimization strategy, ensuring a monotonic decrease in energy without step-size tuning. Furthermore, we provide a geometric interpretation of the edge-preserving properties of these constraints by analyzing their asymptotic behavior near image singularities. We establish the global convergence of the proposed method to a critical point within the Kurdyka--{\L}ojasiewicz framework. Extensive numerical experiments on Gaussian denoising and non-line-of-sight (NLOS) imaging show that the proposed method achieves PSNR and SSIM competitive with or superior to representative variational methods, especially at high noise levels, and improves the structural reconstruction under dense and sparse scanning.

Chinese Translation

全尺度梯度变差（TSGV）正则化器源于对分段线性结构的稀疏建模，已被证明在图像恢复中能够有效保留边缘和角点。然而，其高度非凸和非线性的特性带来了严重的计算挑战，因为现有方法往往存在参数敏感性或缺乏收敛保证。为此，我们提出了一种定制的双线性分解方法，解耦了TSGV正则化器中的非线性加权梯度。该方法产生了一个等效的优化问题，其受锥或球体约束控制，具体取决于所选择的缩放函数。特别地，锥约束在表征边缘和角点保留行为中起着核心作用。我们使用交替最小化方法（AMM）结合主化-最小化策略来解决这一重构问题，确保能量单调下降而无需调整步长。此外，我们通过分析这些约束在图像奇异点附近的渐近行为，提供了对其边缘保留特性的几何解释。我们在Kurdyka-{ ext{L}}ojasiewicz框架下建立了所提方法对临界点的全局收敛性。在高斯去噪和非直视（NLOS）成像的广泛数值实验中，结果表明所提方法在峰值信噪比（PSNR）和结构相似性指数（SSIM）上与代表性的变分方法具有竞争力，尤其在高噪声水平下表现优越，并改善了在稠密和稀疏扫描下的结构重建。

View on arXiv Download PDF AI Translation

cs.CV / 2 / 2609.00111

Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving

Qwen-Drive-1.0：迈向自动驾驶视觉-语言基础模型的初步步骤

Zhou, Xin, Zhao, Zongchuang, Yang, Zhibo, Li, Mingsheng, Zhong, Humen, Bai, Shuai, Chu, Du, Chen, Ruizhe, Li, Zhaohai, Tang, Jun, Wang, Qiuyue, Yang, Mingkun, Zhang, Jiazhao, Liu, Dayiheng, Liang, Dingkang, Bai, Xiang

Abstract

We present Qwen-Drive-1.0, an initial step towards a vision-language foundation model for autonomous driving. Qwen-Drive-1.0 retains the architecture of the pretrained vision-language model (VLM) and integrates 3D perception, visual question answering, and motion planning within a unified framework. An external bird's-eye-view (BEV) perception head jointly performs 3D object detection, semantic occupancy prediction, and BEV map segmentation. It serves as a probe of the 3D information accessible from the shared representations and provides an explicit, inspectable interface to 3D scene structure. A Planning Expert conditions on shared VLM representations to generate future ego trajectories. A staged training recipe combines driving supervision with general-purpose vision-language data to acquire driving-specific competence while helping preserve broad visual understanding and instruction-following capabilities. Experiments demonstrate strong 3D perception and driving scene understanding while largely preserving general vision-language capability. Comprehensive evaluations across open-loop, pseudo-closed-loop, and closed-loop settings further show highly competitive motion-planning performance.

Chinese Translation

我们提出了Qwen-Drive-1.0，这是迈向自动驾驶视觉-语言基础模型的初步步骤。Qwen-Drive-1.0保留了预训练视觉-语言模型（VLM）的架构，并在统一框架内整合了3D感知、视觉问答和运动规划。一个外部鸟瞰视图（BEV）感知头共同执行3D物体检测、语义占用预测和BEV地图分割。它作为从共享表示中获取的3D信息的探针，提供了一个明确的、可检查的3D场景结构接口。规划专家基于共享的VLM表示生成未来的自我轨迹。分阶段的训练方案结合了驾驶监督与通用视觉-语言数据，以获取特定于驾驶的能力，同时帮助保持广泛的视觉理解和遵循指令的能力。实验表明，强大的3D感知和驾驶场景理解能力得以实现，同时在很大程度上保留了通用视觉-语言能力。在开放环路、伪闭环和闭环设置下的全面评估进一步显示出高度竞争的运动规划性能。

View on arXiv Download PDF AI Translation

cs.CV / 3 / 2609.00188

ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training

ZimaBlue：通过可扩展视频预训练演化通用世界动作模型

Wu, Xionghao, Yang, Yijun, Zhou, Shiyang, Sun, Haoze, Liu, Jianhui, Yu, Songsong, Zhang, Jiyao, Li, Wenbo, Wang, Bo, Ma, Guoqing, Song, Lin, Liao, Renjie, Zheng, Shenghe, Tang, Wei, Qi, Xiaojuan, Li, Yanwei, Zhang, Yuan, Tian, Zhuotao, Huang, Haoyang, Duan, Nan

Abstract

Robotic manipulation faces a fundamental scaling challenge: robust generalization demands broad physical experience, yet action-labeled robot trajectories are expensive to collect and inherently limited in diversity. Egocentric videos offer a far more scalable source of embodied experience, capturing object interactions, contact dynamics, tool use, and long-horizon behaviors across diverse environments. The central challenge is how to convert this abundant but action-free experience into effective robot control. We introduce ZimaBlue, a scalable framework for learning generalizable World Action Models (WAMs) from large-scale video. ZimaBlue follows a three-stage training curriculum: it first performs causal embodied video pre-training on large-scale human and robot egocentric videos, then grounds the learned visual dynamics in heterogeneous robot trajectories through video-action mid-training with a unified action representation, and finally specializes the model to a target robot for deployment. To make generative WAMs practical for real-time control, ZimaBluefurther adopts an asynchronous Slow-Fast dual-system architecture, where a high-capacity Slow world model provides generalizable spatiotemporal representations and a lightweight Fast branch enables 30 Hz action prediction on NVIDIA RTX 4090. On real-robot zero-shot evaluations, scaling from target-robot data alone to over 120,000 hours of embodied video improves success from 36.1% to 77.8%. ZimaBlue further delivers strong performance across multiple benchmarks, with particularly pronounced gains on unseen tasks.

Chinese Translation

机器人操作面临一个基本的扩展挑战：强大的泛化能力需要广泛的物理经验，而带有动作标签的机器人轨迹收集成本高且本质上在多样性上有限。自我中心视频提供了一个更具可扩展性的具身经验来源，捕捉了对象交互、接触动态、工具使用和在多样环境中的长时间行为。核心挑战在于如何将这种丰富但不带动作的经验转化为有效的机器人控制。我们提出了ZimaBlue，一个从大规模视频中学习通用世界动作模型（WAMs）的可扩展框架。ZimaBlue遵循三阶段的训练课程：首先在大规模人类和机器人自我中心视频上进行因果具身视频预训练，然后通过统一的动作表示在异构机器人轨迹中将学习到的视觉动态进行视频-动作中期训练，最后将模型专门化以适应目标机器人进行部署。为了使生成的WAMs在实时控制中实用，ZimaBlue进一步采用了异步的Slow-Fast双系统架构，其中高容量的Slow世界模型提供可泛化的时空表示，而轻量级的Fast分支则在NVIDIA RTX 4090上实现30 Hz的动作预测。在真实机器人零-shot评估中，仅从目标机器人数据扩展到超过120,000小时的具身视频，成功率从36.1%提高到77.8%。ZimaBlue在多个基准测试中进一步展现出强大的性能，尤其是在未见任务上表现出显著的提升。

View on arXiv Download PDF AI Translation

cs.CV / 4 / 2609.00198

A Lagrangian View of Flow Matching

流匹配的拉格朗日视角

Milanfar, Peyman

Abstract

Modern explicit-time generative models, such as Flow Matching [Lipman et al., 2023] and Rectified Flow [Liu et al., 2023], are typically derived top-down via Optimal Transport and the continuity equation. This standard Eulerian approach focuses on the macroscopic transport of probability mass. In this paper, we present an alternative, bottom-up mechanical derivation grounded in a Lagrangian (particle-centric) perspective. By analyzing the local Taylor expansion of a continuous denoiser, we motivate a strict invariance condition required for optimal, singlestep generation: the conservation of target identity. Enforcing this condition yields a governing quasi-linear advection Partial Differential Equation (PDE). We demonstrate that solving this PDE via the Method of Characteristics analytically yields the straight-line trajectories of Flow Matching. This geometric perspective isolates the Jacobian of the denoiser as the primary source of trajectory curvature, providing a direct mathematical explanation for why straight-line flows enable massive step sizes, and why empirical models require distillation to flatten intersecting characteristics.

Chinese Translation

现代显式时间生成模型，如流匹配（Flow Matching）[Lipman et al., 2023]和修正流（Rectified Flow）[Liu et al., 2023]，通常通过最优传输和连续性方程自上而下推导而来。这种标准的欧拉方法侧重于概率质量的宏观传输。本文提出了一种基于拉格朗日（粒子中心）视角的替代自下而上的机械推导。通过分析连续去噪器的局部泰勒展开，我们激励了一个严格的不变性条件，该条件对于最优单步生成是必需的：目标身份的守恒。强制执行这一条件产生了一个控制的准线性对流偏微分方程（PDE）。我们证明，通过特征方法解析地求解该PDE可以得到流匹配的直线轨迹。这一几何视角将去噪器的雅可比矩阵孤立为轨迹曲率的主要来源，直接数学解释了为什么直线流动能够实现大步长，以及为什么经验模型需要蒸馏以平坦化交叉特征。

View on arXiv Download PDF AI Translation

cs.CV / 5 / 2609.00206

Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM-Based Video Moderation

分布式隐性危害：基于多模态大型语言模型的视频审核中的组合安全盲点

Wang, Ruotong, Zhu, Zihao, Lyu, Siwei, Tao, Xin, Wu, Baoyuan

Abstract

Despite their growing use in video moderation, multimodal large language models (MLLMs) exhibit a compositional safety blind spot: videos composed of seemingly benign components can convey harmful meaning when interpreted as a whole. We refer to this phenomenon as Distributed Implicit Harm (DIH), where harm arises from relations among components distributed along a decomposition axis of the video, rather than from any single explicit cue. Among many possible axes, we study two representative cases: temporally distributed harm across visual segments (DIH-T) and cross-modal harm between audio and visual streams (DIH-M). Studying and mitigating DIH at scale requires data that is difficult to collect: such videos lack compositional harm annotations, evade retrieval based on local visual cues, keywords, or single-modality signals, and are consequently absent from existing safety datasets. To bridge this gap, we develop a multi-agent synthesis framework that composes individually benign components into harmful scenarios and generates diverse DIH videos with explicit reasoning annotations, yielding a dataset of over 9,000 videos spanning visual-only and audio-visual settings. Benchmarking over 30 MLLMs spanning frontier proprietary models and leading open-source systems reveals substantial and consistent deficits in detecting both DIH-T and DIH-M. Notably, this failure persists even among the strongest frontier models: they often correctly assess individual components in isolation but fail to recognize the harmful meaning that emerges from their composition. We further evaluate these models on a manually collected set of real-world DIH videos from social media and observe the same failure mode, highlighting DIH as a practical and underexplored challenge for video moderation.

Chinese Translation

尽管多模态大型语言模型（MLLMs）在视频审核中的使用日益增长，但它们表现出一种组合安全盲点：由看似无害的组件组成的视频在整体解读时可能传达有害的意义。我们将这种现象称为分布式隐性危害（Distributed Implicit Harm, DIH），其危害源于沿视频分解轴分布的组件之间的关系，而非来自任何单一的显性线索。在众多可能的分解轴中，我们研究了两个代表性案例：视觉片段之间的时间分布危害（DIH-T）和音频与视觉流之间的跨模态危害（DIH-M）。在大规模研究和缓解DIH时，需要收集难度较大的数据：此类视频缺乏组合危害的注释，无法基于局部视觉线索、关键词或单一模态信号进行检索，因此在现有的安全数据集中缺失。为了解决这一问题，我们开发了一个多智能体合成框架，将单独无害的组件组合成有害场景，并生成具有明确推理注释的多样化DIH视频，形成一个包含超过9000个视频的数据集，涵盖视觉单一和音频-视觉设置。对30多种MLLM进行基准测试，包括前沿专有模型和领先的开源系统，发现它们在检测DIH-T和DIH-M方面存在显著且一致的不足。值得注意的是，这种失败在最强的前沿模型中依然存在：它们通常能够正确评估单个组件，但未能识别出其组合所产生的有害意义。我们进一步在社交媒体上手动收集了一组真实世界的DIH视频，对这些模型进行评估，观察到相同的失败模式，突显DIH作为视频审核中的一个实际且未被充分探索的挑战。

View on arXiv Download PDF AI Translation

cs.CV / 6 / 2609.00231

Beyond Language Priors: Diagnosing and Fixing Visual-Origin Hallucinations in Multimodal LLM

超越语言先验：诊断与修复多模态大语言模型中的视觉来源幻觉

Xu, Peiyang, Zhu, Xiaopei, Zhu, Jun, Hu, Xiaolin

Abstract

Existing research on object hallucination in multimodal large language models (MLLMs) predominantly attributes the problem to language priors such as over-reliance on textual co-occurrence statistics. We challenge this view by presenting quantitative evidence for a complementary, under-explored cause: visual-origin hallucination, where hallucinations arise from incorrect visual feature extraction and misalignment between image and text embeddings. Through cosine similarity analysis and Smooth Grad-CAM entropy measurements, we show that hallucinated samples exhibit systematically lower image-text similarity (average 0.158 vs. -0.122) and inverted attention patterns, where attention is dispersed when the target object is present but wrongly concentrated when it is absent. Guided by this diagnosis, we propose Adversarial Contrastive Fine-Tuning (ACFT). ACFT uses an Adversarial Hallucination Attribute Flipping (AHAF) procedure, involving minimal, targeted adversarial perturbations that flip an image's hallucination attribute, to construct perfectly aligned positive-negative pairs, which are then used for contrastive fine-tuning. AHAF simultaneously serves as a diagnostic probe, revealing that MLLM visual representations lie dangerously close to hallucination decision boundaries. Requiring only 0.9% of the COCO dataset and adding zero inference overhead, ACFT achieves state-of-the-art performance on POPE, MME, and four description-level hallucination benchmarks across LLaVA, MiniGPT-4, and Qwen2.5-VL. Code is available at https://github.com/zxp555/ACFT_MM

Chinese Translation

现有关于多模态大语言模型（MLLM）中物体幻觉的研究主要将问题归因于语言先验，例如对文本共现统计的过度依赖。我们通过提供定量证据挑战这一观点，提出一种补充的、尚未充分探索的原因：视觉来源幻觉，即幻觉源于不正确的视觉特征提取以及图像与文本嵌入之间的错位。通过余弦相似度分析和光滑梯度加权类激活映射（Smooth Grad-CAM）熵测量，我们显示出幻觉样本表现出系统性较低的图像-文本相似度（平均值0.158对比-0.122）和反转的注意力模式，即在目标物体存在时注意力分散，而在目标物体缺失时注意力却错误集中。基于这一诊断，我们提出了对抗性对比微调（Adversarial Contrastive Fine-Tuning, ACFT）。ACFT使用对抗性幻觉属性翻转（Adversarial Hallucination Attribute Flipping, AHAF）过程，涉及最小的、有针对性的对抗扰动，以翻转图像的幻觉属性，从而构建完美对齐的正负样本对，随后用于对比微调。AHAF同时作为诊断探针，揭示了MLLM的视觉表示与幻觉决策边界存在危险接近。ACFT仅需0.9%的COCO数据集，并且没有增加推理开销，在LLaVA、MiniGPT-4和Qwen2.5-VL上实现了POPE、MME和四个描述级幻觉基准的最新性能。代码可在 https://github.com/zxp555/ACFT_MM 获取。

View on arXiv Download PDF AI Translation

cs.CV / 7 / 2609.00232

Beyond Blind Compliance: Benchmarking Task Verification in OCR Reasoning

超越盲目遵从：OCR推理中的任务验证基准测试

Zhou, Yue, Wu, Yuan, Chang, Yi

Abstract

Multimodal Large Language Models (MLLMs) have achieved strong performance on OCR-centric document understanding and text-rich visual reasoning benchmarks. Yet existing evaluations largely assume that every task is valid and answerable. In real-world OCR scenarios, this assumption often fails: questions may rely on illegible text, occluded evidence, nonexistent visual targets, contradictory premises, or missing variables. We study this reliability gap as OCR-grounded Task Verification: before answering, a model should determine whether the Image Premise (IP), Textual Premise (TP), and Question (Q) jointly define an executable task. We introduce VeriOCRBench, a 1,800-sample human-verified benchmark built from source images drawn from 8 OCR-related datasets and spanning 8 real-world image domains, with controlled, image-grounded diagnostic tasks. It contains 1,600 trap-injected invalid tasks across 8 trap types and four verification dimensions---Visual, Contextual, Factual, and Logical---plus 200 trap-free controls for measuring over-refusal. Built with a Visual Atomic Fact (VAF)-anchored pipeline and full human auditing, VeriOCRBench enables decoupled evaluation of task verification, root-cause diagnosis, and over-refusal. Evaluating 15 leading MLLMs reveals persistent blind compliance, diagnosis failures, and prompt-induced over-refusal, exposing a critical reliability gap in current OCR reasoning systems. The code is available at: https://github.com/zy001122/Beyond-Blind-Compliance.

Chinese Translation

多模态大型语言模型（MLLMs）在以OCR为中心的文档理解和文本丰富的视觉推理基准测试中表现出色。然而，现有的评估大多假设每个任务都是有效且可回答的。在现实世界的OCR场景中，这一假设往往不成立：问题可能依赖于难以辨认的文本、遮挡的证据、不存在的视觉目标、矛盾的前提或缺失的变量。我们将这种可靠性差距研究为基于OCR的任务验证：在回答之前，模型应确定图像前提（Image Premise, IP）、文本前提（Textual Premise, TP）和问题（Question, Q）是否共同定义了一个可执行的任务。我们引入了VeriOCRBench，这是一个包含1,800个样本的人类验证基准，基于来自8个与OCR相关的数据集的源图像，涵盖8个现实世界的图像领域，并设有受控的、图像基础的诊断任务。它包含1,600个注入陷阱的无效任务，分为8种陷阱类型和四个验证维度——视觉（Visual）、上下文（Contextual）、事实（Factual）和逻辑（Logical），以及200个无陷阱的控制任务用于测量过度拒绝。VeriOCRBench采用以视觉原子事实（Visual Atomic Fact, VAF）为基础的管道和全面的人类审计，能够实现任务验证、根本原因诊断和过度拒绝的解耦评估。对15个领先的MLLMs的评估揭示了持续的盲目遵从、诊断失败和提示引发的过度拒绝，暴露了当前OCR推理系统中的关键可靠性差距。代码可在以下链接获取：https://github.com/zy001122/Beyond-Blind-Compliance。

View on arXiv Download PDF AI Translation

cs.CV / 8 / 2609.00242

CoLT-Drive: Counterfactual Long-Tail Benchmarking and Knowledge-Preserving Adaptation for Driving Affordance Prediction

CoLT-Drive：反事实长尾基准测试与知识保留适应用于驾驶能力预测

Tang, Zhengxu, Cui, Guofeng, Gong, Ziyu, Zhang, Xiaozhou, Deng, Ruifeng, Qi, Chengzhi, Chen, Ke, Patil, Sachin, Xiao, Tianjun, Liu, Langechuan, Wang, Pichao

Abstract

Long-tail autonomous driving failures are often framed as rare-object recognition errors. We argue that this view is incomplete: the decision-critical question is not only whether a model recognizes an unusual object, but whether it infers how that object changes the ego vehicle's feasible high-level actions. We formalize this problem as decision-level driving affordance prediction, where a model maps a front-view image, ego-motion history, and navigation command to a structured longitudinal--lateral meta-action. To evaluate this capability, we introduce CoLT-Drive, a 3,536-sample counterfactual long-tail benchmark that inserts rare objects into otherwise fixed driving scenes and measures whether models predict acceptable action pairs. To improve deployable small VLMs, we propose KPA, a knowledge-preserving adaptation framework that combines structured perception-to-decision prompting, SLERP-based expert merging, and RegMoE, a regime-aware LoRA mixture-of-experts module. KPA preserves the pretrained model's open-world knowledge while allocating lightweight adaptation capacity to different driving decision regimes. Experiments on an in-domain driving split and CoLT-Drive show that KPA achieves 60.8\% pair accuracy on CoLT-Drive, outperforming the pretrained Qwen3-VL-2B baseline (50.3\%) and LoRA SFT (32.4\%) while maintaining competitive in-domain accuracy. Our benchmark and code are available at https://huggingface.co/datasets/tangzx2024/CoLT-Drive and https://github.com/tangzhengxu/CoLT-Drive.

Chinese Translation

长尾自主驾驶失败通常被视为稀有物体识别错误。我们认为这种观点是不完整的：决策关键问题不仅在于模型是否识别出不寻常的物体，还在于它是否推断出该物体如何改变自我车辆的可行高层次动作。我们将这个问题形式化为决策层面的驾驶能力预测，其中模型将前视图像、自我运动历史和导航命令映射到结构化的纵向-横向元动作。为了评估这一能力，我们引入了CoLT-Drive，一个包含3,536个样本的反事实长尾基准，向固定驾驶场景中插入稀有物体，并测量模型是否预测出可接受的动作对。为了改善可部署的小型VLM（视觉语言模型），我们提出了KPA（知识保留适应框架），该框架结合了结构化感知到决策的提示、基于SLERP的专家合并和RegMoE（基于模式的LoRA专家混合模块）。KPA在分配轻量级适应能力到不同驾驶决策模式的同时，保留了预训练模型的开放世界知识。在一个领域内的驾驶分割和CoLT-Drive上的实验表明，KPA在CoLT-Drive上达到了60.8%的配对准确率，超越了预训练的Qwen3-VL-2B基线（50.3%）和LoRA SFT（32.4%），同时保持了竞争性的领域内准确率。我们的基准和代码可在https://huggingface.co/datasets/tangzx2024/CoLT-Drive和https://github.com/tangzhengxu/CoLT-Drive获取。

View on arXiv Download PDF AI Translation

cs.CV / 9 / 2609.00272

CrossFeat: Bridging Imaging Modalities in Feature Descriptor Space

CrossFeat：在特征描述符空间中桥接成像模态

Schneider, Paul, Haouchine, Nazim

Abstract

Most advances in keypoint descriptions address monomodal settings, where image variations arise from viewpoint, illumination, or contrast changes. Multimodal scenarios involve images produced by fundamentally different sensing processes, such as multispectral imaging, RGB-depth, satellite imagery, or medical imaging, causing the same structures to appear differently. A common solution to cross-modal description is to train descriptors for each modality pair, which requires retraining whenever the modalities change, or to train large models, which incur a significant increase in runtime. Instead, we propose CrossFeat, a framework that enables an existing monomodal descriptor to operate across modalities. Our method learns a crossing function in descriptor space that maps features from one modality to a representation compatible with another. To preserve the structural information captured by the original descriptor, CrossFeat introduces a geometry-appearance disentanglement such that only appearance is altered while the geometric properties are preserved. Experiments across multiple domains and datasets demonstrate improved performance in multimodal matching.

Chinese Translation

大多数关键点描述的进展集中在单模态设置上，其中图像变化源于视角、光照或对比度的变化。多模态场景涉及由根本不同的传感过程产生的图像，例如多光谱成像、RGB-深度、卫星图像或医学成像，导致相同结构的表现方式不同。跨模态描述的常见解决方案是为每对模态训练描述符，这需要在模态变化时进行重新训练，或者训练大型模型，这会显著增加运行时间。相反，我们提出了CrossFeat，一个使现有单模态描述符能够跨模态操作的框架。我们的方法在描述符空间中学习一个交叉函数，将一种模态的特征映射到与另一种模态兼容的表示。为了保留原始描述符捕获的结构信息，CrossFeat引入了一种几何-外观解耦，使得只有外观被改变，而几何属性得以保留。跨多个领域和数据集的实验表明，在多模态匹配中性能得到了改善。

View on arXiv Download PDF AI Translation

cs.CV / 10 / 2609.00291

StreamScout: Learning When to Look Deeper for Streaming Video Understanding

StreamScout：学习在流媒体视频理解中何时深入探索

Zhang, Ce, Bi, Jing, He, Jinxi, Zhang, Jianshu, Lin, Jingyang, Xiao, Yunzhong, Fu, Minghao, Xie, Yaqi, Xie, Zhentao, Chen, Weicong, Sycara, Katia, Zhou, Ming

Abstract

Streaming video understanding requires answering questions that arrive at arbitrary moments over an unbounded video stream. Existing systems primarily focus on what to retain in a bounded memory, yet access that memory using the same fixed-cost procedure for every query, despite substantial variation in the evidence required. We argue that deciding how deeply to access memory for each query is as important as deciding what the memory should store. To this end, we introduce StreamScout, an adaptive inference framework that maintains only a lightweight textual timeline in context as the stream unfolds. At query time, StreamScout progressively augments the timeline with up to three increasingly informative visual views: a glance at recent frames, a uniform look-back over the past stream, and query-salient retrieval. At each stage, the model answers immediately if the available evidence is sufficient; otherwise, it escalates to the next view. To improve this stop-or-escalate policy, we probe the cascade on an auxiliary set and distill the model's empirical competence boundary into supervision for a lightweight LoRA adaptation, yielding StreamScout-S. We further refine the policy through reinforcement learning, allowing the model to explore stopping behaviors beyond imitation of the distilled decisions, yielding StreamScout-R. Across three backbones and three streaming benchmarks, StreamScout and its variants consistently outperform prior streaming methods while substantially reducing inference cost and token consumption; on OVO-Bench, for instance, StreamScout-S improves Qwen3-VL-8B by 14.65 points while using 59% fewer tokens than uniform sampling and answering in 1.04 s on average.

Chinese Translation

流媒体视频理解需要在无限的视频流中回答在任意时刻出现的问题。现有系统主要关注在有限内存中保留什么，但对于每个查询使用相同的固定成本程序访问该内存，尽管所需证据存在显著差异。我们认为，决定对每个查询访问内存的深度与决定内存应存储什么同样重要。为此，我们提出了StreamScout，一个自适应推理框架，在流媒体展开时仅维护一个轻量级的文本时间线。在查询时，StreamScout逐步用最多三种越来越具信息量的视觉视图增强时间线：查看最近的帧、对过去流媒体的均匀回顾，以及与查询相关的检索。在每个阶段，如果可用证据足够，模型会立即给出答案；否则，它会升级到下一个视图。为了改进这一停止或升级策略，我们在辅助集上探测级联，并将模型的经验能力边界提炼为轻量级LoRA适应的监督，从而产生StreamScout-S。我们进一步通过强化学习优化该策略，使模型能够探索超越提炼决策的停止行为，从而产生StreamScout-R。在三个基础模型和三个流媒体基准上，StreamScout及其变体始终优于先前的流媒体方法，同时显著降低推理成本和令牌消耗；例如，在OVO-Bench上，StreamScout-S使Qwen3-VL-8B的得分提高了14.65分，同时使用的令牌比均匀采样少59%，平均回答时间为1.04秒。

View on arXiv Download PDF AI Translation

cs.CV / 11 / 2609.00300

TRUST: Threshold-Recalibrated Uncertainty-Safe Training for Certified Dismissal in Breast Cancer Screening

TRUST：阈值重新校准的不确定性安全训练用于乳腺癌筛查中的认证拒绝

Hajishafiezahramini, Parham, Hamilton, Matthew, Kendall, Edward, Doyle, Gregory, Pastor, Oscar Meruvia

Abstract

Reducing the review of clearly cancer-negative screening mammograms could lower radiologist workload without compromising cancer detection. We propose a closed-loop threshold-aware training strategy in which the dismissal threshold is recalculated during training and used to penalize cancer-positive images that approach the dismissal region. We evaluated the method on NLBS and RSNA using five controlled training configurations, with case-level assessment based on a one-sided 99\% Clopper--Pearson upper bound for cancer prevalence among dismissed cases. The proposed model achieved the highest case-level dismissal rates at both 98\% and 95\% recall targets. On NLBS, dismissal reached 19.74\% and 21.70\%, while the cross-entropy baseline did not meet either recall target. On RSNA, dismissal improved from 7.04\% to 14.31\% and from 13.49\% to 19.69\%. In external RSNA$\to$NLBS evaluation, the proposed model achieved dismissal rates of 12.95\% and 19.87\% at the 98\% and 95\% recall targets, respectively. These results support closed-loop threshold-aware training for high-recall selective dismissal.

Chinese Translation

减少对明显癌症阴性筛查乳腺X光片的复审可以降低放射科医生的工作负担，而不影响癌症检测。我们提出了一种闭环阈值感知训练策略，在该策略中，拒绝阈值在训练过程中重新计算，并用于惩罚接近拒绝区域的癌症阳性图像。我们在NLBS和RSNA上评估了该方法，使用五种受控训练配置进行案例级评估，基于被拒绝案例中癌症流行率的一侧99% Clopper--Pearson上限。所提出的模型在98%和95%的召回目标下实现了最高的案例级拒绝率。在NLBS上，拒绝率达到了19.74%和21.70%，而交叉熵基线未达到任何召回目标。在RSNA上，拒绝率从7.04%提高到14.31%，从13.49%提高到19.69%。在外部RSNA$ o$NLBS评估中，所提出的模型在98%和95%召回目标下分别达到了12.95%和19.87%的拒绝率。这些结果支持高召回选择性拒绝的闭环阈值感知训练。

View on arXiv Download PDF AI Translation

cs.CV / 12 / 2609.00358

Where Should Experience Live? Hierarchical Hebbian Memory for Continual Vision Transformers

经验应存于何处？用于持续视觉变换器的层次海贝尔记忆

Mujawar, Mohammed Yusuf, Golilarz, Noorbakhsh Amiri

Abstract

Vision Transformers provide strong visual representations but typically rely on slowly updated parameters, limiting their ability to organize newly acquired information across different memory timescales. This work proposes \textit{Hierarchical Hebbian Memory}, a three-level memory architecture composed of rapid Working Memory, persistent Routed Episodic Memory, and slower Semantic Memory. A learned controller regulates memory contribution, read and write routing, plasticity, retention, and consolidation. A causal read-before-write lifecycle ensures that the current outcome cannot influence the prediction it supervises. The architecture is evaluated on Omniglot 5-way 1-shot recognition and CORe50 continual object recognition. With Swin-Tiny, the hierarchical model reaches 97.39\% accuracy on Omniglot and 95.37\% final accuracy on CORe50 when combined with experience replay. Learned multi-bank retrieval reaches 47.50\% delayed-association accuracy, compared with 24.17\% for a single persistent bank and 25.00\% without memory. After intervening distractors, Episodic Memory retains approximately 0.96 cosine similarity with stored associations, while Working Memory falls to approximately 0.05. These results show that Hebbian association and learned memory routing can jointly organize online visual experience across rapid, persistent, and consolidated memory timescales within Vision Transformers.

Chinese Translation

视觉变换器提供了强大的视觉表征，但通常依赖于缓慢更新的参数，这限制了它们在不同记忆时间尺度上组织新获得信息的能力。本研究提出了 extit{层次海贝尔记忆}，一种由快速的工作记忆、持久的路由情节记忆和较慢的语义记忆组成的三层记忆架构。一个学习的控制器调节记忆贡献、读写路由、可塑性、保留和巩固。因果的先读后写生命周期确保当前结果不会影响其监督的预测。该架构在Omniglot 5-way 1-shot识别和CORe50持续对象识别上进行了评估。在Swin-Tiny的情况下，层次模型在Omniglot上达到了97.39\%的准确率，在结合经验重放时，CORe50的最终准确率为95.37\\%。学习的多库检索达到了47.50\\%的延迟关联准确率，而单一持久库为24.17\\%，没有记忆时为25.00\\%。在干扰物介入后，情节记忆与存储关联的余弦相似度保持在约0.96，而工作记忆降至约0.05。这些结果表明，海贝尔关联和学习的记忆路由可以共同组织视觉变换器内快速、持久和巩固记忆时间尺度上的在线视觉经验。

View on arXiv Download PDF AI Translation

cs.CV / 13 / 2609.00369

Puppeteer: Object-Grounded Posture-Aware Co-Speech Gesture Generation

木偶师：基于对象的姿态感知共语手势生成

Adeli, Vida, Mehraban, Soroush, Rommann, Jacob, Sanborn, Harrison, Clifford, Cole, Taati, Babak

Abstract

Generating co-speech gestures that are temporally coherent, semantically aligned with speech, and grounded with surrounding objects remains challenging. Prior speech-driven gesture models emphasize audio-gesture alignment but do not explicitly account for posture constraints or surrounding objects, failing to capture the inherent correlation between body gestures and the physical space. We present Puppeteer, a posture-aware, object-grounded co-speech gesture diffusion model operating in a causal latent space. We decompose long gestures into structured primitives and learn a causal variational autoencoder that encodes them into temporally ordered latent tokens, each depending only on the past. We then perform conditional diffusion directly in the causal latent space, conditioning on speech signals, motion history, an initial posture reference, and object geometry to synthesize physically consistent gestures. This temporally ordered latent formulation enables explicit temporal control and supports tasks such as gesture in-betweening and gesture completion. To better assess co-speech gesture synthesis beyond existing measures, we introduce new evaluation metrics tailored to this task. We also created SceneGes, the first curated synthetic 3D dataset of embodied co-speech gestures and corresponding 3D objects, enabling object-grounded gesture generation. Experiments show that Puppeteer generates more diverse and temporally synchronized gestures than prior methods, while enabling object-grounded gesture synthesis.

Chinese Translation

生成在时间上连贯、语义上与语言对齐且与周围对象相结合的共语手势仍然具有挑战性。以往的语音驱动手势模型强调音频与手势的对齐，但并未明确考虑姿态约束或周围对象，未能捕捉身体手势与物理空间之间的内在关联。我们提出了木偶师（Puppeteer），一种姿态感知、基于对象的共语手势扩散模型，操作于因果潜在空间。我们将长手势分解为结构化原语，并学习一个因果变分自编码器，将其编码为时间顺序的潜在标记，每个标记仅依赖于过去。然后，我们直接在因果潜在空间中进行条件扩散，基于语音信号、运动历史、初始姿态参考和对象几何形状合成物理一致的手势。这种时间顺序的潜在形式使得显式时间控制成为可能，并支持手势插值和手势完成等任务。为了更好地评估共语手势合成，我们引入了针对该任务的新评估指标。我们还创建了SceneGes，这是第一个策划的合成3D数据集，包含具身共语手势及其对应的3D对象，从而实现基于对象的手势生成。实验表明，木偶师生成的手势在多样性和时间同步性上优于以往方法，同时实现了基于对象的手势合成。

View on arXiv Download PDF AI Translation

cs.CV / 14 / 2609.00377

FoldingAgent: Inferring Parametric Origami Procedures from Demonstration Videos

FoldingAgent：从演示视频推断参数化折纸程序

Moriya, Maya, Raab, Sigal, Vinker, Yael, Dekel, Tali

Abstract

We present FoldingAgent, an agentic framework for inferring explicit parametric folding programs directly from origami demonstration videos. Our framework leverages the reasoning power of a pre-trained Vision-Language Model (VLM) equipped with a suite of specialized tools that enable the agent to simulate geometric transitions, verify physical plausibility, retrieve and compare visual content, and evaluate its own predictions. To translate visual content into folding programs, we define a parametric space that consists of the paper's geometry and a set of parametric folding actions. Unlike models that predict static crease patterns, our agent operates sequentially and possesses the ability to re-plan its actions, effectively mitigating the compounding errors inherent in multi-step folding. Our approach takes a step toward closing the gap between human origami knowledge, which is primarily shared through unstructured visual demonstrations, and computational methods, which typically rely on structured, parametric representations such as a crease pattern or an executable parametric plan. We evaluate our approach on PurelandFold, a newly curated benchmark of diverse Pureland origami videos with ground-truth geometry and action labels. Our results demonstrate that by combining VLM reasoning with a set of specialized tools and physical simulation, we can successfully transform unstructured visual demonstrations into executable, physically plausible folding procedures.

Chinese Translation

我们提出了FoldingAgent，一个用于直接从折纸演示视频中推断显式参数化折叠程序的智能框架。我们的框架利用了一个预训练的视觉-语言模型（Vision-Language Model, VLM）的推理能力，并配备了一套专门工具，使得该智能体能够模拟几何过渡、验证物理合理性、检索和比较视觉内容，并评估自身的预测。为了将视觉内容转化为折叠程序，我们定义了一个参数空间，该空间由纸张的几何形状和一组参数化折叠动作组成。与预测静态折痕模式的模型不同，我们的智能体按顺序操作，并具备重新规划其动作的能力，有效减轻了多步骤折叠中固有的累积误差。我们的方法向缩小人类折纸知识（主要通过非结构化视觉演示共享）与计算方法（通常依赖于结构化的参数化表示，如折痕模式或可执行的参数化计划）之间的差距迈出了重要一步。我们在PurelandFold上评估了我们的方法，这是一个新近整理的多样化Pureland折纸视频基准，包含真实几何和动作标签。我们的结果表明，通过将VLM推理与一套专门工具和物理模拟相结合，我们能够成功地将非结构化视觉演示转化为可执行的、物理上合理的折叠程序。

View on arXiv Download PDF AI Translation

cs.CV / 15 / 2609.00396

SlideMix: Enhancing Whole Slide Image Analysis via Multimodal Shuffling

SlideMix：通过多模态洗牌增强全切片图像分析

Wong, Chad, Chen, Sicheng, Zhang, Tianyi, Chai, Enhui, Jin, Yueming, Liu, Zeyu, Xia, Fei

Abstract

Histopathological whole slide images (WSIs) are central to cancer diagnosis, but their gigapixel scale, tissue heterogeneity, weak slide-level supervision, sparse diagnostic regions, and multi-scale evidence make robust automated analysis challenging. Multiple instance learning (MIL) is widely used to aggregate tile-level features into slide-level predictions, yet existing augmentation strategies often perturb tissue regions without preserving diagnostic relevance, slide context, or cross-scale structure. We propose SlideMix, a model-agnostic multimodal augmentation framework for MIL-based WSI analysis. SlideMix uses a retrieval-augmented vision-language model (VLM)-based Visual-Language Adaptive Region selector to identify diagnostically relevant regions and reduce weak-label noise. It then performs In-place Tile Shuffling within meaningful tissue regions to mix feature embeddings while preserving slide-level context. A VLM-based soft-labeling module supervises mixed samples, while a multi-factor, loss-driven online Curriculum-Learning Feedback scheme adaptively controls shuffle granularity, feature similarity, and shuffle ratio to promote cross-scale representation learning. Across 11 WSI datasets comprising 20,523 slides, 8 diagnostic tasks, and 10 WSI backbones, SlideMix improves accuracy and generalization in most settings and compares favorably with established augmentation baselines, providing a simple plug-and-play approach for more robust and scalable digital pathology models. Source code: https://github.com/Xia-Research-Lab/SlideMix

Chinese Translation

组织病理学全切片图像（WSIs）在癌症诊断中至关重要，但其千兆像素级别、组织异质性、薄弱的切片级监督、稀疏的诊断区域以及多尺度证据使得稳健的自动化分析面临挑战。多实例学习（MIL）被广泛用于将瓦片级特征聚合为切片级预测，然而现有的增强策略往往扰动组织区域，而未能保持诊断相关性、切片上下文或跨尺度结构。我们提出了SlideMix，这是一种针对基于MIL的WSI分析的模型无关多模态增强框架。SlideMix使用基于检索增强的视觉-语言模型（VLM）的视觉-语言自适应区域选择器来识别诊断相关区域并减少弱标签噪声。然后，它在有意义的组织区域内执行就地瓦片洗牌，以混合特征嵌入，同时保持切片级上下文。基于VLM的软标签模块对混合样本进行监督，而多因素、基于损失的在线课程学习反馈方案自适应地控制洗牌粒度、特征相似性和洗牌比例，以促进跨尺度表示学习。在包含20,523个切片的11个WSI数据集、8个诊断任务和10个WSI骨干网络中，SlideMix在大多数设置中提高了准确性和泛化能力，并与已建立的增强基线相比表现良好，为更稳健和可扩展的数字病理模型提供了一种简单的即插即用方法。源代码：https://github.com/Xia-Research-Lab/SlideMix

View on arXiv Download PDF AI Translation

cs.CV / 16 / 2609.00411

Unmasking Face Embeddings: Reading, Rendering and Naming with Foundation Models

揭示面部嵌入：与基础模型的阅读、渲染和命名

Rubab, Fizza, Tong, Yiying, Ross, Arun

Abstract

Modern face recognition (FR) owes much of its success to deep neural networks that learn to extract compact identity embeddings from face images. These models are typically trained for identity discrimination, producing embeddings that are highly effective for biometric matching but largely opaque to semantic interpretation. In contrast, foundation models, pretrained on broad visual or vision--language tasks, provide rich interfaces for describing, retrieving, generating, and organizing visual content. This contrast raises a natural question: what capabilities become available when face embeddings from domain-specific FR models are made interoperable with foundation models? Building on recent work on embedding compatibility across models, we use simple pre-computed linear transformations, estimated from paired embeddings alone, to connect existing FR models with off-the-shelf foundation models. Once aligned with a foundation model, a face embedding can be 'unmasked' in multiple ways, without training or modifying either model: it can be read in natural language, enabling free-form text queries over a gallery of FR embeddings; rendered into a face image that recovers a person's appearance, using an unmodified diffusion decoder; and converted to a name, enabling identification even in the absence of an enrolled face gallery. In effect, one linear transformation turns an identity embedding into a rich embedding for web-scale foundation models. This interoperability exposes face embeddings as semantically and visually rich biometric representations, with direct implications for interpretability, retrieval, reconstruction, and template security.

Chinese Translation

现代人脸识别（FR）的成功在很大程度上归功于深度神经网络，这些网络学习从人脸图像中提取紧凑的身份嵌入。这些模型通常是为了身份区分而训练的，生成的嵌入在生物识别匹配中非常有效，但在语义解释上却大多不透明。相比之下，基础模型（foundation models）在广泛的视觉或视觉-语言任务上进行预训练，提供了丰富的接口，用于描述、检索、生成和组织视觉内容。这种对比引发了一个自然的问题：当领域特定的FR模型的人脸嵌入与基础模型实现互操作时，会有哪些能力可用？基于近期关于模型间嵌入兼容性的研究，我们使用简单的预计算线性变换，仅通过成对嵌入估计，来连接现有的FR模型与现成的基础模型。一旦与基础模型对齐，人脸嵌入可以以多种方式被“揭示”，而无需训练或修改任一模型：它可以用自然语言进行阅读，使得对FR嵌入库的自由文本查询成为可能；可以渲染成恢复一个人外貌的人脸图像，使用未修改的扩散解码器；并且可以转换为一个名字，即使在没有注册人脸库的情况下也能实现身份识别。实际上，一个线性变换将身份嵌入转变为适用于网络规模基础模型的丰富嵌入。这种互操作性将人脸嵌入揭示为语义和视觉上丰富的生物识别表示，对可解释性、检索、重建和模板安全性具有直接影响。

View on arXiv Download PDF AI Translation

cs.CV / 17 / 2609.00447

Instance-Guided Report Anchoring for Text-Free 3D Abnormality Segmentation in Chest CT

基于实例引导的报告锚定用于无文本胸部CT的3D异常分割

Bu, Zhenyu, Ding, Haoyan, Shen, Chushu, Zheng, Xinyuan, Duan, Peiyu, Guo, Xueqi, Farhand, Sepehr, Shinagawa, Yoshihisa, Hermosillo, Gerardo, Wu, Chaowei

Abstract

Accurate 3D abnormality segmentation in chest CT requires dense spatial supervision, but obtaining expert voxel-level labels is costly. Radiology reports, however, are routinely generated during clinical interpretation and contain instance-specific descriptions that can provide additional guidance without new dense annotation. Existing vision-language grounding methods typically require report-derived findings at inference, making localization dependent on paired text and limiting each forward pass to a queried finding. We propose Instance-Guided Report Anchoring (IGRA), a model-agnostic module that preserves the correspondence between each annotated abnormality instance and the report finding that describes it. IGRA pools each instance representation and anchors it to the corresponding finding embedding during training; all text-related components are discarded at inference. We further reformulate free-text grounding on ReXGroundingCT as multi-label volumetric segmentation by merging same-category instances, allowing all abnormality categories to be predicted in one image-only forward pass. IGRA improves Dice by 22.5% over the strongest image-only baseline (30.93 vs. 25.25) and is comparable to VoxTell on the single-finding subset (30.29 vs. 30.43). Applied unchanged to four standard 3D segmentation backbones, IGRA improves Dice and hit rate across all architectures. Zero-shot evaluation on LIDC-IDRI, PleThora, and a private in-house dataset further shows consistent gains over image-only baselines.

Chinese Translation

在胸部CT中，准确的3D异常分割需要密集的空间监督，但获取专家级的体素级标签成本高昂。然而，放射学报告在临床解读过程中会定期生成，并包含特定实例的描述，这可以在不需要新的密集标注的情况下提供额外的指导。现有的视觉-语言基础方法通常在推理时需要报告衍生的发现，使得定位依赖于配对文本，并限制每次前向传播仅针对查询的发现。我们提出了实例引导的报告锚定（Instance-Guided Report Anchoring, IGRA），这是一个与模型无关的模块，保留每个标注异常实例与描述它的报告发现之间的对应关系。IGRA在训练期间对每个实例表示进行汇聚，并将其锚定到相应的发现嵌入；在推理时，所有与文本相关的组件被丢弃。我们进一步将ReXGroundingCT上的自由文本锚定重新表述为多标签体积分割，通过合并同类实例，允许在一次仅图像的前向传播中预测所有异常类别。IGRA在最强的仅图像基线（30.93 vs. 25.25）上提高了22.5%的Dice系数，并且在单一发现子集上与VoxTell相当（30.29 vs. 30.43）。IGRA在四个标准3D分割骨干网络上保持不变，提升了所有架构的Dice系数和命中率。在LIDC-IDRI、PleThora和一个私有内部数据集上的零-shot评估进一步显示出相对于仅图像基线的一致提升。

View on arXiv Download PDF AI Translation

cs.CV / 18 / 2609.00469

SAM3-LoRA: Parameter-Efficient Adaptation of a Concept-Promptable Foundation Model for Multi-Class Structural Defect Segmentation

SAM3-LoRA：用于多类结构缺陷分割的概念可提示基础模型的参数高效适应

Malaisree, P., Youwai, S., Janrungautai, S., Amorndechaphon, D., Rojanavasu, P., Songkitti, W.

Abstract

Promptable segmentation foundation models such as SAM3 accept an open-vocabulary text concept and return every instance matching it, but adapting them to a specialized domain by full fine-tuning is computationally prohibitive for the organizations that would benefit most. This study applies Low-Rank Adaptation (LoRA) to SAM3 for multi-class structural defect segmentation and examines both how such a model can be supervised from conventional annotation and whether the resulting efficiency gain transfers across datasets. Two contributions are methodological. First, we describe a supervision procedure that trains a concept-promptable model directly from COCO-style class-labeled instance segmentation by using the category name itself as the prompt, requiring no prompt templates, no synonym expansion, and no learned class embeddings. Second, we identify and mitigate a failure mode specific to this setting: because a conventional annotation file yields positive prompts exclusively, the model's presence prediction decouples from the text condition and degenerates into responding to any prompt, a collapse that is invisible to every metric computed on positive prompts alone. Exhaustive hard-negative prompting, in which every dataset category absent from an image is issued as a zero-detection query, addresses this at no annotation cost. Two adapter placements were compared under an identical protocol, updating 0.121% and 1.341% of model parameters. On a purpose-built tunnel lining dataset, pixel intersection-over-union improved from 0.017 to 0.338 and instance-level recall from 0.375 to 0.672; on the independent public Structural Defects Dataset, from 0.017 to 0.855 and from 0.574 to 1.000. Improvements were directionally consistent across ten metrics on both datasets, and the largest per-category gains occurred precisely where zero-shot competence was absent.

Chinese Translation

可提示分割基础模型如SAM3接受开放词汇的文本概念并返回与之匹配的每个实例，但通过完全微调将其适应于特定领域对那些最能受益的组织来说计算成本过高。本研究将低秩适应（LoRA）应用于SAM3，以进行多类结构缺陷分割，并考察该模型如何能够通过传统注释进行监督，以及所获得的效率提升是否能够跨数据集转移。两个贡献是方法论性的。首先，我们描述了一种监督程序，该程序通过使用类别名称本身作为提示，直接从COCO风格的类别标记实例分割中训练概念可提示模型，要求不使用提示模板、不进行同义词扩展，也不需要学习的类别嵌入。其次，我们识别并缓解了这一设置特有的失败模式：由于传统注释文件仅产生正提示，模型的存在预测与文本条件解耦，并退化为对任何提示的响应，这种崩溃在仅基于正提示计算的每个指标中都是不可见的。全面的硬负提示，即对图像中缺失的每个数据集类别发出零检测查询，在没有注释成本的情况下解决了这一问题。在相同协议下比较了两种适配器放置，更新了0.121%和1.341%的模型参数。在一个专门构建的隧道衬里数据集中，像素交并比从0.017提高到0.338，实例级召回率从0.375提高到0.672；在独立的公共结构缺陷数据集中，从0.017提高到0.855，召回率从0.574提高到1.000。两个数据集上的十个指标的改进方向一致，且每个类别的最大增益恰好发生在零-shot能力缺失的地方。

View on arXiv Download PDF AI Translation

cs.CV / 19 / 2609.00476

Less Is More: Balancing Positive and Negative Space in Visual Concept Blending

少即是多：在视觉概念融合中平衡正负空间

Xiao, Shishi, Coscia, Adam J., Laidlaw, David H.

Abstract

Graphic designers often blend visual concepts to communicate multiple ideas within a single image, leveraging positive and negative space to create balance, emphasis, and aesthetic appeal. While computational methods have begun to support automatic concept blending, they largely overlook the role of spatial composition in the design. To address this gap, we present an automatic pipeline that explicitly applies positive and negative space throughout the blending process. Our approach first identifies plausible regions for concept integration by combining semantic reasoning from vision-language models with geometric constraints derived from real-world examples. Conditioned on these regions, the system generates blended compositions using a hybrid pixel-vector pipeline: diffusion-based inpainting produces a fast, coarse initialization, which is then refined through vector-based optimization at the point level to ensure structural coherence and balanced semantic expression. A multimodal agent orchestrates this process as a planner and evaluator, enabling iterative improvement and interpretable control. Through an evaluation using both baseline comparisons and a user study, we demonstrate greater expressiveness, creativity, and concept recognizability by effectively leveraging positive and negative space. We further demonstrate the generalizability of our approach across diverse applications, including controllable image and infographic generation.

Chinese Translation

平面设计师常常融合视觉概念，以在单一图像中传达多重思想，利用正负空间创造平衡、强调和美学吸引力。尽管计算方法已开始支持自动概念融合，但它们在设计中往往忽视了空间构图的作用。为了解决这一问题，我们提出了一种自动化流程，明确地在整个融合过程中应用正负空间。我们的方法首先通过结合视觉-语言模型的语义推理与源自现实世界示例的几何约束，识别出概念整合的合理区域。在这些区域的基础上，系统使用混合像素-矢量流程生成融合的构图：基于扩散的修复技术提供快速、粗略的初始化，然后通过基于矢量的点级优化进行细化，以确保结构一致性和语义表达的平衡。一个多模态代理作为规划者和评估者协调这一过程，实现迭代改进和可解释的控制。通过基准比较和用户研究的评估，我们展示了通过有效利用正负空间，获得更强的表现力、创造力和概念可识别性。我们进一步展示了我们的方法在多种应用中的普遍适用性，包括可控的图像和信息图生成。

View on arXiv Download PDF AI Translation

cs.CV / 20 / 2609.00505

ViTAL-X: Video-Text Alignment with Cross-Modal Temporal Edits

ViTAL-X：具有跨模态时间编辑的视频-文本对齐

T V, Sethuraman, Khosla, Savya, Susladkar, Onkar Kishor, Tiwari, Aditi, Oh, Seoung Wug, Kafle, Kushal, Lee, Joon-Young, Hoiem, Derek, Jenni, Simon

Abstract

Video-text models adapted from image-text architectures (e.g., CLIP) frequently exhibit temporal blindness, the inability to perceive fundamental cues like order, direction, and motion dynamics. Standard datasets mask this limitation by enabling models to exploit static spatial shortcuts. To systematically evaluate this, we introduce XTE-Bench, a diagnostic probe revealing that even large-scale video-language models struggle with basic temporal reasoning, indicating that parameter scaling alone is insufficient to resolve this flaw. To address this, we propose Cross-Modal Temporal Edits (XTE), a self-supervised framework that injects precise temporal supervision. By performing synchronized video-text transformations, XTE generates hard temporal negatives without manual annotation. We instantiate this with ViTAL-X, a lightweight model that equips frozen image-text backbones with temporal awareness while preserving their foundational spatial knowledge. Across six temporal benchmarks, ViTAL-X achieves state-of-the-art performance. Utilizing only 0.4B parameters and 1M training clips, ViTAL-X outperforms 7B-parameter models and surpasses baselines trained on 600x more data. These results demonstrate that targeted, high-quality temporal alignment provides a highly efficient alternative to pure scaling.

Chinese Translation

从图像-文本架构（例如，CLIP）改编的视频-文本模型常常表现出时间盲ness，即无法感知基本线索如顺序、方向和运动动态。标准数据集通过使模型能够利用静态空间捷径来掩盖这一局限性。为了系统地评估这一点，我们引入了XTE-Bench，一个诊断探针，揭示即使是大规模视频-语言模型在基本时间推理方面也存在困难，表明仅仅增加参数规模不足以解决这一缺陷。为了解决这个问题，我们提出了跨模态时间编辑（Cross-Modal Temporal Edits, XTE），这是一个自监督框架，注入精确的时间监督。通过执行同步的视频-文本转换，XTE生成了困难的时间负样本，而无需手动标注。我们用ViTAL-X实例化这一方法，这是一种轻量级模型，赋予冻结的图像-文本骨干网络时间感知，同时保留其基础的空间知识。在六个时间基准测试中，ViTAL-X达到了最先进的性能。仅利用0.4B参数和1M训练片段，ViTAL-X超越了7B参数模型，并且超过了在600倍更多数据上训练的基线。这些结果表明，针对性的高质量时间对齐提供了一种比单纯扩展更高效的替代方案。

View on arXiv Download PDF AI Translation

cs.CV / 21 / 2609.00521

Soft-Argmax for the Projective Plane via the Veronese Embedding

通过 Veronese 嵌入实现投影平面的软极大值

El-Zein, Benjamin, Eckert, Dominik, Zech, Paul, Syben, Christopher, Geiger, Bernhard, Kappler, Steffen, Stober, Sebastian

Abstract

From horizon detection to fibre structures in X-ray imaging, many vision tasks recover lines via peak detection in Hough space $H=S^1\times\mathbb{R}$, the domain of orientation-offset pairs $(\theta,\rho)$. Differentiable pipelines extract coordinates via \emph{soft-argmax}, a probability-weighted average that is only meaningful in a globally linear space. However, $(\theta,\rho)$ and $(\theta+\pi,-\rho)$ describe the same undirected line, so $H$ double-covers the space of undirected lines $H/\mathbb{Z}_2$: a M\"obius strip, obtained by identifying each pair under $\mathbb{Z}_2$ action. Soft-argmax operates on the cover $H$, but since $H/\mathbb{Z}_2$ admits no linear structure, it tears geometrically adjacent lines apart. Thus we need a $\mathbb{Z}_2$-invariant embedding of lines into a linear space, on which soft-argmax is well-defined. We achieve this by parametrising lines via unit-norm homogeneous vectors $\ell=(1+\rho^2)^{-1/2}(\cos\theta,\sin\theta,-\rho)^{\top}\in\mathbb{R}^3$ and applying the Veronese map $v_2(\ell)=\ell\ell^{\top}$ that satisfies $v_2(\ell)=v_2(-\ell)$. This descends continuously to an embedding of the quotient $H/\mathbb{Z}_2$ into the linear space $\mathrm{Sym}^2(\mathbb{R}^3)$, where the antipodal ambiguity vanishes. Line extraction becomes a barycentre in $\mathrm{Sym}^2(\mathbb{R}^3)$, projected back via its leading eigenvector. We validate our \emph{Veronese soft-argmax} in a Hough transform-based network across all resolvable lines, confirming uniform and seam-free recovery. We further derive that the $L_2$-loss on isometrically weighted Veronese embeddings equals the squared chordal distance between lines in projective space, enabling a geometrically precise training objective.

Chinese Translation

从地平线检测到 X 射线成像中的纤维结构，许多视觉任务通过在霍夫空间 $H=S^1 imes ext{R}$ 中的峰值检测来恢复直线，该空间是方向-偏移对 $( heta, ho)$ 的定义域。可微分管道通过 extit{soft-argmax} 提取坐标，这是一种在全局线性空间中才有意义的概率加权平均。然而，$( heta, ho)$ 和 $( heta+ ext{π},- ho)$ 描述的是同一条无向直线，因此 $H$ 双重覆盖了无向直线的空间 $H/ ext{Z}_2$：一个莫比乌斯带，通过在 $ ext{Z}_2$ 操作下识别每对而获得。软极大值在覆盖空间 $H$ 上操作，但由于 $H/ ext{Z}_2$ 不具有线性结构，它会将几何上相邻的直线撕裂。因此，我们需要一个 $ ext{Z}_2$ 不变的直线嵌入到线性空间中，使得软极大值在其上定义良好。我们通过使用单位范数齐次向量 $oldsymbol{ ext{ℓ}}=(1+ ho^2)^{-1/2}( ext{cos} heta, ext{sin} heta,- ho)^{ op} ext{∈} ext{R}^3$ 参数化直线，并应用满足 $v_2(oldsymbol{ ext{ℓ}})=v_2(-oldsymbol{ ext{ℓ}})$ 的 Veronese 映射 $v_2(oldsymbol{ ext{ℓ}})=oldsymbol{ ext{ℓ}}oldsymbol{ ext{ℓ}}^{ op}$。这连续降维到商空间 $H/ ext{Z}_2$ 的嵌入到线性空间 $ ext{Sym}^2( ext{R}^3)$，在这里对立模糊消失。直线提取变成了在 $ ext{Sym}^2( ext{R}^3)$ 中的重心，通过其主特征向量投影回去。我们在基于霍夫变换的网络中验证了我们的 extit{Veronese soft-argmax}，涵盖所有可解析的直线，确认了均匀且无缝的恢复。我们进一步推导出，等距加权的 Veronese 嵌入上的 $L_2$ 损失等于投影空间中直线之间的平方弦距，从而实现几何精确的训练目标。

View on arXiv Download PDF AI Translation

cs.CV / 22 / 2609.00525

GenScale: A Benchmark for Relative Object Scale in Image Generation and Editing

GenScale：图像生成与编辑中相对物体尺度的基准测试

Li, Lingxiao, Whitton, Max, Wu, Ledell, Gong, Boqing

Abstract

Modern image generation and editing systems can produce photorealistic, prompt-aligned images, but still often render familiar objects at implausible relative sizes. To measure this failure mode, we introduce GenScale, a benchmark and evaluation protocol for real-world relative object scale in image generation and editing. GenScale contains 900 image-level entries and 1,643 pairwise anchor-target scale relations across common-object generation, human-product generation with metric dimensions, and scale correction from failed generations. We further design a human-calibrated ordinal judge for scalable pairwise scale evaluation. Last but not the least, we introduce Rescale, a model-agnostic post-processing agent for localized scale correction without modifying the source generator. Experiments reveal that state-of-the-art image generators and editors cannot reliably observe relative scale yet, while Rescale consistently improves scale plausibility across generated and edited images. Together, GenScale establishes relative object scale as a distinct, measurable, and actionable capability for image generation systems.

Chinese Translation

现代图像生成和编辑系统能够生成与提示一致的照片级真实图像，但仍然常常以不合理的相对尺寸呈现熟悉的物体。为了衡量这种失败模式，我们引入了GenScale，一个用于图像生成和编辑中现实世界相对物体尺度的基准测试和评估协议。GenScale包含900个图像级条目和1,643对常见物体生成、人类产品生成（具有度量尺寸）以及从失败生成中进行尺度校正的锚点-目标尺度关系。我们进一步设计了一个人类校准的序数评估者，用于可扩展的成对尺度评估。最后但同样重要的是，我们引入了Rescale，一个模型无关的后处理代理，用于局部尺度校正，而不修改源生成器。实验结果表明，最先进的图像生成器和编辑器尚无法可靠地观察相对尺度，而Rescale在生成和编辑的图像中始终提高了尺度的合理性。综上所述，GenScale确立了相对物体尺度作为图像生成系统的一个独特、可测量和可操作的能力。

View on arXiv Download PDF AI Translation

cs.CV / 23 / 2609.00591

A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss

你所需要的只是一个瞬间：使用 SimLoss 的单次精细图像描述

Jain, Suryaansh, Barkur, Rahasya, G, Vishal, Rossi, Ryan, Dernoncourt, Franck, Wang, Jack, Goswami, Koustava, Lipka, Nedim, Mathur, Puneet, Basu, Samyadeep, Yoon, Seunghyun

Abstract

An image may be worth a thousand words, but most captioning models describe it in only a few. Modern vision-language models produce fluent high-level captions, yet routinely miss the attributes, counts, textures, materials, and spatial relations that make an image visually specific. Recent multi-stage systems recover some of these details through generation, decomposition, verification, and rewriting, but they do so at the expense of substantially higher inference latency. We propose SimLoss, a reference-free embedding-space objective for single-pass fine-grained image captioning. SimLoss trains a vision-language model to align its projected hidden-state representation with a frozen image embedding through an InfoNCE contrastive loss, supplying a dense visual supervision signal before any text is decoded, and requiring neither human-written fine-grained captions nor pseudo-captions from a multi-stage pipeline. We instantiate it as SimLoss FFT, which backpropagates through a locally available embedding model, and SimLoss GRPO, which treats that model as a black-box reward. Compared with single-pass, multi-stage verification, reward-optimized, and perception-aware baselines, the fully differentiable fine-tuning variant, SimLoss FFT, achieves the highest precision while nearly matching the F1 score of the multi-stage method, all while retaining single-pass inference and running roughly 20 times faster than the multi-stage pipeline. The reward-based variant SimLoss GRPO attains the strongest recall. Together, these results show that embedding-space supervision can recover the quality of multi-stage verification at the latency of a single-pass captioner.

Chinese Translation

一幅图像可能胜过千言万语，但大多数描述模型仅用几句话来描述它。现代视觉-语言模型生成流畅的高层次描述，但常常忽略了使图像在视觉上独特的属性、数量、纹理、材料和空间关系。近期的多阶段系统通过生成、分解、验证和重写来恢复这些细节，但这样做的代价是显著增加推理延迟。我们提出了 SimLoss，这是一种无参考的嵌入空间目标，用于单次精细图像描述。SimLoss 训练视觉-语言模型，使其投影的隐藏状态表示与冻结的图像嵌入通过 InfoNCE 对比损失对齐，在任何文本被解码之前提供密集的视觉监督信号，并且不需要人工编写的精细描述或来自多阶段管道的伪描述。我们将其实例化为 SimLoss FFT，该方法通过本地可用的嵌入模型进行反向传播，以及 SimLoss GRPO，该方法将该模型视为黑箱奖励。与单次、多阶段验证、奖励优化和感知感知基线相比，完全可微调的精细调整变体 SimLoss FFT 在保持单次推理的同时，达到了最高的精度，并且几乎匹配了多阶段方法的 F1 分数，运行速度大约是多阶段管道的 20 倍。基于奖励的变体 SimLoss GRPO 达到了最强的召回率。这些结果表明，嵌入空间监督可以在单次描述器的延迟下恢复多阶段验证的质量。

View on arXiv Download PDF AI Translation

cs.CV / 24 / 2609.00593

BrainDiff: Longitudinal Report Generation for Multimodal Brain MRI

BrainDiff：多模态脑MRI的纵向报告生成

Patel, Krish, Liu, Peirong

Abstract

Neuroradiologists rarely read a brain MRI in isolation, yet automated brain-MRI report generation has been built almost entirely for single studies. Temporal analysis has been explored on chest radiography and chest CT, but to our knowledge, longitudinal reporting for brain MRI, where interval change is often subtle and spatially distributed, remains unaddressed. We present BrainDiff, the first longitudinal vision-language system for brain MRI. BrainDiff outperforms both frontier general-purpose and single-study neuroimaging models on the same patient pairs. Moreover, BrainDiff retains 91% of internal RadGraph-XL entity+relation F1 (rg_er) on an external, cross-hospital cohort. Beyond the system, we contribute three analyses. First, we identify two independent grounding levers: a counterfactual objective with prior-report dropout, which increases measured image reliance by ~47%, and a staged curriculum. Together, these interventions raise image reliance 2.5-fold from the baseline. Second, we provide a factorial over prior-report availability and image identity, isolating a visual contribution of +0.0387 rg_er, which grows when the prior report is withheld. Third, a cheap change-decodability test for candidate backbones shows that interval change is decodable far more weakly than single-study pathology (0.60 vs. 0.77 AUROC). Code is publicly available at https://github.com/jhuldr/BrainDiff.

Chinese Translation

神经放射科医生很少单独解读脑MRI，然而，自动化的脑MRI报告生成几乎完全是为单一研究而构建的。虽然在胸部X光和胸部CT中已经探索了时间分析，但据我们所知，针对脑MRI的纵向报告（其中间隔变化通常较为微妙且空间分布广泛）仍未得到解决。我们提出了BrainDiff，这是第一个用于脑MRI的纵向视觉-语言系统。BrainDiff在相同患者对比中超越了前沿的通用和单一研究神经影像模型。此外，BrainDiff在一个外部跨医院队列中保留了91%的内部RadGraph-XL实体+关系F1（rg_er）。除了系统本身，我们还贡献了三项分析。首先，我们识别出两个独立的基础支撑杠杆：一个具有先前报告丢失的反事实目标，增加了约47%的图像依赖性，以及一个分阶段的课程。这些干预措施共同将图像依赖性从基线提高了2.5倍。其次，我们提供了对先前报告可用性和图像身份的因子分析，隔离出视觉贡献+0.0387 rg_er，当先前报告被扣留时，该贡献会增加。第三，对于候选骨干网络的低成本变化可解码性测试显示，间隔变化的可解码性远低于单一研究病理（0.60 vs. 0.77 AUROC）。代码已公开可用，网址为 https://github.com/jhuldr/BrainDiff。

View on arXiv Download PDF AI Translation

cs.CV / 25 / 2609.00610

Streaming4D: Accelerate 4D World Models via Block-wise Video Generation and Incremental Reconstruction

Streaming4D：通过块级视频生成和增量重建加速4D世界模型

Liu, Xiaoyan, Liu, Jiaxin, Li, Kangrui, Zhou, Sifan

Abstract

Current 4D generation paradigms are often bottlenecked by a sequential decoupling design: video is generated first, followed by 3D reconstruction, leading to high interaction latency. This limits applications in interactive real-time scenarios. To this end, we propose \textbf{Streaming4D}, a tightly coupled synchronous pipeline that integrates block-wise autoregressive video generation with incremental 3D reconstruction. Unlike traditional frame-by-frame emission and delayed geometry recovery, Streaming4D generates temporal video blocks and immediately triggers reconstruction for each completed block, enabling parallel execution between synthesis and geometric updates. This approach allows the world representation to evolve online with the video stream, reducing feedback latency while preserving geometric fidelity. We instantiate \textbf{Streaming4D} using a Self-Forcing-style autoregressive generator and an incremental reconstruction backend. Experiments show consistent runtime improvements across resolutions on a single RTX 4090 (1.24$\times$ speedup), while maintaining high-quality 4D geometry and multi-view consistency.

Chinese Translation

当前的4D生成范式常常受到顺序解耦设计的瓶颈：视频首先生成，然后进行3D重建，这导致高交互延迟。这限制了在交互实时场景中的应用。为此，我们提出了 extbf{Streaming4D}，一个紧密耦合的同步管道，集成了块级自回归视频生成与增量3D重建。与传统的逐帧生成和延迟几何恢复不同，Streaming4D生成时间视频块，并立即触发每个完成块的重建，实现合成与几何更新之间的并行执行。这种方法使得世界表示能够随着视频流在线演变，减少反馈延迟，同时保持几何保真度。我们使用自强式自回归生成器和增量重建后端实例化了 extbf{Streaming4D}。实验表明，在单个RTX 4090上，各种分辨率下的运行时间均有一致的提升（1.24$ imes$加速），同时保持高质量的4D几何和多视图一致性。

View on arXiv Download PDF AI Translation

cs.CV / 26 / 2609.00628

Restrict, Don't Retrain: Inference-Time VLM Guidance for Zero-Shot Aerial Segmentation

限制，而非重训练：零-shot 空中分割的推理时 VLM 指导

DiMeola, Teresa, Walter, Charles, Xiao, Hong

Abstract

Global welfare often depends on the correct interpretation of aerial and satellite imagery. Acting on such imagery (mapping flooded ground, crop extent, or damaged infrastructure) demands pixel-level segmentation to ensure perfect class localization. Pretrained general foundation models, when applied directly, often miss important features and cannot always find all the classes belonging to a given scene, overlooking smaller objects that matter most. We use a single consumer-grade GPU running a vision-language model (VLM) to supply this missing guidance, improving segmentation while producing structured, auditable evidence that drives the result and can be inspected on its own. We fuse three approaches: the frozen foundation model that labels every pixel, and two queries to a VLM, one to choose the classes that matter, and one to locate the small objects the base model misses. Evaluating across four aerial datasets, we see consistent gains at each stage where the base model is competent.

Chinese Translation

全球福利往往依赖于对航空和卫星影像的正确解读。对这些影像采取行动（如绘制淹没的土地、作物范围或受损基础设施）需要像素级的分割，以确保完美的类别定位。当预训练的通用基础模型直接应用时，往往会遗漏重要特征，并不总能找到属于特定场景的所有类别，忽视了那些最重要的小物体。我们使用一台单一的消费级 GPU 运行视觉-语言模型（VLM）来提供这些缺失的指导，从而改善分割效果，同时生成结构化、可审计的证据，推动结果并可以单独检查。我们融合了三种方法：冻结的基础模型为每个像素标记，两个对 VLM 的查询，一个用于选择重要类别，另一个用于定位基础模型遗漏的小物体。在四个航空数据集上的评估中，我们在基础模型能够胜任的每个阶段都观察到了持续的提升。

View on arXiv Download PDF AI Translation

cs.CV / 27 / 2609.00634

Beyond Landmark Extraction: A Framework for Robust Geometric Feature Construction in Structured Image Classification

超越地标提取：结构化图像分类中稳健几何特征构建的框架

Mauree, Saravana, Arya, Sakshi

Abstract

Much of the literature on structured image recognition has disproportionately focused on the comparison of classification algorithms. Rather than investigating which classifier performs best, this paper instead asks: what should a classifier know before it ever makes a prediction? In structured vision problems such as gesture recognition, facial expression categorization, and medical image analysis, discriminative information lies less in individual pixels and more in spatial relationships between semantic parts. Raw pixel spaces are high-dimensional, sensitive to nuisance variation, and often obfuscate the geometric structures that make visual tasks interpretable. Landmark extraction provides one form of dimension reduction, but it does not by itself determine the information preserved. This paper studies the post-landmark feature map as the central object of analysis and proposes a systematic framework for constructing and interpreting landmark-derived representations as an, informed, feature-based ``dimension reduction'' step. Using static hand gesture recognition as a case study, we evaluate coordinate, distance, angle, and hybrid representations through perturbation and ablation experiments. The results show that visually variable data exposes substantial gaps between raw coordinate features and their geometrically invariant counterparts, while hybrid representations achieve the strongest overall performance by combining complementary geometric components. These findings frame feature construction as a fundamental modeling decision and ultimately suggests that the question of what representation should a classifier learn from is one worth asking. The code used for feature construction and evaluation is available at https://github.com/ShivMaureeCWRU/Feature_based_dimension_reduction

Chinese Translation

关于结构化图像识别的文献大多不成比例地集中于分类算法的比较。本文并不探讨哪种分类器表现最佳，而是提出了一个问题：分类器在做出预测之前应该了解什么？在手势识别、面部表情分类和医学图像分析等结构化视觉问题中，区分信息往往不在于单个像素，而在于语义部分之间的空间关系。原始像素空间是高维的，对干扰变化敏感，且常常模糊了使视觉任务可解释的几何结构。地标提取提供了一种维度降低的形式，但它本身并不能决定所保留的信息。本文将后地标特征图作为分析的核心对象，提出了一个系统框架，用于构建和解释基于地标的表示，作为一个经过深思熟虑的特征基础“维度降低”步骤。以静态手势识别为案例研究，我们通过扰动和消融实验评估坐标、距离、角度和混合表示。结果表明，视觉可变数据暴露了原始坐标特征与其几何不变对应物之间的显著差距，而混合表示通过结合互补的几何组件实现了最强的整体性能。这些发现将特征构建框架视为一个基本的建模决策，并最终建议提出分类器应学习什么表示是一个值得探讨的问题。用于特征构建和评估的代码可在 https://github.com/ShivMaureeCWRU/Feature_based_dimension_reduction 获取。

View on arXiv Download PDF AI Translation

cs.CV / 28 / 2609.00649

You Cannot Photograph the Same Street Twice: Reliability Limits in Vision-Language Measurement of Urban Change

你无法两次拍摄同一条街道：视觉-语言测量城市变化的可靠性限制

Tan, Kaizhen

Abstract

Vision-language models are increasingly used to measure urban change from repeated street-level imagery, but their longitudinal reliability is not well understood. We test how much a perception score can change when the street itself does not undergo substantial redevelopment. Using 4,648 consecutive-epoch image pairs from 435 Google Street View standpoints across five US cities, we find that re-photographing the same street changes a perception score by 0.80 points on average, equivalent to 66.5% of the difference between two different streets in the same city. Repeated model calls contribute almost no variation, while image re-encoding and prompt-order changes each account for about one fifth of the between-street difference. Six image statistics describing scattering, contrast, colour, exposure, sharpness and specularity explain almost none of the remaining epoch-to-epoch variation. A small systematic drift of about 0.1 points remains and increases with the interval between captures, consistent with minor physical changes not recorded by redevelopment labels. Controlled experiments further show that acquisition conditions can shift scores when camera and image properties are allowed to vary, and that the direction of these shifts depends on the model. In crowdsourced imagery, camera geometry alone causes a model to report physical change in 45% of identical-scene pairs; normalising both images to a common virtual camera reduces this rate to 7.5%. Despite poor reliability at the individual-location level, aggregation recovers a coherent redevelopment signal: changed streets are judged wealthier, better maintained, more enclosed and less green. These results show that vision-language measurement of urban change is reliable at the scale of hundreds of paired observations, but not at the scale of individual sample points.

Chinese Translation

视觉-语言模型越来越多地用于通过重复的街景图像测量城市变化，但其纵向可靠性尚不清楚。我们测试了在街道本身没有经历重大重建的情况下，感知评分可以变化多少。使用来自五个美国城市的435个Google街景视角的4,648对连续时期图像，我们发现重新拍摄同一条街道平均会使感知评分变化0.80分，相当于同一城市中两条不同街道之间差异的66.5%。重复模型调用几乎不贡献变化，而图像重新编码和提示顺序的变化各自占街道间差异的约五分之一。六个描述散射、对比度、颜色、曝光、清晰度和光泽度的图像统计几乎无法解释剩余的时期间变化。约0.1分的小系统性漂移仍然存在，并随着拍摄间隔的增加而增加，这与未被重建标签记录的小物理变化一致。控制实验进一步表明，当相机和图像属性允许变化时，获取条件可以改变评分，而这些变化的方向依赖于模型。在众包图像中，仅相机几何就导致模型在45%的相同场景对中报告物理变化；将两幅图像规范化为共同的虚拟相机将这一比例降低到7.5%。尽管在单个位置层面的可靠性较差，但聚合恢复了一个连贯的重建信号：变化的街道被评估为更富裕、维护更好、封闭性更强且绿化更少。这些结果表明，视觉-语言测量城市变化在数百对观察的规模上是可靠的，但在单个样本点的规模上则不然。

View on arXiv Download PDF AI Translation

cs.CV / 29 / 2609.00656

Physically Plausible Video Generation via Visual-Semantic Chain-of-Events Conditioning

通过视觉-语义事件链条件生成物理上可信的视频

Wang, Zixuan, Hu, Yixin, Li, Wen, Chen, Feng, Liu, Yan, Peng, Duo, Lei, Yinjie

Abstract

Physically Plausible Video Generation (PPVG) seeks to synthesize videos consistent with physical principles, yet remains challenging due to underspecified natural language conditioning. Advanced chain-of-thought (CoT) frameworks augment prompts with physical knowledge. However, such prompts describe physical phenomena holistically, overlooking intermediate states and transition dynamics. In this paper, we reformulate PPVG as event-centric generation by representing physical evolution as a chain of causally connected and physically constrained events. Our framework comprises three key modules: (1) Physics-driven Event Chain Reasoning. This module decomposes physical phenomena into causally connected events represented by evolving scene graphs. Formula-derived physical quantities are bound to relevant objects and interactions, characterizing the direction and magnitude of each event transition. (2) Transition-aware Routed Keyframe Conditioning. This module routes each event to a specialized keyframe synthesis operator for appearance variation or object transformation. Consecutive keyframes are injected as residual guidance during denoising, enabling smooth visual transitions between event-boundary states. (3) Physics-injected Contrastive Semantic Guidance. This module constructs physics-informed positive and counterfactual negative prompts for classifier-free guidance, steering generation toward plausible dynamics and away from physics-violating counterparts. Experiments on PhyGenBench, VideoPhy, PhyWorldBench, and Physics-IQ demonstrate that our framework generates videos with superior physical plausibility across diverse domains.

Chinese Translation

物理上可信的视频生成（PPVG）旨在合成符合物理原则的视频，但由于自然语言条件的不明确性，这一任务仍然具有挑战性。先进的思维链（CoT）框架通过物理知识增强提示。然而，这些提示整体上描述物理现象，忽视了中间状态和过渡动态。在本文中，我们将PPVG重新表述为以事件为中心的生成，通过将物理演变表示为一系列因果连接且受物理约束的事件。我们的框架包括三个关键模块：（1）基于物理驱动的事件链推理。该模块将物理现象分解为由不断演变的场景图表示的因果连接事件。公式推导的物理量与相关对象和交互绑定，表征每个事件过渡的方向和幅度。（2）过渡感知的路由关键帧条件。该模块将每个事件路由到专门的关键帧合成操作符，以实现外观变化或对象变换。在去噪过程中，连续的关键帧作为残差指导注入，使事件边界状态之间的视觉过渡平滑。（3）注入物理的对比语义指导。该模块构建基于物理的正向和反事实负向提示，用于无分类器指导，引导生成朝向可信的动态，远离违反物理的对立面。在PhyGenBench、VideoPhy、PhyWorldBench和Physics-IQ上的实验表明，我们的框架在不同领域生成的视频具有更高的物理可信性。

View on arXiv Download PDF AI Translation

cs.CV / 30 / 2609.00658

Teaching Vision-Language Models to Use the Scale They Are Given: Label-Free Equivariance Training for Metric Physical Reasoning

教会视觉-语言模型利用其所给的尺度：无标签等变训练用于度量物理推理

Tan, Kaizhen, Feng, Yang, Du, Heqing, Tao, Siru, Xu, Xin, Hong, Hanzhe

Abstract

Metric questions about video require vision-language models to use supplied real-world references to convert visual measurements into physical units. Yet we find that current models use this scale information only partially. When every world-space quantity in a prompt is rescaled by a common factor, the video remains equally valid and the correct answer changes by exactly that factor, but model predictions move only part of the way and accuracy remains concentrated near the familiar scale of the depicted objects. Across eight vision-language models, this under-response persists over four orders of magnitude. The same models recover the correct closed-form scaling laws when the identical physics is asked in a scale-free form, indicating that the main deficit lies in metric grounding rather than physical mechanism knowledge. We use this exact scaling relation as supervision without requiring metric annotations. Under a common rescaling of the supplied world-space quantities, the correct metric answer must change by the same factor. EquiSD exploits this constraint by projecting a model's own prediction onto the scale-equivariant family and fine-tuning the model on the resulting targets. It requires no ground-truth answers and only one model query per training video. On held-out simulated videos, EquiSD increases a 3B model's median response slope from 0.66 to 0.94 and improves mean relative accuracy by 9.2 points across scales. The learned relation generalizes to unseen world scales and transfers without adaptation to real QuantiPhy videos, where accuracy increases by 6.4 points. These results show that an exact physical symmetry can provide label-free supervision for improving metric grounding in vision-language models.

Chinese Translation

关于视频的度量问题要求视觉-语言模型使用提供的现实世界参考，将视觉测量转换为物理单位。然而，我们发现当前模型仅部分利用了这一尺度信息。当提示中的每个世界空间量都按一个共同因子重新缩放时，视频依然有效，正确答案也正好按该因子变化，但模型预测仅部分响应，准确性仍集中在所描绘对象的熟悉尺度附近。在八个视觉-语言模型中，这种不足的响应在四个数量级上持续存在。这些模型在以无尺度形式询问相同物理问题时能够恢复正确的封闭形式缩放定律，表明主要缺陷在于度量基础而非物理机制知识。我们利用这一确切的缩放关系作为监督，而无需度量注释。在提供的世界空间量的共同重新缩放下，正确的度量答案必须按相同因子变化。EquiSD利用这一约束，通过将模型自身的预测投影到尺度等变家族上，并在生成的目标上微调模型。它不需要真实答案，并且每个训练视频只需一次模型查询。在保留的模拟视频上，EquiSD将一个3B模型的中位响应斜率从0.66提高到0.94，并在各尺度上提高了9.2个点的平均相对准确性。学习到的关系能够推广到未见过的世界尺度，并在真实的QuantiPhy视频中无须适应地转移，准确性提高了6.4个点。这些结果表明，确切的物理对称性可以为改善视觉-语言模型中的度量基础提供无标签的监督。

View on arXiv Download PDF AI Translation

cs.CV / 31 / 2609.00661

Do Satellites See Commuters? A Critical Benchmark of Vision Foundation Models

卫星能看到通勤者吗？视觉基础模型的关键基准评估

Iqbal, Ashiq Shukoor, Wongso, Wilson, Salim, Flora D.

Abstract

Satellite foundation models offer a globally available alternative to census data for commuting origin-destination (OD) generation, yet no study has systematically compared encoder paradigms within a single downstream pipeline. We ablate four satellite vision encoders: language-supervised (RemoteCLIP), self-supervised (DINOv3), and geographically grounded (SatCLIP, AlphaEarth) within an identical WeDAN graph diffusion framework across 1,925 US counties, 325 UK districts, and 14 global cities under five random seeds. Three main findings emerge. First, language-supervised features achieve the strongest in-distribution performance (RemoteCLIP CPC 0.602), while geographically grounded encoders transfer more reliably zero-shot: AlphaEarth improves CPC by 33% over RemoteCLIP on UK districts. Second, pretraining corpus scale alone is insufficient: DINOv3, trained on a substantially larger satellite corpus, underperforms RemoteCLIP by 0.091 CPC in-distribution and collapses to CPC 0.022 globally. Third, no encoder transfers usefully to global cities (best CPC 0.122 for RemoteCLIP, 0.022 for DINOv3), confirming cross-continental OD generation remains an open problem. We additionally clarify the semantics of the census noise parameter $\eta$, whose ordering reverses under cross-continental evaluation, a distinction critical to correctly interpreting prior results. Training scripts and evaluation logs will be released.

Chinese Translation

卫星基础模型为通勤起讫点（OD）生成提供了一个全球可用的替代人口普查数据的方案，但尚无研究在单一下游管道中系统性地比较编码器范式。我们在相同的WeDAN图扩散框架下，对四种卫星视觉编码器进行了消融实验：语言监督（RemoteCLIP）、自监督（DINOv3）和地理基础（SatCLIP，AlphaEarth），涵盖1,925个美国县、325个英国区和14个全球城市，使用五个随机种子。研究得出三个主要发现。首先，语言监督特征在分布内表现最强（RemoteCLIP CPC 0.602），而地理基础编码器在零样本转移上更为可靠：AlphaEarth在英国区的CPC比RemoteCLIP提高了33%。其次，仅靠预训练语料库的规模是不够的：DINOv3在一个显著更大的卫星语料库上训练，但在分布内的表现比RemoteCLIP低0.091 CPC，并在全球范围内崩溃至CPC 0.022。第三，没有编码器能有效转移到全球城市（RemoteCLIP最佳CPC为0.122，DINOv3为0.022），确认跨洲OD生成仍然是一个未解决的问题。我们还澄清了人口普查噪声参数$ heta$的语义，其排序在跨洲评估中发生反转，这一区别对于正确解读先前结果至关重要。训练脚本和评估日志将被发布。

View on arXiv Download PDF AI Translation

cs.CV / 32 / 2609.00663

Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures

在视觉-语言模型中将感知与推理分离：一种无模型的渲染上限用于晶体结构

Polat, Can, Kurban, Mustafa, Serpedin, Erchin, Kurban, Hasan

Abstract

Multimodal evaluations cannot say whether a vision-language model misread an image or misreasoned about it, because every existing method for separating the two places a second model in the loop. We introduce the render ceiling, a model-free reference for benchmarks built by rendering known objects: inverting the frozen cameras and re-solving cross-view correspondence recovers exactly the answer the images support. We prove the ceiling fails only through an enumerable set of projection coincidences and certify that set empty on 2,160 rendered crystal structures, so every point of a model's deficit belongs to the model. Across fourteen vision-language models, supplying exact geometry as text lifts every model yet closes under half the gap for thirteen, while a supervised vision model with no language component reads the same images at 0.8952, above every vision-language model. The instrument exposes extraction-stage fabrication that downstream accuracy would misattribute to reasoning, yields camera-placement rules for benchmark builders, and transfers to any benchmark with an invertible forward rendering.

Chinese Translation

多模态评估无法判断视觉-语言模型是误读了图像还是对其进行了错误推理，因为现有的所有分离方法都在循环中引入了第二个模型。我们引入了渲染上限（render ceiling），这是一个无模型的基准参考，通过渲染已知对象构建：反转固定摄像机并重新解决视角对应关系可以准确恢复图像所支持的答案。我们证明渲染上限仅在可枚举的投影重合情况下失效，并在2160个渲染的晶体结构上认证该集合为空，因此模型的每一个不足之处都归属于模型本身。在十四个视觉-语言模型中，提供精确几何信息作为文本使每个模型的表现提升，但对于十三个模型仍然只能缩小一半的差距，而一个没有语言组件的监督视觉模型在读取相同图像时得分为0.8952，超过了所有视觉-语言模型。该工具揭示了提取阶段的制造问题，这些问题会被下游准确性错误归因于推理，并为基准构建者提供了摄像机放置规则，同时可转移到任何具有可逆前向渲染的基准。

View on arXiv Download PDF AI Translation

cs.CV / 33 / 2609.00666

DGNet: Dual-knowledge Guided Network for Infrared Small Target Detection

DGNet：双知识引导网络用于红外小目标检测

Yu, Chenglong, Xu, Mingzhu, Wang, Jing, Wang, Tongtong, Miao, Pingping, Nie, Liqiang

Abstract

InfRared Small Target Detection (IRSTD) is a prominent and challenging task in computer vision. In recent years, text-guided methods have significantly improved detection performance. However, they still suffer from two key limitations. First, a single text description simultaneously modeling both background and target leads to semantic entanglement, which contradicts the objective of background suppression and target enhancement. Second, reliance on image-specific textual prompts (requiring additional external models such as CLIP during inference) results in deployment constraints. To address these issues, we propose a novel Dual-knowledge Guided Network (DGNet) based on multiple generalizable texts. Specifically, we design a Prior-knowledge Wavelet Modulation (PWM) module, which leverages dual textual priors that separately characterize large-scale backgrounds and sparse targets to effectively disentangle and modulate entangled semantics in the frequency domain. Furthermore, we introduce a Consensus-knowledge Directional Alignment (CDA) loss, which models the initial state and the ideal target across samples as `complex background' and `bright target', respectively, thereby constructing a clear and unified directional optimization trajectory for the model. Extensive experiments on three public datasets demonstrate the superior performance of DGNet and the effectiveness of each component. The source code is available at https://github.com/iLearn-Lab/MM26-DGNet.

Chinese Translation

红外小目标检测（IRSTD）是计算机视觉领域一个重要且具有挑战性的任务。近年来，文本引导的方法显著提高了检测性能。然而，它们仍然面临两个主要限制。首先，单一文本描述同时建模背景和目标导致语义纠缠，这与背景抑制和目标增强的目标相悖。其次，依赖于特定图像的文本提示（在推理过程中需要额外的外部模型，如CLIP）导致了部署上的限制。为了解决这些问题，我们提出了一种基于多种可泛化文本的新型双知识引导网络（DGNet）。具体而言，我们设计了一个先验知识小波调制（PWM）模块，该模块利用分别表征大规模背景和稀疏目标的双重文本先验，有效地解开并调制频域中的纠缠语义。此外，我们引入了一种共识知识方向对齐（CDA）损失，它将样本的初始状态和理想目标建模为“复杂背景”和“明亮目标”，从而为模型构建一个清晰统一的方向优化轨迹。在三个公共数据集上的大量实验表明，DGNet的优越性能以及各个组件的有效性。源代码可在 https://github.com/iLearn-Lab/MM26-DGNet 获取。

View on arXiv Download PDF AI Translation

cs.CV / 34 / 2609.00704

FTU-Seek: Foundation Model-Guided Hard-Negative Learning for Sparse Functional Tissue Unit Segmentation

FTU-Seek：基础模型引导的稀疏功能组织单元分割中的困难负样本学习

Liu, Zonghao, Su, Lei, Yu, Jiguang, Geng, Xuqing, Wang, Louis Shuo, Wang, Jianmin, Liu, Jingfeng

Abstract

Functional tissue units (FTUs), including tertiary lymphoid structures (TLSs), blood vessels, and glands, encode localized immune, vascular, and epithelial organization in histopathology. Accurate quantification of these structures is important for studying tissue architecture and disease-associated tissue organization. However, FTUs are frequently sparse, heterogeneous, and surrounded by large amounts of morphologically similar background tissue, making automated segmentation in whole-slide images (WSIs) challenging. We therefore developed FTU-Seek, a pathology foundation model-guided framework that treats morphology-aware negative-patch selection as a key component of sparse FTU segmentation. FTU-Seek uses frozen multi-depth features from the UNI pathology foundation model to train a patch-level classifier that distinguishes FTU-containing from FTU-absent tissue. Target-absent patches are subsequently ranked according to their predicted target-containing probabilities, and the highest-scoring hard negatives are selected through a static Top$K$ strategy to construct compact segmentation training sets. The framework was evaluated using five-fold cross-validation and internal test cohorts across TLS, blood-vessel, and gland segmentation tasks, with an additional independent 30-WSI held-out cohort for TLS. Positive-only, all-tissue, random-negative, and matched random Top$K$ sampling strategies served as comparators. Segmentation-derived phenotypes were further explored in external TCGA cohorts.

Chinese Translation

功能组织单元（FTUs），包括三级淋巴结构（TLSs）、血管和腺体，在组织病理学中编码了局部的免疫、血管和上皮组织结构。这些结构的准确量化对于研究组织架构和与疾病相关的组织组织至关重要。然而，FTUs 通常稀疏、异质，并且被大量形态相似的背景组织所包围，这使得在全切片图像（WSIs）中进行自动分割变得具有挑战性。因此，我们开发了 FTU-Seek，这是一种基于病理基础模型的框架，将形态感知的负样本选择视为稀疏 FTU 分割的关键组成部分。FTU-Seek 利用来自 UNI 病理基础模型的冻结多深度特征，训练一个补丁级分类器，以区分含 FTU 的组织和不含 FTU 的组织。随后，根据其预测的目标含量概率对缺失目标的补丁进行排名，并通过静态 Top$K$ 策略选择得分最高的困难负样本，以构建紧凑的分割训练集。该框架通过五折交叉验证和内部测试队列评估，涵盖了 TLS、血管和腺体分割任务，并为 TLS 设立了额外的独立 30-WSI 保留队列。仅正样本、全组织、随机负样本和匹配随机 Top$K$ 采样策略作为比较。分割衍生的表型在外部 TCGA 队列中进一步探讨。

View on arXiv Download PDF AI Translation

cs.CV / 35 / 2609.00709

Controllable Image Captioning with Prompt-Conditioned Scene Rewards

基于提示条件场景奖励的可控图像描述生成

Hyun, Jongyeop, Kim, Taeyoung, Kim, Hyounghun

Abstract

Large Vision-Language Models produce fluent image descriptions but offer limited semantic control: users cannot reliably specify whether captions should emphasize attributes, relations, or particular image regions. We present Fine-grained Captioning Control Using Scene Rewards (FoCUS), a controllable image captioning method that lets users steer captions toward specific semantic emphases through natural-language control prompts. The core idea is a prompt-conditioned control objective based on scene-graph-aligned component scores. Generated captions are parsed and aligned to scene-graph components such as objects, attributes, and relations. These components are differentially weighted, including negative weights, according to the requested emphasis. We optimize this objective with GRPO and further improve its reliability through a stricter object validity threshold and reasoning-based verification for attribute and relation scoring. To evaluate controllability, we introduce Semantic Control and Precision Evaluation (SCoPE), a benchmark with contrastive Include/Avoid constraints for measuring both target content coverage and out-of-scope suppression. Experiments on two VLM backbones show that FoCUS consistently improves controllability and fine-grained caption quality without degrading general caption performance.

Chinese Translation

大型视觉语言模型能够生成流畅的图像描述，但在语义控制方面存在局限：用户无法可靠地指定描述应强调属性、关系或特定图像区域。我们提出了一种名为基于场景奖励的细粒度描述控制方法（FoCUS），该方法允许用户通过自然语言控制提示引导描述朝向特定的语义重点。其核心思想是基于场景图对齐的组件得分的提示条件控制目标。生成的描述被解析并与场景图组件（如对象、属性和关系）对齐。这些组件根据请求的强调程度被赋予不同的权重，包括负权重。我们使用GRPO优化该目标，并通过更严格的对象有效性阈值和基于推理的属性和关系评分验证进一步提高其可靠性。为了评估可控性，我们引入了语义控制与精度评估（SCoPE），这是一个具有对比性包含/避免约束的基准，用于测量目标内容覆盖和超出范围的抑制。对两个视觉语言模型骨干的实验表明，FoCUS在不降低一般描述性能的情况下，始终提高了可控性和细粒度描述质量。

View on arXiv Download PDF AI Translation

cs.CV / 36 / 2609.00712

EarthLD: Towards Unified Open-World Landslide Understanding via Vision-Language Guided Diffusion Models

EarthLD：通过视觉-语言引导的扩散模型实现统一的开放世界滑坡理解

Su, Yuanchao, Gao, Lianru, Jiang, Mengying, Chen, Jiangyi, Cheng, Jiaxin, Zhou, Yicong

Abstract

Landslides are widespread geological hazards, yet their automated detection and mapping in remote sensing imagery remain challenging because of their irregular morphology, ambiguous spectral signatures, and substantial domain shifts across imaging platforms. To overcome these challenges, we propose EarthLD, a vision-language-guided diffusion framework for open-world landslide understanding, enabling unified landslide recognition, mapping, and trigger interpretation. At its core, EarthLD formulates landslide understanding as a diffusion process that progressively infers the presence, spatial extent, and pixel-level boundaries of landslides from noisy latent representations. This probabilistic formulation enables the model to jointly perform image-level landslide recognition and mapping while characterizing predictive uncertainty. By integrating visual observations with contextual knowledge in the denoising process, EarthLD distinguishes diverse landslides from backgrounds, produces confidence-aware predictions for suspected regions, and maps landslide ranges. We additionally construct a global-scale open-world landslide benchmark by systematically harmonizing multiple publicly available remote sensing data collected by diverse institutions. Extensive experiments across regions, sensors, and triggering events demonstrate that EarthLD consistently outperforms existing landslide detection methods, highlighting its potential as a unified and robust solution for global geological-hazard monitoring and emergency response.

Chinese Translation

滑坡是一种广泛存在的地质灾害，但由于其不规则的形态、模糊的光谱特征以及不同成像平台之间的显著领域转移，滑坡在遥感影像中的自动检测和制图仍然面临挑战。为了解决这些问题，我们提出了EarthLD，这是一种用于开放世界滑坡理解的视觉-语言引导的扩散框架，能够实现滑坡的统一识别、制图和触发解释。EarthLD的核心是将滑坡理解形式化为一个扩散过程，该过程逐步从嘈杂的潜在表示中推断滑坡的存在、空间范围和像素级边界。这种概率性表述使模型能够同时进行图像级滑坡识别和制图，同时表征预测的不确定性。通过在去噪过程中将视觉观察与上下文知识相结合，EarthLD能够区分多样的滑坡与背景，为可疑区域生成置信度感知的预测，并绘制滑坡范围。此外，我们通过系统性地协调多个由不同机构收集的公开遥感数据，构建了一个全球范围的开放世界滑坡基准。针对不同地区、传感器和触发事件的广泛实验表明，EarthLD始终优于现有的滑坡检测方法，突显了其作为全球地质灾害监测和应急响应的统一且稳健解决方案的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 37 / 2609.00713

Efficient and Robust Absolute Pose Estimation via Gravity-Prior-Driven Transformation Decoupling and Pose Refinement

通过重力先验驱动的变换解耦和姿态优化实现高效且鲁棒的绝对姿态估计

Cao, Hu, Yang, Qianyi, Li, Xinyi, Liu, Jiong, Liu, Yinlong, Knoll, Alois

Abstract

Estimation of the absolute pose of an object is an essential task for various robotic applications. Recently, incorporating gravity direction as prior information has emerged as a popular approach to simplify absolute pose estimation. However, developing a robust and efficient algorithm to solve this challenging problem remains a difficult question due to large amounts of mismatches. In addition, obtaining an accurate pose solution from selected inlier correspondences with gravity prior is still a research gap. In this paper, we propose a novel transformation strategy that exploits geometric relations derived from the gravity prior. Through transformation decoupling, the original 6 degrees of freedom (DoF) absolute pose estimation problem is simplified into a 4-DoFs problem: 1-DoF for the rotation angle and 3-DoFs for translation, significantly improving the efficiency. For the 1-DoF rotation angle, we apply a one-dimensional global voting algorithm for optimal estimation. Once the optimal rotation is obtained, the mismatched correspondences are preliminarily filtered, and translation estimation, a linear problem, can be easily solved. Furthermore, to obtain accurate pose results, we introduce a novel pose refinement algorithm to enhance the accuracy of both rotation and translation. Extensive experiments on synthetic data and three publicly available real-world datasets (TUM RGB-D, ETH3D, and RobotCar) demonstrate that the proposed method achieves stronger performance compared to existing state-of-the-art (SOTA) approaches. To further validate our method, we integrated it into ORB-SLAM2. The results on the KITTI dataset show it effectively reduces drift and improves trajectory alignment during relocalization. The source code will be released upon acceptance.

Chinese Translation

物体绝对姿态的估计是各种机器人应用中的一项重要任务。最近，利用重力方向作为先验信息已成为简化绝对姿态估计的热门方法。然而，由于大量的匹配错误，开发一种鲁棒且高效的算法来解决这一挑战性问题仍然是一个难题。此外，从选定的内点对应关系中获得准确的姿态解仍然是一个研究空白。本文提出了一种新颖的变换策略，利用从重力先验推导的几何关系。通过变换解耦，原始的6自由度（DoF）绝对姿态估计问题被简化为一个4自由度的问题：1自由度用于旋转角度，3自由度用于平移，显著提高了效率。对于1自由度的旋转角度，我们应用了一种一维全局投票算法进行最优估计。一旦获得最优旋转，初步过滤掉不匹配的对应关系，平移估计作为一个线性问题可以轻松解决。此外，为了获得准确的姿态结果，我们引入了一种新颖的姿态优化算法，以提高旋转和平移的准确性。在合成数据和三个公开可用的真实世界数据集（TUM RGB-D、ETH3D和RobotCar）上的大量实验表明，所提出的方法相比现有的最先进（SOTA）方法具有更强的性能。为了进一步验证我们的方法，我们将其集成到ORB-SLAM2中。在KITTI数据集上的结果表明，它有效减少了漂移并改善了重定位过程中的轨迹对齐。源代码将在接受后发布。

View on arXiv Download PDF AI Translation

cs.CV / 38 / 2609.00730

Design and Implementation of a Kalman Filter-Infused Algorithm for Tilt Estimation

基于卡尔曼滤波的倾斜角度估计算法的设计与实现

Ma, Yuehan, Dai, Hongji

Abstract

Accurate tilt angle estimation is important in many engineering applications, such as robotics, motion tracking, and embedded control systems. However, measurements from low-cost inertial sensors are often degraded by noise and drift. This paper presents a single-axis tilt angle estimation system based on the MPU6050 inertial measurement unit, implemented on an RP2040 microcontroller platform, with sensor fusion achieved through a Kalman filter. The accelerometer provides a direct estimate of tilt angle from gravity but is sensitive to noise and short-term fluctuations. The gyroscope provides smooth angular rate measurements, but integration over time introduces drift. To overcome these limitations, a Kalman filter is used to combine measurements from both sensors, leveraging the long-term stability of the accelerometer and the short-term smoothness of the gyroscope. Both simulation and hardware experiments are performed. In simulation, sensor noise and drift are modeled to evaluate the filter performance under control conditions. In the hardware implementation, real-time MPU6050 data is acquired and processed by the RP2040 platform, and the estimated tilt angle is compared with accelerometer-only and gyroscope-only outputs. The results show that the proposed method effectively reduces noise measurements and suppresses long-term drift while preserving good dynamic response. Overall, the system provides more stable and accurate tilt estimation than either sensor alone, demonstrating a practical and accessible approach for Kalman filter based sensor fusion in embedded application.

Chinese Translation

准确的倾斜角度估计在许多工程应用中至关重要，例如机器人技术、运动跟踪和嵌入式控制系统。然而，低成本惯性传感器的测量数据常常受到噪声和漂移的影响。本文提出了一种基于MPU6050惯性测量单元的单轴倾斜角度估计系统，该系统在RP2040微控制器平台上实现，通过卡尔曼滤波实现传感器融合。加速度计提供了基于重力的倾斜角度的直接估计，但对噪声和短期波动敏感。陀螺仪提供平滑的角速度测量，但随时间积分会引入漂移。为克服这些限制，采用卡尔曼滤波器结合两种传感器的测量数据，利用加速度计的长期稳定性和陀螺仪的短期平滑性。进行了仿真和硬件实验。在仿真中，建模传感器噪声和漂移，以评估滤波器在控制条件下的性能。在硬件实现中，实时获取MPU6050数据并由RP2040平台处理，估计的倾斜角度与仅使用加速度计和仅使用陀螺仪的输出进行比较。结果表明，所提出的方法有效减少了噪声测量，抑制了长期漂移，同时保持了良好的动态响应。总体而言，该系统提供了比单一传感器更稳定和准确的倾斜估计，展示了一种在嵌入式应用中基于卡尔曼滤波的传感器融合的实用且可行的方法。

View on arXiv Download PDF AI Translation

cs.CV / 39 / 2609.00742

Mind the Rift: Cross-Scale Coupling Mismatch for AI-Generated Video Detection

注意裂缝：AI生成视频检测中的跨尺度耦合不匹配

Li, Siyu, Yang, Jin, Liang, Weiheng

Abstract

As AI video generators achieve cinematic realism, reliable detection becomes essential for safeguarding digital trust. We identify cross-scale coupling mismatch as a new forensic signal, where scale refers to the level of abstraction (semantic dynamics vs. pixel-level residuals): in natural videos, macro-level temporal dynamics and micro-level residual patterns are intrinsically coupled by the unified imaging physics pipeline, whereas AI generators, whose training objectives do not explicitly preserve this joint distribution, systematically violate this coupling. Detecting such mismatch is challenging because it requires independently extracting information at both scales while simultaneously quantifying their cross-scale relationship. We propose RIFT (Representation Inconsistency Forensics on Trajectories), an orthogonal forensic framework that addresses this through three interlocking components: a macro stream that builds a dynamic baseline of expected temporal evolution via differential geometry and persistent homology on learned manifold trajectories, a micro stream that acts as a sensitive forensic probe via steganalytic filtering and temporal modeling, and a coupling divergence module that measures the conditional dependency between the two streams. Gram-Schmidt orthogonality guarantees the information-theoretic validity of this measurement. Experiments on two benchmarks (VidProM, 120K videos, 7 generators; GenVidBench, 68K videos, 4 generators) demonstrate that RIFT achieves 99.33% and 99.72% F1-score respectively, with 97.87% unseen-generator detection rate in leave-one-out evaluation, while exhibiting encoder agnosticism: scaling from ViT-S/14 (22M) to ViT-L/14 (300M) changes F1 by less than 0.1%, and switching to a different encoder family (DINOv1) reduces F1 by only 0.73 pp. Code is available at https://github.com/Litsay/RIFT

Chinese Translation

随着AI视频生成器实现电影级的真实感，可靠的检测变得至关重要，以维护数字信任。我们将跨尺度耦合不匹配识别为一种新的取证信号，其中尺度指的是抽象层次（语义动态与像素级残差）：在自然视频中，宏观层次的时间动态与微观层次的残差模式通过统一的成像物理管道内在耦合，而AI生成器的训练目标并未明确保持这种联合分布，从而系统性地违反了这种耦合。检测这种不匹配具有挑战性，因为它需要独立提取两个尺度的信息，同时量化它们的跨尺度关系。我们提出了RIFT（Representation Inconsistency Forensics on Trajectories），一个正交的取证框架，通过三个相互关联的组件来解决这一问题：一个宏观流，通过对学习的流形轨迹进行微分几何和持久同调，建立预期时间演变的动态基线；一个微观流，通过隐写分析过滤和时间建模，充当敏感的取证探针；以及一个耦合发散模块，测量两个流之间的条件依赖性。Gram-Schmidt正交性保证了该测量的信息论有效性。在两个基准（VidProM，12万视频，7个生成器；GenVidBench，6.8万视频，4个生成器）上的实验表明，RIFT分别达到了99.33%和99.72%的F1分数，在留一法评估中，未见生成器的检测率为97.87%，同时表现出编码器的无关性：从ViT-S/14（2200万）扩展到ViT-L/14（3亿）时，F1变化小于0.1%，而切换到不同的编码器系列（DINOv1）仅减少0.73个百分点。代码可在https://github.com/Litsay/RIFT获取。

View on arXiv Download PDF AI Translation

cs.CV / 40 / 2609.00745

Feed-Forward Multi-view Multi-person Reconstruction with Contrastive Human-Aware 3D Representation

基于对比人类感知3D表示的前馈多视角多人物重建

Yang, Yuanwang, Huang, Buzhen, Ren, Zongxuan, Huang, Jing, Li, Kun

Abstract

Multi-view human reconstruction has been extensively studied under simplified settings, yet robust and efficient multi-person reconstruction in unconstrained environments remains challenging. Existing bottom-up methods often rely on accurate camera calibration and explicit cross-view matching, and therefore struggle with severe occlusions and ambiguities. We propose a new top-down paradigm that maintains a unified, instance-centric human-aware 3D space, enabling simultaneous camera calibration, cross-view association, and human reconstruction via cross-modal contrastive learning. Observations from multiple views are lifted and fused into this shared 3D space, where geometric structure, visual appearance, and human-centric semantic cues are jointly encoded at the instance level. We further introduce a spatial contrastive learning strategy that aligns 3D features corresponding to the same human instance across different views and modalities while separating different instances. This enables correspondence reasoning, semantic aggregation, and instance discrimination to be performed natively in 3D, improving cross-view consistency and robustness under severe occlusions. Finally, structured human body models are recovered in a feed-forward manner by regressing SMPL parameters from instance-level 3D human tokens. Extensive experiments demonstrate robust, accurate, and efficient multi-view human reconstruction in challenging real-world scenarios.

Chinese Translation

多视角人类重建在简化环境下得到了广泛研究，但在不受限环境中进行稳健且高效的多人物重建仍然具有挑战性。现有的自下而上方法往往依赖于准确的相机标定和明确的视图间匹配，因此在严重遮挡和歧义情况下表现不佳。我们提出了一种新的自上而下范式，维持一个统一的、以实例为中心的人类感知3D空间，能够通过跨模态对比学习实现相机的同时标定、视图间关联和人类重建。来自多个视角的观测被提升并融合到这个共享的3D空间中，在这里，几何结构、视觉外观和以人类为中心的语义线索在实例级别上共同编码。我们进一步引入了一种空间对比学习策略，该策略对齐不同视角和模态下对应同一人类实例的3D特征，同时区分不同实例。这使得在3D空间中本地进行对应推理、语义聚合和实例区分成为可能，从而提高了视图间的一致性和在严重遮挡下的稳健性。最后，通过从实例级别的3D人类标记回归SMPL参数，以前馈方式恢复结构化的人体模型。大量实验表明，在具有挑战性的真实场景中实现了稳健、准确和高效的多视角人类重建。

View on arXiv Download PDF AI Translation

cs.CV / 41 / 2609.00775

VOIM: Training-Free Open-Vocabulary 3D Instance Mapping for RGB-D and Monocular SLAM

VOIM：无训练开放词汇3D实例映射的RGB-D和单目SLAM

Song, Sangmin, Kodagoda, Sarath, Carmichael, Marc G., Thiyagarajan, Karthick, Gunatilake, Amal, Prentice, Kelly, Martin, Jodi

Abstract

We present Voxel-Grounded Online Instance Manager (VOIM), a training-free voxel-grounded instance manager that builds open-vocabulary 3D instance maps from RGB-D or from monocular RGB alone, a regime no prior training-free system addresses. Online systems typically segment object instances and label them at first detection, committing when evidence is weakest. VOIM instead defers label and instance decisions until soft evidence from unmodified, off-the-shelf perception has accumulated per voxel across views. We show that the mapping stage, rather than the particular perception models, carries the result: across four perception configurations on ScanNet++, varying the region descriptor, the detector label prior and the mask source, the map exceeds the strongest online RGB-D system, OVO-SLAM, by between 4.8 and 11.7 mIoU. Perception is not neutral, and substituting that baseline's own descriptor family costs 4.1 of the margin, yet the baseline carries the marginally better 2D descriptor (33.7 vs. 31.5 mIoU over three scenes) and still realizes the weaker map. Under a like-for-like protocol VOIM reaches 44.07 mIoU on ScanNet++ against 32.37, winning all ten scenes and both aggregations (pooled 33.31 vs. 25.97), and the same system runs unchanged to fully monocular RGB, matching that baseline pooled on Replica (27.80 vs. 27.50). The advantage is regime-specific: under Replica's all-classes scoring, matched inputs give a split result, 28.60 vs. 27.50 pooled against 24.59 vs. 30.11 on the per-scene mean. Room scale is label-limited and building scale drift-limited. Labeling does not run in real time, dominated by per-class detection over the full vocabulary. The maps export occupancy grids and resolve free-form queries to object instances.

Chinese Translation

我们提出了体素基础在线实例管理器（VOIM），这是一种无训练的体素基础实例管理器，可以从RGB-D或单目RGB单独构建开放词汇的3D实例地图，这是之前无训练系统未曾解决的领域。在线系统通常在首次检测时对物体实例进行分割和标记，在证据最薄弱时做出承诺。VOIM则推迟标签和实例决策，直到来自未修改的、现成的感知的软证据在每个体素上跨视图累积。我们展示了映射阶段，而非特定的感知模型，才是结果的关键：在ScanNet++上的四种感知配置中，改变区域描述符、检测器标签先验和掩膜来源，地图的mIoU超过了最强的在线RGB-D系统OVO-SLAM，提升幅度在4.8到11.7 mIoU之间。感知并非中立，替换该基线的描述符家族损失了4.1的优势，然而基线仍然拥有略微更好的2D描述符（在三个场景中分别为33.7与31.5 mIoU），但实现的地图却较弱。在相同协议下，VOIM在ScanNet++上达到了44.07 mIoU，而基线为32.37，赢得了所有十个场景和两种聚合方式（合并后为33.31对25.97），同一系统在完全单目RGB下运行不变，在Replica上与该基线相匹配（27.80对27.50）。这一优势是特定于模式的：在Replica的全类评分下，匹配输入给出了分裂结果，28.60对27.50合并，场景均值为24.59对30.11。房间规模受标签限制，构建规模受漂移限制。标记并不实时运行，主要受每类检测的影响，涵盖整个词汇。地图导出占用网格，并解决对物体实例的自由形式查询。

View on arXiv Download PDF AI Translation

cs.CV / 42 / 2609.00776

Solaris: Towards Interfaces That Are Generated, Not Coded

Solaris：朝着生成而非编码的接口迈进

Alaluf, Yuval, Avrahami, Omri, Leshem, Guy Bukchin, Geyer, Michal, Goldberg, Kfir, Richardson, Elad, Alarcón, Diego, Alvarez, Alejandro, Garry, Cole, Germanidis, Anastasis, Goldsen, Tenaya, Gurau, Corina, Kahlow, Robin, Kwartler, Joel, Lewis, Kathleen, Ortiz, Alejandro Matamala, McMahon, Eugene, Prom, Thon, Saltonstall-Wurm, Sarah, Umpherson, Jamie, Yeo, Hudson

Abstract

Digital interfaces are traditionally implemented through intermediate representations such as code, requiring their appearance and behavior to be specified in advance. We introduce Solaris, an interface world model that instead generates an interactive UI directly, frame by frame, in response to user actions. Solaris treats mouse interactions as conditioning signals and autoregressively synthesizes the resulting visual state at interactive speeds. To enable real-time generation while maintaining visual coherence over extended interactions, we combine autoregressive frame generation with few-step distillation and training on the model's own outputs. A language model complements the visual world model by interpreting user intent and specifying how interactions should affect the generated environment, separating high-level reasoning from visual rendering. By generating both the appearance and behavior of an interface dynamically, Solaris enables open-ended interactions that need not be explicitly programmed in advance. We view interface world models as a step toward a new paradigm for software, where interfaces are generated and adapted continuously around user intent rather than implemented as fixed collections of predefined states and

Chinese Translation

数字接口传统上通过中间表示（如代码）实现，这要求其外观和行为必须事先指定。我们介绍了Solaris，一个接口世界模型，它直接生成交互式用户界面，逐帧响应用户操作。Solaris将鼠标交互视为条件信号，并以自回归的方式以交互速度合成结果视觉状态。为了在保持视觉一致性的同时实现实时生成，我们将自回归帧生成与少步蒸馏和在模型自身输出上进行训练相结合。语言模型通过解释用户意图并指定交互如何影响生成环境，补充了视觉世界模型，从而将高层推理与视觉渲染分离。通过动态生成接口的外观和行为，Solaris实现了开放式交互，而无需事先明确编程。我们将接口世界模型视为软件新范式的一步，在这种范式中，接口是围绕用户意图持续生成和适应的，而不是作为固定的预定义状态集合实现的。

View on arXiv Download PDF AI Translation

cs.CV / 43 / 2609.00788

Forbid Your Attention: Fooling Multimodal Large Language Models by Selectively Removing Intrinsic Focus in Spectral Domain

禁止你的注意力：通过选择性去除频谱域内在焦点来欺骗多模态大型语言模型

Liu, Daizong, Dong, Junhao, Ma, Zhiyuan, Qu, Xiaoye, Fang, Xiang, Guan, Runwei, Tang, Keke, Dong, Jianfeng, Ong, Yew-Soon

Abstract

Multimodal large language models (MLLMs) have extended the capability of large language models (LLMs) to process more contextual multimodal information, showing remarkable progress in diverse realistic multimodal applications. Despite their strong perception and reasoning abilities, recent studies reveal that MLLMs remain highly vulnerable to adversarial inputs, especially those targeting visual components. However, existing attacks mainly focus on global perturbations, lacking an understanding of how MLLMs internally interpret visual structures. In this paper, we make the attempt to investigate the intrinsic focus of MLLMs in the frequency domain and discover that their predictions are particularly sensitive to phase information, which encodes essential structural and semantic cues. Based on this observation, we propose a novel phase-aware adversarial attack framework that explicitly restricts adversarial perturbations to structure-relevant phase regions to suppress the MLLMs' focus for effective and imperceptible attacks. To further amplify the structural influence, we also introduce an auxiliary adversarial prompt learning module to guide multimodal misalignment around phase-sensitive regions, misleading the MLLM's attention toward targeted structural patterns. Extensive experiments on multiple representative MLLM models and datasets demonstrate the superior effectiveness of our method compared to existing attacks.

Chinese Translation

多模态大型语言模型（MLLMs）扩展了大型语言模型（LLMs）处理更多上下文多模态信息的能力，在多种现实多模态应用中展现出显著进展。尽管它们具有强大的感知和推理能力，近期研究表明，MLLMs仍然对对抗性输入高度脆弱，尤其是针对视觉组件的输入。然而，现有攻击主要集中在全局扰动上，缺乏对MLLMs如何内部解释视觉结构的理解。本文尝试研究MLLMs在频域中的内在焦点，并发现其预测对相位信息特别敏感，而相位信息编码了重要的结构和语义线索。基于这一观察，我们提出了一种新颖的相位感知对抗攻击框架，明确将对抗扰动限制在与结构相关的相位区域，以抑制MLLMs的焦点，从而实现有效且不可察觉的攻击。为了进一步增强结构影响，我们还引入了一个辅助对抗提示学习模块，以指导多模态在相位敏感区域的错位，误导MLLM的注意力朝向目标结构模式。对多个代表性MLLM模型和数据集的广泛实验表明，我们的方法相比现有攻击具有更优越的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 44 / 2609.00798

Advanced Pixel Diffusion Model with Guided Sparse Global Refinement

带引导稀疏全局精炼的高级像素扩散模型

You, Weiyi, Zhang, Jinhua, Zhou, Xingyu, Long, Wei, Lou, Junyu, Gu, Shuhang

Abstract

Pixel-space diffusion has recently emerged as a promising direction for high-fidelity image generation by modeling images directly in the original pixel domain. However, pixel-space diffusion is computationally demanding due to the extremely high dimensionality of natural images. For efficiency, existing pixel diffusion models either compromise fine details with large-patch tokenization or confine subsequent refinement within individual patches. Such intra-patch refinement inevitably restricts structural continuity across patch boundaries and long-range token interactions, limiting refinement quality. To address these issues, we propose PixSGR, a novel Pixel diffusion framework with Sparse Global Refinement tailored for modeling the distribution of natural images directly in pixel space. PixSGR starts from a supervised low-channel bottleneck to efficiently capture the low-dimensional manifold of natural images. It then progressively expands the channel dimensionality and spatial resolution to recover increasingly fine-grained structures. At the spatial refinement stage, coarse-scale attention maps preselect globally relevant interactions to pre-sparsify fine-scale attention, enabling non-local refinement beyond isolated patches without the quadratic cost of dense attention. Extensive experiments on ImageNet validate the effectiveness of PixSGR. It achieves an FID of 1.51 at 256$\times$256 and maintains performance when scaled to 512$\times$512, attaining an FID of 1.60.

Chinese Translation

像素空间扩散最近作为高保真图像生成的一个有前景的方向出现，通过直接在原始像素域中建模图像。然而，由于自然图像的极高维度，像素空间扩散在计算上是非常耗费资源的。为了提高效率，现有的像素扩散模型要么通过大块标记化妥协细节，要么将后续的精炼限制在单个块内。这种块内精炼不可避免地限制了块边界之间的结构连续性和长距离标记交互，从而限制了精炼质量。为了解决这些问题，我们提出了PixSGR，一种新颖的像素扩散框架，采用稀疏全局精炼，旨在直接在像素空间中建模自然图像的分布。PixSGR从一个监督的低通道瓶颈开始，以高效捕捉自然图像的低维流形。然后，它逐步扩展通道维度和空间分辨率，以恢复越来越精细的结构。在空间精炼阶段，粗尺度注意力图预先选择全球相关的交互，以预稀疏细尺度注意力，从而实现超越孤立块的非局部精炼，而无需密集注意力的二次成本。在ImageNet上的大量实验验证了PixSGR的有效性。在256×256时，它达到了1.51的FID，并且在扩展到512×512时保持性能，达到了1.60的FID。

View on arXiv Download PDF AI Translation

cs.CV / 45 / 2609.00811

ReBridge-Flow: Re-Coupling Posterior Bridges in Flow Matching for Image Restoration

ReBridge-Flow：图像恢复中的流匹配后桥重耦合

Zhang, Jiaqi, Wang, Yiqi, Wu, Hongjie, Guo, Bohan, Wang, Xinan, Luo, Zichen, Cai, Taotao, Chen, Zhi, Zheng, Mingkai

Abstract

Flow Matching provides an efficient generative prior for image restoration by learning continuous transport between source and data distributions. However, existing methods typically incorporate measurement constraints through local corrections. Such corrections may disrupt the source-clean endpoint coupling implicitly encoded by the pretrained flow, making the corrected endpoint pair incompatible with the current state. To address this issue, we propose ReBridge-Flow, a posterior bridge re-coupling method. Specifically, given the current state, ReBridge-Flow first decodes the corresponding local source and clean endpoints. It then incorporates measurement information through clean-side anchoring and synchronously re-couples the source endpoint, yielding a measurement-aware endpoint pair with improved local bridge compatibility. The re-coupled endpoints further define a posterior-informed transport direction for advancing the sampling process. We also introduce the Posterior Bridge Defect, which jointly characterizes measurement error, deviation from the flow prior, and bridge mismatch, and leads to explicit updates for clean-side anchoring and source-side re-coupling. Extensive experiments on multiple natural and medical image restoration tasks demonstrate that ReBridge-Flow effectively alleviates bridge mismatch and improves the structural consistency of restored images.

Chinese Translation

流匹配通过学习源分布与数据分布之间的连续传输，为图像恢复提供了一种高效的生成先验。然而，现有方法通常通过局部修正来纳入测量约束。这种修正可能会破坏由预训练流隐式编码的源-清晰端点耦合，使得修正后的端点对与当前状态不兼容。为了解决这个问题，我们提出了ReBridge-Flow，一种后桥重耦合方法。具体而言，给定当前状态，ReBridge-Flow首先解码相应的局部源端点和清晰端点。然后，它通过清晰侧锚定纳入测量信息，并同步重新耦合源端点，从而产生一个具有改进局部桥兼容性的测量感知端点对。重耦合的端点进一步定义了一个后验信息驱动的传输方向，以推进采样过程。我们还引入了后桥缺陷（Posterior Bridge Defect），该缺陷共同表征测量误差、与流先验的偏差以及桥不匹配，并导致清晰侧锚定和源侧重耦合的显式更新。在多个自然和医学图像恢复任务上的广泛实验表明，ReBridge-Flow有效缓解了桥不匹配问题，并改善了恢复图像的结构一致性。

View on arXiv Download PDF AI Translation

cs.CV / 46 / 2609.00814

RingMoClaw: An Experience-Inspired Multi-Agent Framework for Self-Evolving Research in Remote Sensing

RingMoClaw：一种基于经验的多智能体自我演化框架用于遥感研究

Kang, Kaiyue, He, Qixuan, Wang, Peijin, Feng, Yingchao, Ren, Chao, Wang, Kangxin, Diao, Wenhui, Wang, Yixiao, Zhao, Liangjin, Wei, Kaiwen, Liu, Nayu, Sun, Xian

Abstract

Remote sensing visual models have continuously advanced various interpretation tasks. However, the research process behind model improvement still heavily relies on manual expertise, requiring extensive trial-and-error iterations in model design, data processing, and performance diagnosis. Existing agent-based approaches mainly focus on task execution and workflow orchestration, while lacking the capability of autonomous research iteration for continuous performance optimization. To address this issue, we propose RingMoClaw, an experience-inspired self-evolving multi-agent framework for remote sensing visual interpretation. RingMoClaw integrates a research branch, a quality-control branch, and a dual-stream dynamic experience bus to establish a closed-loop optimization process covering strategy generation, experiment execution, independent review, and experience accumulation. The heterogeneous Critic mechanism provides stage-wise diagnosis and feedback, while the dual-stream experience bus incorporates external knowledge and internal experimental experience to guide strategy evolution and eliminate ineffective searches. Extensive experiments on four remote sensing downstream tasks, including object detection, scene classification, semantic segmentation, and change detection, demonstrate the effectiveness and generalization of RingMoClaw. Compared with the corresponding baseline models, RingMoClaw improves performance by 1.84\% mAP$_{50}$ on object detection and achieves consistent gains across the other three tasks, while reducing the required evolution steps by over 40\% compared with existing research automation frameworks. These results suggest that RingMoClaw offers a feasible route from task execution toward continuous research driven model evolution in remote sensing.

Chinese Translation

遥感视觉模型不断推动各种解释任务的进展。然而，模型改进背后的研究过程仍然高度依赖人工专业知识，需要在模型设计、数据处理和性能诊断中进行大量的试错迭代。现有的基于智能体的方法主要集中在任务执行和工作流程编排上，缺乏自主研究迭代以实现持续性能优化的能力。为了解决这一问题，我们提出了RingMoClaw，一种基于经验的自我演化多智能体框架，用于遥感视觉解释。RingMoClaw集成了研究分支、质量控制分支和双流动态经验总线，以建立一个涵盖策略生成、实验执行、独立评审和经验积累的闭环优化过程。异构的Critic机制提供阶段性诊断和反馈，而双流经验总线则结合外部知识和内部实验经验，以指导策略演变并消除无效搜索。在四个遥感下游任务（包括目标检测、场景分类、语义分割和变化检测）上的大量实验表明，RingMoClaw的有效性和泛化能力。与相应的基线模型相比，RingMoClaw在目标检测上提高了1.84\%的mAP$_{50}$，并在其他三个任务中实现了一致的性能提升，同时与现有的研究自动化框架相比，所需的演化步骤减少了超过40\\%。这些结果表明，RingMoClaw为遥感领域从任务执行到持续研究驱动的模型演化提供了一条可行的路径。

View on arXiv Download PDF AI Translation

cs.CV / 47 / 2609.00816

Can Scene Text Recognition Read Rare Compositions?

场景文本识别能否读取稀有组合？

Zhang, Genpei

Abstract

Scene text recognition is reported as 89--97% accurate on the six standard benchmarks, and the problem is widely treated as saturated. We present an alternative reading. When the same test images are stratified jointly by ground-truth word rarity and character n-gram novelty against a reference corpus, accuracy at the rare-word x rare-trigram corner of the resulting 5x5 grid drops 10--18 pt below the q3/q3 centre across nine English specialised recognisers, and the same direction (corner below centre) holds on all 13 of 13 (language, model) pairs we test across four writing systems (Latin, Han, Han+kana, Arabic). The drop is not a capacity bottleneck. A 6x vision-backbone scale-up (CLIP4STR-Base 158M -> CLIP4STR-Huge 1.0B, OpenCLIP ViT-H/14 LAION-2B) leads every benchmark in aggregate accuracy yet leaves the stress corner unchanged (86.9 -> 86.5, within paired-bootstrap noise). Four converging probes--layer-wise probing, confidence-when-wrong, attention re-balancing, and a cross-script commit-vs-abstain error split--localise the failure to the autoregressive decoder's lexical prior. We then ask how much of the gap existing techniques recover. Of 16 non-architectural mitigations, the largest mean q5/q5 gain is +1.3 pt and none clears the paired-bootstrap noise floor; the only intervention that does is the architectural shift from autoregressive to CTC decoding (SVTRv2, +2.5 pt, p=0.02, n=474). A confidence-routed AR-CTC ensemble adds a directionally consistent +0.6 pt that stays within noise, and its dominant learned coefficient is each model's own minimum-softmax confidence--independently echoing the mechanism above. No configuration we test improves both the compositional corner and aggregate accuracy. The rare-input long tail thus points to architectural change rather than added capacity.

Chinese Translation

场景文本识别在六个标准基准测试中的准确率报告为89%至97%，该问题被广泛视为饱和。我们提出了一种替代的解读。当将相同的测试图像根据真实词稀有性和字符n-gram新颖性在参考语料库中进行联合分层时，结果5x5网格中稀有词与稀有三元组的交点处的准确率在九个英语专业识别器中下降了10至18个百分点，而在我们测试的四种书写系统（拉丁文、汉字、汉字+假名、阿拉伯文）中的13对（语言，模型）中，均保持相同的趋势（交点低于中心）。这一下降并不是能力瓶颈。将视觉主干网络规模扩大6倍（CLIP4STR-Base 158M -> CLIP4STR-Huge 1.0B，OpenCLIP ViT-H/14 LAION-2B）在总体准确率上领先于每个基准测试，但在压力交点处保持不变（86.9 -> 86.5，在配对自助法噪声范围内）。四个收敛探针——逐层探测、错误时的置信度、注意力重平衡和跨脚本的承诺与放弃错误分裂——将失败定位于自回归解码器的词汇先验。我们接着询问现有技术能恢复多少差距。在16种非架构性缓解措施中，最大的平均q5/q5增益为+1.3个百分点，且没有一种超过配对自助法的噪声底线；唯一能做到这一点的干预是从自回归到CTC解码的架构转变（SVTRv2，+2.5个百分点，p=0.02，n=474）。一个置信度引导的AR-CTC集成增加了一个方向一致的+0.6个百分点，且其主导学习系数是每个模型自身的最小软最大置信度——独立地回响上述机制。我们测试的没有任何配置同时改善了组合交点和总体准确率。因此，稀有输入的长尾指向的是架构变化而非增加能力。

View on arXiv Download PDF AI Translation

cs.CV / 48 / 2609.00830

Visual Attention Faithfulness in Vision-Language Models is Heterogeneous

视觉-语言模型中的视觉注意力可信度是异质的

Song, Xurui, Wang, Weishi, Yue, Zhongqi, Binici, Kuluhan, Bai, Tao, Shao, Hongxin, Dahlmeier, Daniel, Luo, Jun

Abstract

Whether attention weights faithfully reflect model reasoning has been actively debated in NLP, yet this question remains largely unexplored for the visual modality in Vision-Language Models (VLMs). We address this gap through causal perturbation analysis on current VLMs, evaluating both the comprehensiveness and sufficiency gap of attention-ranked visual tokens. Our analysis reveals that visual attention faithfulness is heterogeneous, manifesting in three distinct processing modes: Faithful-Sufficient, where top-$k$ attention tokens are both necessary and sufficient for prediction; Faithful-Distributed, where they are necessary but broader visual context remains required; and Non-Focal, where no localized attention region is individually necessary while visual information remains an essential trigger for prediction. Furthermore, human-annotated ground-truth regions satisfy comprehensiveness in only $\sim 60$% of cases compared with model attention rankings, revealing systematic divergence between model visual reliance and human intuition. We demonstrate these patterns across both general VQA on VQAv2 and document tasks on VRDU and ChartQA, showing that visual attention faithfulness varies systematically with processing demands and model architectures rather than being uniformly faithful or unfaithful.

Chinese Translation

在自然语言处理领域，关于注意力权重是否忠实反映模型推理的问题一直备受争议，但在视觉-语言模型（VLMs）的视觉模态中，这一问题仍然 largely 未被探索。我们通过对当前 VLMs 进行因果扰动分析来填补这一空白，评估注意力排名的视觉标记的全面性和充分性差距。我们的分析揭示了视觉注意力的可信度是异质的，表现为三种不同的处理模式：忠实-充分（Faithful-Sufficient），其中前 $k$ 个注意力标记既是预测所必需的，也是充分的；忠实-分布（Faithful-Distributed），其中它们是必要的，但仍需要更广泛的视觉上下文；以及非聚焦（Non-Focal），在这种情况下，没有单个的局部注意区域是必需的，而视觉信息仍然是预测的重要触发因素。此外，与模型注意力排名相比，人类标注的真实区域在约 $60 ext{ extperthousand}$ 的情况下满足全面性，揭示了模型视觉依赖与人类直觉之间的系统性偏差。我们在 VQAv2 的一般视觉问答（VQA）和 VRDU 及 ChartQA 的文档任务中展示了这些模式，表明视觉注意力的可信度随着处理需求和模型架构的变化而系统性变化，而不是统一地忠实或不忠实。

View on arXiv Download PDF AI Translation

cs.CV / 49 / 2609.00839

Residual Kalman Dynamics for Event-Based UAV Forecasting

基于残差卡尔曼动态的事件驱动无人机预测

Nyblom, Per, Ovrén, Hannes, Gustafsson, David

Abstract

We study short- and mid-horizon UAV bounding-box forecasting on the FRED event-camera dataset. We use a constant-velocity Kalman filter over a full center-size box state as a strong physical baseline, and train a residual model to predict acceleration-like corrections from recent box history, filtered state features, and local event representations. This simple residual formulation consistently improves over the Kalman baseline, with event-conditioned models giving the strongest results among the evaluated methods. We further show that part of the residual target is predictable from anchor position and velocity alone, indicating that canonical FRED results can reflect both visual evidence and dataset-specific motion priors. To analyze this effect, we introduce decorrelated subsets as a diagnostic stress test, showing that event-conditioned residual models retain useful predictive signal even when measured position- and velocity-based shortcuts are weakened.

Chinese Translation

我们研究了在FRED事件相机数据集上进行短期和中期无人机边界框预测。我们使用全中心大小框状态的恒速卡尔曼滤波器作为强有力的物理基线，并训练一个残差模型，以预测来自最近框历史、过滤状态特征和局部事件表示的加速度类修正。这种简单的残差形式在卡尔曼基线之上始终表现出改善，其中事件条件模型在评估的方法中表现出最强的结果。我们进一步表明，残差目标的一部分仅从锚点位置和速度就可以预测，这表明经典的FRED结果可以反映视觉证据和数据集特定的运动先验。为了分析这一效应，我们引入了去相关子集作为诊断压力测试，显示即使在测量的位置和速度基础捷径被削弱时，事件条件残差模型仍然保留有用的预测信号。

View on arXiv Download PDF AI Translation

cs.CV / 50 / 2609.00846

An Intelligent Decision Support System for Emotion Monitoring using Microscopic Fixational Dynamics

基于显微注视动态的情感监测智能决策支持系统

Shen, Xiangyu, Deng, Feiyang, Dai, Zijian, Chen, Aibin, Yi, Jizheng, Li, Jie, Jiang, Hongbo

Abstract

The rising prevalence of psychological disorders necessitates effective emotion monitoring, yet current methods relying on facial or physiological signals often suffer from intrusiveness and privacy issues. This paper proposes an intelligent decision support system and pervasive edge-computing framework that leverages smart glasses and a companion smartphone to infer emotional states from microscopic visual fixation patterns. Moving beyond traditional macroscopic gaze metrics, the proposed system extracts and decomposes three distinct neurophysiological micro-movements: microsaccades, ocular drifts, and ocular microtremors. We introduce an interpretable hybrid artificial intelligence pipeline combining a multi-head attention mechanism, extreme gradient boosting, and a support vector machine to extract deep temporal features, quantify their physiological importance, and perform efficient on-device classification. Through an extensive evaluation involving 60 volunteers, we rigorously validate the framework under a strict leave-one-subject-out cross-validation protocol across both controlled and naturalistic mobile scenarios. Ablation studies unequivocally demonstrate that these fixational micro-movements are substantially more discriminative for emotion inference than traditional macroscopic features. Furthermore, aligned with contemporary affective science, the system incorporates a few-shot personalization mechanism to bridge universal physiological baselines with individual emotional heterogeneity, achieving a highly robust personalized F1-score of 83.6%. This work establishes a physiologically interpretable, unobtrusive, and deployable paradigm for continuous real-time emotion monitoring.

Chinese Translation

心理障碍的日益普遍使得有效的情感监测变得必要，然而目前依赖面部或生理信号的方法往往存在侵入性和隐私问题。本文提出了一种智能决策支持系统和普适边缘计算框架，利用智能眼镜和配套智能手机从显微视觉注视模式中推断情感状态。该系统超越了传统的宏观注视指标，提取并分解三种不同的神经生理微运动：微眨眼（microsaccades）、眼球漂移（ocular drifts）和眼球微颤（ocular microtremors）。我们引入了一种可解释的混合人工智能管道，结合了多头注意力机制（multi-head attention mechanism）、极端梯度提升（extreme gradient boosting）和支持向量机（support vector machine），以提取深层时间特征、量化其生理重要性，并进行高效的设备内分类。通过对60名志愿者进行广泛评估，我们在严格的留一法交叉验证协议下，对该框架在受控和自然移动场景中进行了严谨验证。消融研究明确表明，这些注视微运动在情感推断中比传统的宏观特征具有更强的区分能力。此外，结合当代情感科学，该系统还引入了少样本个性化机制，以将普遍生理基线与个体情感异质性相结合，达到了83.6%的高度稳健个性化F1分数。这项工作建立了一种生理可解释的、非侵入性的、可部署的连续实时情感监测范式。

View on arXiv Download PDF AI Translation

cs.CV / 51 / 2609.00853

ADGNet: Asymmetric Dual-text Guided Network for Infrared Small Target Detection

ADGNet：用于红外小目标检测的非对称双文本引导网络

Wang, Tongtong, Xu, Mingzhu, Yu, Chenglong, Wang, Jing, Lin, Xiaohui, Guan, Weili

Abstract

InfRared Small Target Detection (IRSTD) is a challenging task. Relying solely on pixel-level information, vision-only methods struggle to distinguish targets from clutter. Current multimodal methods typically describe both targets and backgrounds with a single textual prompt. Such an approach lacks dedicated regional guidance and ignores infrared semantic asymmetry. Consequently, it provides insufficient background suppression information and introduces severe feature optimization conflicts, overwhelming small targets with noise. To address these issues, we propose a novel Asymmetric Dual-text Guided Network (ADGNet). Specifically, accounting for the infrared semantic asymmetry, we first design the Asymmetric Dual-text Prompt (ADP), comprising an image-agnostic abstract target prompt and an image-specific detailed background prompt. To leverage these prompts, we introduce an Asymmetric Dual-Branch Interaction (ADBI) module to separately guide visual features with their respective text priors, protecting targets from noise while suppressing background clutter. Subsequently, we introduce an Adaptive Feature Aggregation (AFA) module to dynamically fuse features from the two branches. Furthermore, we construct a multimodal Asymmetric Image-Text Infrared (AITIR) dataset by providing asymmetric text annotations for three public datasets (IRSTD-1K, NUDT-SIRST, and SIRST). Extensive experiments demonstrate that ADGNet outperforms 21 state-of-the-art (SOTA) methods. Code is available at https://github.com/iLearn-Lab/MM26-ADGNet.

Chinese Translation

红外小目标检测（IRSTD）是一项具有挑战性的任务。仅依赖于像素级信息的视觉方法在从杂波中区分目标方面面临困难。目前的多模态方法通常使用单一文本提示描述目标和背景。这种方法缺乏专门的区域引导，并忽视了红外语义的不对称性。因此，它提供的背景抑制信息不足，并引入严重的特征优化冲突，使小目标被噪声淹没。为了解决这些问题，我们提出了一种新颖的非对称双文本引导网络（ADGNet）。具体而言，考虑到红外语义的不对称性，我们首先设计了非对称双文本提示（ADP），包括一个与图像无关的抽象目标提示和一个特定于图像的详细背景提示。为了利用这些提示，我们引入了一个非对称双分支交互（ADBI）模块，分别用各自的文本先验引导视觉特征，保护目标免受噪声影响，同时抑制背景杂波。随后，我们引入了一个自适应特征聚合（AFA）模块，以动态融合两个分支的特征。此外，我们通过为三个公共数据集（IRSTD-1K、NUDT-SIRST 和 SIRST）提供不对称文本注释，构建了一个多模态非对称图像-文本红外（AITIR）数据集。大量实验表明，ADGNet在21种最先进（SOTA）方法中表现优越。代码可在 https://github.com/iLearn-Lab/MM26-ADGNet 获取。

View on arXiv Download PDF AI Translation

cs.CV / 52 / 2609.00862

Training-Free Inpainting Across Domains with a Frozen Text-to-Image Diffusion Model

基于冻结文本到图像扩散模型的无训练跨领域修复

Wang, Zhenhuan, Yuan, Fengyi

Abstract

We show that a frozen generic text-to-image diffusion model can perform conditional inpainting across three evaluated natural-image domains with one fixed controller configuration, without inpainting-specific weight training, dataset-specific weight adaptation, or learned inpainting-specific conditioning channels. Step-PI augments known-region projection with boundary-interior latent feedback, persistent PI state, and a predefined four-field release schedule that modulates controller signals along the reverse trajectory. Developed only on Main35-disjoint CelebA-HQ pilots, the controller transfers unchanged to AFHQ and Places2. Across two field-identical comparisons on the same 3,500 cases, adding persistent state and replacing uniform release with the predefined schedule each improve all 15 dataset-metric cells; 95% bootstrap intervals exclude zero for all five metrics in both comparisons. In descriptive native-route comparisons, Step-PI leads LanPaint and PILOT (the closest evaluated training-free baselines using vanilla SD1.5) on all five equal-dataset macro metrics. Inpainting-trained systems retain the absolute metric leads but rely on substantial inpainting-specific offline optimization. Our method provides a complementary approach for repurposing a frozen generic text-to-image model for cross-domain inpainting through test-time latent control.

Chinese Translation

我们展示了一个冻结的通用文本到图像扩散模型可以在三个评估的自然图像领域中执行条件修复，且只需一个固定的控制器配置，无需特定于修复的权重训练、特定于数据集的权重适应或学习的修复特定条件通道。Step-PI通过边界内部潜在反馈、持久的PI状态和预定义的四字段释放计划增强已知区域投影，该计划在反向轨迹上调节控制器信号。该控制器仅在Main35-不相交的CelebA-HQ试点上开发，且在AFHQ和Places2中保持不变。在对同一3,500个案例的两个领域相同的比较中，添加持久状态并用预定义计划替换均匀释放均改善了所有15个数据集度量单元；在两个比较中，95%的自助法区间排除了零。在描述性的本地路径比较中，Step-PI在所有五个相同数据集的宏观度量上超越了LanPaint和PILOT（最接近的评估无训练基线，使用普通的SD1.5）。经过修复训练的系统保持绝对度量领先，但依赖于大量特定于修复的离线优化。我们的方法为重新利用冻结的通用文本到图像模型进行跨领域修复提供了一种补充方法，通过测试时的潜在控制实现。

View on arXiv Download PDF AI Translation

cs.CV / 53 / 2609.00866

Benchmarking Vision-Language Models for Automated Pathology Diagnosis and Report Generation

自动化病理诊断与报告生成的视觉-语言模型基准测试

Lee, Yumi, Oh, Harim, Kim, Hyoryung, Kim, Minji, Kim, Eunsu, Lee, Hyeseong, Fukuoka, Junya, Bychkov, Andrey, Munkhdelger, Jijgee, Kaushal, Rajiv Kumar, Sahay, Ayushi, Yadav, Rajni, Prabakaran, Bharathi, Sarioglu, Sulen, Balcı, Serdar, Turkmen, Ilknur, Tolkach, Yuri, Harder, Christian, Westerdorf, Julian, Buettner, Reinhard, Henriksen, Audun Ljone, De Raedt, Sepp, Lee, Byung Hyun, Lim, Sungjin, Lee, Joohoon, Kim, Gwanghyun, Chun, Se Young, Singh, Suryakant, Kapse, Saarthak, Prasanna, Prateek, Kim, Kyung A, Kang, Yousun, Yoo, Sehwan, Hong, Sungman, Innani, Shubham, Feldman, Michael, Bakas, Spyridon, Baid, Ujjwal, Dutande, Prasad, Gajare, Suhas, Baheti, Bhakti, Sökmen, Serkan, Cebeci, Ece Tuğba, Halıcı, Ahmet, Balcı, Musa, Peçenek, Kardelen, Sainath, Srividhya, Jang, Kyongseok, Lee, Messi H. J., Wahab, Noorul, Du, Bodong, Zhang, Jiaming, Zhang, Qixiang, Choi, Jang-Hwan, Ahn, Sangjeong

Abstract

The rapid advancement of vision-language models (VLMs) has accelerated progress in computational pathology; however, whole-slide image (WSI)-based pathology report generation remains limited by the scarcity of large-scale WSI--report datasets and the complexity of mapping spatially distributed visual patterns to structured clinical text. To address this, we introduce a clinically curated Pan-Asia WSI--report dataset of approximately 10,500 pairs from five institutions and establish the REG 2025 benchmark through a MICCAI challenge for systematic evaluation of multimodal models. We analyze submitted methods spanning pretrained VLMs, multiple-instance learning frameworks, hierarchical expert models, retrieval-augmented generation, and cross-modal Transformers. Rather than indicating that VLM use alone was sufficient for superior performance, the results suggest that top-performing methods benefited from structured report representations, hierarchical diagnostic decomposition, and effective multimodal grounding. We identify key limitations, including instability in quantitative attribute estimation (e.g., numeric hallucination) and a tendency toward diagnostic overspecification, with some errors resembling known diagnostic pitfalls in routine pathology. These findings establish REG 2025 as a benchmark for evaluating WSI-based structured report generation and vision-language understanding in computational pathology, providing insights for the design of clinically grounded multimodal pathology models.

Chinese Translation

视觉-语言模型（VLMs）的快速发展加速了计算病理学的进步；然而，基于全切片图像（WSI）的病理报告生成仍受到大规模WSI-报告数据集稀缺和将空间分布的视觉模式映射到结构化临床文本的复杂性限制。为了解决这一问题，我们引入了一个临床策划的泛亚WSI-报告数据集，包含来自五个机构的大约10,500对数据，并通过MICCAI挑战建立了REG 2025基准，以系统评估多模态模型。我们分析了提交的方法，涵盖了预训练的VLMs、多实例学习框架、分层专家模型、检索增强生成和跨模态变换器。结果表明，顶尖表现的方法并非仅依赖于VLM的使用，而是受益于结构化报告表示、分层诊断分解和有效的多模态基础。我们识别出关键限制，包括定量属性估计的不稳定性（例如，数字幻觉）和对诊断过度具体化的倾向，某些错误类似于常规病理中的已知诊断陷阱。这些发现确立了REG 2025作为评估基于WSI的结构化报告生成和计算病理学中的视觉-语言理解的基准，为临床基础的多模态病理模型设计提供了见解。

View on arXiv Download PDF AI Translation

cs.CV / 54 / 2609.00868

The Visual Insensitivity Gap: Diagnosing When Vision-Language Models Fail to Use Visual Evidence

视觉无敏感性差距：诊断视觉-语言模型何时未能利用视觉证据

Zhang, Genpei

Abstract

Vision-language models are evaluated by aggregate accuracy on multimodal benchmarks, a practice that implicitly assumes the model uses its visual input. We show this assumption fails on 40%--97% of samples across six VLMs and three perceptual benchmarks: blurring the question-relevant visual region leaves the next-token distribution nearly unchanged. We name this phenomenon the Visual Insensitivity Gap and quantify it with a per-sample Visual Sensitivity Index (VSI). The gap is a property of samples, not of models: VSI ranks correlate across models (grand-mean Spearman rho=+0.40, permutation p<10^-3), so the same samples are flagged insensitive by VLMs sharing no architectural detail beyond a contrastively pretrained vision tower. The mechanism is concrete: on the insensitive samples, a linear probe on each model's own vision tower distinguishes perturbed from clean images at 0.72--0.79 accuracy, yet the model's argmax token changes on only 2%--11% of the same samples, an encoder--LLM gap above 0.65 on every model. Mapping VSI's diagnostic utility cell by cell surfaces a strong regime (multi-choice reasoning on capable VLMs: AUROC=0.85--0.87) and a weak regime (well-calibrated factuality, where softmax confidence already leads). VSI is not a universal best abstention signal; it is a sample-intrinsic indicator of vision-ignoring failure, best used as a conditional ensemble component.

Chinese Translation

视觉-语言模型通过在多模态基准上的整体准确性进行评估，这一做法隐含地假设模型使用其视觉输入。我们展示了这一假设在六个视觉-语言模型（VLMs）和三个感知基准的40%至97%的样本上失效：模糊与问题相关的视觉区域几乎不改变下一个标记的分布。我们将这一现象称为视觉无敏感性差距，并通过每个样本的视觉敏感性指数（Visual Sensitivity Index, VSI）对其进行量化。该差距是样本的特性，而非模型的特性：VSI 排名在不同模型之间相关（总体平均斯皮尔曼相关系数 rho=+0.40，置换 p<10^-3），因此相同的样本在没有共享建筑细节的 VLMs 中被标记为无敏感性，除了对比预训练的视觉塔。其机制是具体的：在无敏感样本上，每个模型自身的视觉塔上的线性探针以0.72至0.79的准确率区分扰动图像与干净图像，但模型的 argmax 标记仅在2%至11%的相同样本上发生变化，在每个模型上都存在超过0.65的编码器-LLM差距。逐细胞映射 VSI 的诊断效用显示出强的范畴（在有能力的 VLMs 上进行多选推理：AUROC=0.85至0.87）和弱的范畴（良好校准的事实性，其中 softmax 信心已经领先）。VSI 不是一个通用的最佳弃权信号；它是一个样本内在的视觉忽视失败的指示，最好作为条件集成组件使用。

View on arXiv Download PDF AI Translation

cs.CV / 55 / 2609.00898

Vision-Language-Guided Pseudo-Labels for Unsupervised Domain Adaptation in Semantic Segmentation for Waste Sorting

基于视觉-语言引导的伪标签在废物分类语义分割中的无监督领域适应

Schlegel, Udo, Shubhangi, Dax, Gabriel, Kaminwar, Sai Rahul, Karl, Florian, Seidl, Thomas

Abstract

Obtaining labeled data for semantic segmentation in applied settings (e.g., autonomous driving, industrial waste sorting) is expensive and often infeasible at scale. We present a cross-modal pseudo-labeling pipeline that enables unsupervised domain adaptation without any target-domain annotations. The pipeline is built on two core foundation models: SAM generates class-agnostic region proposals, and EVA-CLIP assigns semantic labels based on region-text similarity, with confidence filtering ensuring that only reliable pseudo-labels are used for self-training a segmentation model. As an optional extension, BLIP provides language-grounded verification for ambiguous regions, thereby improving pseudo-label quality without altering the overall pipeline. Evaluated on two domain shifts, synthetic-to-real autonomous driving and, with a primary focus, lab-to-factory industrial waste sorting, the pipeline consistently improves over source-only baselines. Our results demonstrate that pseudo-label quality, not quantity, is a decisive factor in self-training under domain shift, and that cross-modal language grounding offers a practical path to reliable automatic annotation in deployment-critical applications.

Chinese Translation

在应用场景中（例如，自动驾驶、工业废物分类）获取语义分割的标注数据既昂贵又往往难以大规模实现。我们提出了一种跨模态伪标签生成管道，使得在没有任何目标领域注释的情况下实现无监督领域适应。该管道基于两个核心基础模型构建：SAM生成与类别无关的区域提议，EVA-CLIP根据区域与文本的相似性分配语义标签，并通过置信度过滤确保仅使用可靠的伪标签来进行分割模型的自我训练。作为可选扩展，BLIP提供了对模糊区域的语言基础验证，从而提高伪标签的质量而不改变整体管道。我们在两个领域迁移上进行了评估，分别是合成到真实的自动驾驶，以及主要关注的实验室到工厂的工业废物分类，该管道在源域基线之上始终表现出改进。我们的结果表明，在领域迁移下，伪标签的质量而非数量是自我训练的决定性因素，而跨模态语言基础提供了一条在关键应用中实现可靠自动注释的实用路径。

View on arXiv Download PDF AI Translation

cs.CV / 56 / 2609.00901

HELIOS: From midnight to noon, continuous outdoor urban scene relighting

HELIOS：从午夜到正午，连续的户外城市场景重光照

Djeghim, Hala, Piasco, Nathan, Roldão, Luis, Bennehar, Moussab, Tsishkou, Dzmitry, Loscos, Céline, Sidibé, Désiré

Abstract

Modifying the illumination of driving images is a fundamental challenge, as most datasets are captured at specific times of day. Existing methods rely on synthetic data or paired multi-illumination supervision, which limits their generalization to the diverse and challenging conditions of real-world scenarios. To address this, we propose HELIOS, a novel image relighting approach that relies on unlabeled real-world datasets without requiring any paired images for training. Our approach integrates albedo-based conditioning into a cycle-consistent diffusion pipeline to prevent identity collapse and ensure accurate domain translation. To handle low-visibility nighttime conditions, we introduce a robust albedo distillation strategy that transfers structural stability from the daytime domain. Additionally, we replace traditional text prompts with a fine-grained control mechanism based on GPS-derived solar angles, enabling smooth and continuous lighting manipulation across the day-night cycle. Through extensive evaluation and a user study, we demonstrate that HELIOS produces structurally consistent and realistic results in both night-to-day and day-to-night tasks, outperforming state-of-the-art methods.

Chinese Translation

修改驾驶图像的照明是一个基本挑战，因为大多数数据集是在特定的时间段捕获的。现有方法依赖于合成数据或配对的多照明监督，这限制了它们在真实场景中多样且具有挑战性的条件下的泛化能力。为了解决这个问题，我们提出了HELIOS，一种新颖的图像重光照方法，该方法依赖于未标记的真实世界数据集，而无需任何配对图像进行训练。我们的方法将基于反照率的条件集成到一个循环一致的扩散管道中，以防止身份崩溃并确保准确的领域转换。为了处理低能见度的夜间条件，我们引入了一种稳健的反照率蒸馏策略，从白天领域转移结构稳定性。此外，我们用基于GPS导出的太阳角度的细粒度控制机制替换了传统的文本提示，使得在昼夜循环中实现平滑和连续的照明操控。通过广泛的评估和用户研究，我们证明HELIOS在夜到日和日到夜的任务中产生了结构一致且逼真的结果，超越了最先进的方法。

View on arXiv Download PDF AI Translation

cs.CV / 57 / 2609.00909

A multicenter benchmark and clinically structured metric for coronary CTA report generation

冠状动脉CTA报告生成的多中心基准和临床结构化指标

Ye, Zhiyu, Sun, Yue, Zou, Limiao, Xu, Cheng, Xu, Keting, Hu, Tong, Yu, Yue, Zheng, Hairong, Wang, Yining, Zhang, Tong

Abstract

Reliable evaluation of automated coronary computed tomography angiography (CCTA) report generation requires standardized multicentre benchmarks and clinically structured metrics. We established a four-centre benchmark comprising 3,021 CCTA series from 818 patient-report pairs to evaluate seven open-source three-dimensional vision-language models. We developed CSM$_{\text{CCTA}}$, a clinically structured metric for CCTA report evaluation, with patient-, vessel-, and segment-level variables defined according to clinical guidelines. Report pairs are compared at the finest shared anatomical level, and the contributions of different clinical components are weighted based on expert assessments. We estimated these weights using 70 expert-scored cases and evaluated clinical alignment in a non-overlapping set of 30 cases. CSM$_{\text{CCTA}}$ showed a strong correlation with radiologist scores (Pearson's $r=0.97$, $p<0.001$), exceeding the next-best metric, FORTE ($r=0.70$), by 0.27, and agreed with expert preferences in 115 of 160 pairwise comparisons (71.9\%). Under controlled perturbations, CSM$_{\text{CCTA}}$ remained stable to clinically equivalent wording and decreased monotonically with progressive information omission. In the multicenter benchmark, the CCTA-trained C2RG model achieved the highest CSM$_{\text{CCTA}}$ scores across all four hospitals, although its performance remained far from optimal. In contrast, CCTA-irrelevant reports accounted for up to 98.7\% of the outputs from generalist models. Together, the benchmark provides a standardized setting for model comparison, while CSM$_{\text{CCTA}}$ enables clinically structured evaluation of finding agreement and anatomical specificity. These results support a more clinically aligned and anatomically resolved approach to evaluating CCTA report generation. Code is available at https://openi.pcl.ac.cn/OpenMedIA/CSM_CCTA.

Chinese Translation

对自动化冠状动脉计算机断层扫描血管造影（CCTA）报告生成的可靠评估需要标准化的多中心基准和临床结构化指标。我们建立了一个包含来自818名患者报告对的3,021个CCTA系列的四中心基准，以评估七个开源三维视觉-语言模型。我们开发了CSM$_{ ext{CCTA}}$，这是一个用于CCTA报告评估的临床结构化指标，患者、血管和分段级变量根据临床指南定义。报告对在最细的共享解剖层面进行比较，不同临床组件的贡献根据专家评估进行加权。我们使用70个专家评分的案例估计了这些权重，并在一个不重叠的30个案例集上评估了临床一致性。CSM$_{ ext{CCTA}}$与放射科医生评分显示出强相关性（Pearson's $r=0.97$, $p<0.001$），超出下一个最佳指标FORTE（$r=0.70$）0.27，并在160个成对比较中与专家偏好一致115次（71.9%）。在受控扰动下，CSM$_{ ext{CCTA}}$对临床等效措辞保持稳定，并随着信息逐渐缺失而单调下降。在多中心基准中，经过CCTA训练的C2RG模型在所有四家医院中获得了最高的CSM$_{ ext{CCTA}}$分数，尽管其性能仍远未达到最佳。相比之下，CCTA无关的报告占一般模型输出的高达98.7%。综上所述，该基准为模型比较提供了标准化的环境，而CSM$_{ ext{CCTA}}$则使得对发现一致性和解剖特异性的临床结构化评估成为可能。这些结果支持了一种更具临床一致性和解剖解析的方法来评估CCTA报告生成。代码可在 https://openi.pcl.ac.cn/OpenMedIA/CSM_CCTA 获取。

View on arXiv Download PDF AI Translation

cs.CV / 58 / 2609.00923

On-the-Fly3R: Towards Robust Online 3D Reconstruction with Feed-Forward 3R Models for Large-Scale UAV Scenarios

On-the-Fly3R：面向大规模无人机场景的鲁棒在线3D重建与前馈3R模型

Shen, Zhe, Lou, Liyuan, Yu, Yifei, Wang, Guanbo, Ji, Quanjian, Wang, Xin, Zhan, Zongqian

Abstract

While feed-forward 3D reconstruction (3R) offers efficient end-to-end modeling, its application in large-scale UAV mapping is hindered by the prohibitive memory cost of Transformer attention. Current scalable streaming 3R methods assume temporally and spatially continuous inputs, rendering them ineffective for the weakly ordered or unordered image streams common in cross-strip UAV operations. To address this, we propose On-the-Fly3R, a training-free, progressive online 3D reconstruction framework for large-scale UAV images that upgrades various 3R backbones for large-scale UAV scenarios. Our method enables reconstruction from unordered inputs via retrieval-guided dynamic subset construction, which adaptively selects spatially relevant images. To further improve the robustness, a validation-rejection-retry mechanism is designed to guarantee global consistency, performing a pre-integration consistency check and automatically rejecting misaligned images and retrying with alternative subset. Finally, inspired by VSLAM, pose graph optimization based on the retrieval loop closure is employed to mitigate camera drift. Evaluations on several UAV benchmarks show that our On-the-Fly3R successfully scales various 3R models to over 5,000 images across square-kilometer UAV scenes, delivering substantially superior accuracy compared to several SOTA streaming 3R methods. Code is available at https://github.com/Sh1nZzz/On_the_Fly3R

Chinese Translation

尽管前馈3D重建（3R）提供了高效的端到端建模，但其在大规模无人机映射中的应用受到Transformer注意力机制高昂内存成本的限制。目前的可扩展流式3R方法假设输入在时间和空间上是连续的，这使得它们在跨条带无人机操作中常见的弱序或无序图像流中效果不佳。为了解决这一问题，我们提出了On-the-Fly3R，这是一种无训练的渐进式在线3D重建框架，旨在处理大规模无人机图像，并对各种3R骨干网络进行了升级，以适应大规模无人机场景。我们的方法通过检索引导的动态子集构建，使得能够从无序输入中进行重建，适应性地选择空间相关的图像。为了进一步提高鲁棒性，我们设计了一种验证-拒绝-重试机制，以确保全局一致性，执行预集成一致性检查，自动拒绝不对齐的图像，并使用替代子集进行重试。最后，受到VSLAM的启发，基于检索回路闭合的位姿图优化被采用，以减轻相机漂移。在多个无人机基准测试中的评估表明，我们的On-the-Fly3R成功地将各种3R模型扩展到超过5,000幅图像，覆盖平方公里级的无人机场景，提供了显著优于多种最先进流式3R方法的准确性。代码可在 https://github.com/Sh1nZzz/On_the_Fly3R 获取。

View on arXiv Download PDF AI Translation

cs.CV / 59 / 2609.00924

Beyond the Image Plane: World-Grounded Queries for Multi-Object Tracking

超越图像平面：面向多目标跟踪的世界基础查询

Cetintas, Orcun, Brasó, Guillem, Meinhardt, Tim, Leal-Taixé, Laura

Abstract

Monocular videos record 3D scenes as sequences of 2D image-plane projections, obscuring depth and spatial relationships. Multi-object trackers localize and associate objects primarily using appearance and geometry observed only in the image plane, inheriting these ambiguities. To address this limitation, we introduce PLANET, an end-to-end multi-object tracker designed to move beyond the image plane. As an enabling step, we lift existing 2D tracking datasets into 3D. We then form world-grounded queries by embedding reconstructed 3D scene geometry into the features and positional encodings used during query formation. An auxiliary 3D location prediction task further encourages the queries to encode object positions during training. A complementary dual-resolution temporal memory preserves this evidence across longer temporal gaps. As a result, PLANET achieves state-of-the-art performance across three diverse benchmarks.

Chinese Translation

单目视频将三维场景记录为二维图像平面投影的序列，遮蔽了深度和空间关系。多目标跟踪器主要通过图像平面中观察到的外观和几何信息来定位和关联对象，从而继承了这些模糊性。为了解决这一局限性，我们提出了PLANET，一个旨在超越图像平面的端到端多目标跟踪器。作为一项基础步骤，我们将现有的二维跟踪数据集提升至三维。然后，我们通过将重建的三维场景几何嵌入到查询形成过程中使用的特征和位置编码中，形成世界基础查询。一项辅助的三维位置预测任务进一步鼓励查询在训练过程中编码对象位置。一个互补的双分辨率时间记忆在更长的时间间隔内保留这些证据。因此，PLANET在三个不同的基准测试中实现了最先进的性能。

View on arXiv Download PDF AI Translation

cs.CV / 60 / 2609.00948

From Terminology to Diagrams: Visual-Instruction Generation for Scientific Diagram Understanding

从术语到图示：科学图示理解的视觉指令生成

Ortega, Raul, Gómez-Pérez, José Manuel

Abstract

Vision-language models (VLMs) have demonstrated strong performance in visual question answering with natural images. However, they continue to struggle with scientific diagrams, which are designed to convey functional or relational meaning rather than literal scenes. We therefore introduce a framework for generating large-scale diagram-grounded instruction data by leveraging terminology derived from scientific curricula. Our approach systematically extracts domain concepts, synthesizes atomic facts, retrieves relevant diagrams from the web, and generates multimodal supervision in the form of diagram captions and multiple-choice questions. Using this pipeline, we construct SciGram, a dataset of over 194K diagrams and 1.4M visual instructions across life, earth, and physical sciences. Despite relying on noisy web data and synthetic annotations, models fine-tuned on SciGram achieve substantial improvements on diagram-centric benchmarks, including TQA, ScienceQA, and AI2D, outperforming or matching state-of-the-art VLMs while using fewer training instances. Furthermore, augmenting existing models such as LLaVA OneVision with SciGram establishes new state-of-the-art performance on diagram question answering. Our results highlight the effectiveness of terminology-grounded instruction generation as a general strategy for improving vision-language reasoning in scientific domains. To support future research in scientific diagram understanding, we release both the SciGram dataset and models.

Chinese Translation

视觉语言模型（VLMs）在自然图像的视觉问答中表现出色。然而，它们在科学图示方面仍然面临挑战，因为科学图示旨在传达功能或关系意义，而非字面场景。因此，我们提出了一种框架，通过利用源自科学课程的术语，生成大规模的图示基础指令数据。我们的方法系统地提取领域概念，合成原子事实，从网络中检索相关图示，并生成多模态监督，包括图示说明和多项选择题。通过这一流程，我们构建了SciGram，一个包含超过194K个图示和140万条视觉指令的数据集，涵盖生命科学、地球科学和物理科学。尽管依赖于噪声较大的网络数据和合成注释，但在SciGram上微调的模型在以图示为中心的基准测试（包括TQA、ScienceQA和AI2D）上取得了显著的提升，超越或匹配了最先进的VLMs，同时使用了更少的训练实例。此外，将现有模型如LLaVA OneVision与SciGram结合，创造了图示问答的新最先进性能。我们的结果突显了基于术语的指令生成作为改善科学领域视觉语言推理的通用策略的有效性。为了支持未来在科学图示理解方面的研究，我们发布了SciGram数据集和模型。

View on arXiv Download PDF AI Translation

cs.CV / 61 / 2609.00951

CERF: Communication-Efficient and Retraining-Free Collaborative Perception

CERF：一种通信高效且无需重训练的协作感知框架

Hao, Jiuwu, Ni, Ziyi, Sun, Liguo, Wan, Yuting, Wu, Yueyang, Xiang, Ti, Song, Haolin, Lv, Pin

Abstract

Collaborative perception shares information among multiple agents to obtain a comprehensive scene representation, enhancing the perceptual capability of individual agents. However, most existing methods rely on transmitting and fusing dense feature maps for collaboration, which incurs inevitable communication overhead and heterogeneity challenges, limiting their practicality for real-world deployment. To address these challenges, we propose CERF, a novel Communication-Efficient and Retraining-Free framework for open heterogeneous collaborative perception. In CERF, we introduce a new virtual modality (termed Poture), which is generated from the perception outputs of other agents, to augment the extracted Bird's Eye View (BEV) features of the ego agent. To mitigate transmission delays, we employ a Kalman-filter based tracker and a motion forecasting model to derive the current predictions from historical perception results. Extensive experiments demonstrate that CERF achieves performance comparable to mainstream intermediate-collaboration methods while reducing communication overhead by 95% across various downstream tasks. Furthermore, CERF enables seamless integration of unknown heterogeneous agents into the existing collaborative framework without additional retraining costs. Code is available at https://github.com/uestchjw/CERF.

Chinese Translation

协作感知在多个智能体之间共享信息，以获得全面的场景表示，从而增强个体智能体的感知能力。然而，大多数现有方法依赖于传输和融合密集特征图进行协作，这不可避免地带来了通信开销和异构性挑战，限制了其在实际应用中的可行性。为了解决这些挑战，我们提出了CERF，一种新颖的通信高效且无需重训练的开放异构协作感知框架。在CERF中，我们引入了一种新的虚拟模态（称为Poture），该模态由其他智能体的感知输出生成，用于增强自我智能体提取的鸟瞰图（BEV）特征。为了减轻传输延迟，我们采用基于卡尔曼滤波的跟踪器和运动预测模型，从历史感知结果中推导当前预测。大量实验表明，CERF在各种下游任务中实现了与主流中间协作方法相当的性能，同时将通信开销减少了95%。此外，CERF能够无缝集成未知的异构智能体到现有的协作框架中，而无需额外的重训练成本。代码可在 https://github.com/uestchjw/CERF 获取。

View on arXiv Download PDF AI Translation

cs.CV / 62 / 2609.00955

ASSERT: Adaptive Stochastic Sampling for Robust Diffusion Models on Analog Compute-in-Memory Hardware

ASSERT：用于模拟计算内存硬件上鲁棒扩散模型的自适应随机采样

Feng, Yuannuo, Chen, Yizhe, Yao, Wenshuai, Xie, Yuxin, Wong, Ngai, Zhou, Wenyong, Kang, Wang

Abstract

Diffusion models achieve strong image generation quality but incur high iterative denoising costs. Analog compute-in-memory (CIM) can accelerate matrix-vector multiplications, yet spatial memory variations perturb weights and accumulate during sampling. Unlike conventional neural networks, diffusion models' temporal sensitivity to hardware noise remains underexplored. We investigate diffusion inference using a noise model calibrated and validated against measurements collected from multiple physical CIM chips. Our results show that the early, high-noise denoising stage is substantially more vulnerable than the final refinement stage. A first-order trajectory analysis attributes this behavior to the repeated propagation of correlated prediction errors induced by a fixed hardware mapping. Based on this observation, we propose ASSERT, a training-free sampler that uses higher stochasticity early and smoothly transitions to deterministic denoising. The injected stochasticity changes subsequent activation trajectories and thereby reduces their alignment with persistent spatial errors. Across the evaluated settings, ASSERT achieves up to 2.58$\times$ lower FID than deterministic DDIM on high-resolution datasets and 7.68$\times$ lower FID in the CIFAR-10 step-count study, without changing model parameters or the number of network evaluations.

Chinese Translation

扩散模型在图像生成质量上表现出色，但需要高昂的迭代去噪成本。模拟计算内存（CIM）可以加速矩阵-向量乘法，但空间内存的变化会扰动权重并在采样过程中累积。与传统神经网络不同，扩散模型对硬件噪声的时间敏感性尚未得到充分研究。我们使用经过校准和验证的噪声模型，研究了扩散推理，该模型是基于从多个物理CIM芯片收集的测量数据。我们的结果表明，早期的高噪声去噪阶段比最终的精细化阶段更为脆弱。一阶轨迹分析将这种行为归因于由固定硬件映射引起的相关预测误差的重复传播。基于这一观察，我们提出了ASSERT，一种无训练的采样器，早期使用更高的随机性，并平滑过渡到确定性去噪。注入的随机性改变了后续的激活轨迹，从而减少了它们与持续空间误差的对齐。在评估的设置中，ASSERT在高分辨率数据集上实现了比确定性DDIM低2.58倍的FID，并在CIFAR-10步数研究中实现了低7.68倍的FID，而无需更改模型参数或网络评估次数。

View on arXiv Download PDF AI Translation

cs.CV / 63 / 2609.00956

PredErase: Training-Free Object-and-Effect Removal with Predictive Latent Guidance

PredErase：无训练的对象与效果去除方法，基于预测潜在引导

Xiu, Waikit, Lu, Qiang, Chen, Junbiao, Li, Xiying

Abstract

Removing an object is not the same as filling its mask. Cast shadows and contact shading usually lie outside the user-provided instance mask M_obj, so a frozen Fill model that edits only that mask leaves the object's photometric footprint on nearby surfaces. Supervised removers learn this joint erasure from paired clean plates. Training-free editors freeze pretrained weights, yet most still treat M_obj as the entire editable support and steer sampling with CLIP or DINO energies that do not predict the occluded scene. We present PredErase, a training-free inference procedure on frozen FLUX.2 and I-JEPA. The method separates where Fill may rewrite pixels from what structure should occupy the hole. A contact-band expansion M_flux of M_obj exposes local residuals on the supporting plane. I-JEPA, pretrained for masked token prediction, supplies a context-conditioned hole target in representation space; sparse projected gradients align decoded Fill completions with that target inside the instance, while coordinates outside the packed support stay locked. Under instance-only masks on RemovalBench, RORD-Val, and DEFACTO-Val, PredErase improves the native FLUX.2 backbone. Supervised removers remain stronger on several full-image appearance metrics; the supported claim is training-free object-and-effect editing of frozen Fill, not replacement of paired-data erasers.

Chinese Translation

去除一个对象并不等同于填充其遮罩。投影阴影和接触阴影通常位于用户提供的实例遮罩 M_obj 之外，因此仅编辑该遮罩的冻结填充模型会在附近表面留下对象的光度足迹。监督去除器通过配对的干净图像学习这种联合去除。无训练的编辑器冻结预训练权重，但大多数仍将 M_obj 视为整个可编辑支持，并利用 CLIP 或 DINO 能量进行采样，这些能量无法预测被遮挡的场景。我们提出了 PredErase，这是一种在冻结的 FLUX.2 和 I-JEPA 上进行的无训练推理过程。该方法将填充可能重写像素的区域与应占据空洞的结构分开。接触带扩展 M_flux 使 M_obj 暴露出支撑平面上的局部残差。I-JEPA 经过预训练以进行遮罩标记预测，提供了在表示空间中条件化的空洞目标；稀疏投影梯度将解码的填充完成与实例内的该目标对齐，而打包支持外的坐标保持锁定。在 RemovalBench、RORD-Val 和 DEFACTO-Val 上的仅实例遮罩下，PredErase 改进了原生的 FLUX.2 主干。虽然在多个全图外观指标上，监督去除器仍然更强，但支持的主张是无训练的冻结填充对象与效果编辑，而不是替代配对数据去除器。

View on arXiv Download PDF AI Translation

cs.CV / 64 / 2609.00959

Candidate-Expanding Routing with Permutation-Stabilized Experts for Mixed-Format Medical VQA

具有排列稳定专家的候选扩展路由用于混合格式医学视觉问答

Nguyen, Hai-Dang, Pham, Huy-Hieu

Abstract

Mixed-format medical visual question answering (VQA) requires stable option selection and machine-readable free-text output. The two formats fail differently: multiple-choice predictions can change with option symbols or positions, while clinically plausible open answers can fail automated evaluation when serialization is malformed. We address both challenges with an answer-text memory, a permutation-stabilized vision--language expert, and a sparse candidate- expanding router. The cyclic schedule follows prior work; our contribution is to make expert top-2 a routable candidate alongside memory and expert top-1. On a 1,403-case retrospective internal analysis, this expansion improves a matched binary router from 88.95% to 91.73% (+2.78 percentage points; 95% CI 1.57--3.99), with 56 rescued errors and 17 regressions. Oracle coverage rises from 90.31% to 96.15%, and the final submitted configuration reaches 92.23% on the same retrospective split. For open questions, strict generation and deterministic guards produce 475/475 schema- valid participant-facing outputs without repair, retry, or hard-gate failure. Visual ablations reveal substantial textual dependence. Candidate expansion supplies the principal controlled routing gain; open-path evidence establishes output-contract validity rather than clinical correctness in medical use or deployment.

Chinese Translation

混合格式医学视觉问答（VQA）需要稳定的选项选择和机器可读的自由文本输出。这两种格式的失败方式不同：多项选择预测可能因选项符号或位置的变化而改变，而临床上合理的开放答案在序列化不当时可能导致自动评估失败。我们通过答案文本记忆、排列稳定的视觉-语言专家和稀疏候选扩展路由器来解决这两个挑战。循环调度遵循先前的工作；我们的贡献在于使专家的前两名成为可路由的候选项，与记忆和专家的第一名一起。通过对1,403个案例的回顾性内部分析，这种扩展将匹配的二元路由器的准确率从88.95%提高到91.73%（+2.78个百分点；95%置信区间1.57--3.99），挽救了56个错误并出现了17个回归。Oracle覆盖率从90.31%上升到96.15%，最终提交的配置在相同的回顾性分割上达到了92.23%。对于开放问题，严格生成和确定性保护机制在没有修复、重试或硬门失败的情况下产生了475/475个模式有效的参与者面对输出。视觉消融实验揭示了显著的文本依赖性。候选扩展提供了主要的受控路由增益；开放路径证据确立了输出合同的有效性，而非医学使用或部署中的临床正确性。

View on arXiv Download PDF AI Translation

cs.CV / 65 / 2609.00960

Conditional Flow Matching for Cross-Field MRI Harmonisation

跨场强MRI协调的条件流匹配

Imre, Baris, Salehi, Aram, Baljer, Levente, Webb, Andrew, Staring, Marius, Ilicak, Efe

Abstract

Magnetic resonance images of the same subject look markedly different across field strengths, which complicates the comparison and pooling of data across sites. We address cross-field brain-MRI translation for the MRIxFields2026 challenge, and in particular its Task~3: a single model that translates between any directed pair of the five field strengths and across three contrasts. We phrase the problem as a conditional flow matching path: because the source and target volumes are spatially registered, we learn a velocity field that carries the source slice directly to the target slice, rather than starting from noise. To learn this mapping from only three paired subjects, the unified model is trained in three stages: a degradation-bridge pretraining that distills a restoration prior from the abundant unpaired retrospective cohort, a cross-field finetuning over all directed pairs on the paired cohort, and an adversarial refinement that sharpens the output. At inference, we integrate the learned velocity with a second-order Heun solver in a handful of steps. A restoration prior learned without any paired data already reaches a mean SSIM of 0.837, and each subsequent training stage improves on it. A single 6.3M-parameter model thereby covers all 60 field-pair and contrast combinations, with inference in five solver steps per slice. On the challenge evaluation set the model reaches a mean SSIM of 0.909, averaged over the three contrasts, outperforming regression and diffusion baselines built on the identical network on all three challenge metrics.

Chinese Translation

同一受试者的磁共振图像在不同场强下看起来差异显著，这使得跨站点数据的比较和汇总变得复杂。我们针对MRIxFields2026挑战中的跨场脑MRI翻译问题，特别是其任务3：一个能够在五个场强之间的任意定向对以及三种对比度之间进行翻译的单一模型。我们将该问题表述为条件流匹配路径：由于源体积和目标体积在空间上已对齐，我们学习一个速度场，将源切片直接传送到目标切片，而不是从噪声开始。为了仅从三个配对受试者中学习这种映射，统一模型分三个阶段进行训练：首先是降解-桥接的预训练，从丰富的未配对回顾性队列中提炼出恢复先验；其次是在配对队列上对所有定向对进行跨场微调；最后是对输出进行锐化的对抗性精炼。在推理时，我们将学习到的速度与二阶Heun求解器结合，经过少数步骤完成。没有任何配对数据学习到的恢复先验已经达到了0.837的平均结构相似性指数（SSIM），而每个后续训练阶段都在此基础上有所改善。因此，单个参数为630万的模型覆盖了所有60个场对和对比度组合，每个切片在五个求解步骤中完成推理。在挑战评估集上，该模型在三种对比度下的平均SSIM达到了0.909，超越了基于相同网络构建的回归和扩散基线，在所有三个挑战指标上均表现优异。

View on arXiv Download PDF AI Translation

cs.CV / 66 / 2609.00968

ReFlowSET: Representation-Aligned Latent Flow Matching for SAR-to-EO Image Translation

ReFlowSET：用于SAR到EO图像翻译的表征对齐潜在流匹配

Do, Jeonghyeok, Lee, Seungchul, Kim, Munchurl

Abstract

SAR-to-EO image translation aims to generate electro-optical (EO) imagery from synthetic aperture radar (SAR) observations. Existing latent diffusion approaches typically inherit a predetermined autoencoder, although reconstruction fidelity can vary substantially across codecs and modalities. Because the latent codec affects the round-trip preservation of both SAR conditions and EO targets, codec selection constitutes a fundamental design choice; nevertheless, existing methods largely rely on codecs pretrained on natural images. To remedy this, we introduce ReFlowSET, a conditional latent flow-matching framework that selects its codec through a joint SAR--EO reconstruction audit. Rather than inheriting a heavyweight pretrained generator, ReFlowSET trains a substantially smaller conditional DiT from scratch in the selected latent space, using dual-stream SAR conditioning followed by joint feature refinement. To provide semantic guidance for this from-scratch training, intermediate noisy-EO features are aligned with clean target-EO representations extracted by a frozen vision foundation model. This alignment is used only during training and introduces no additional inference cost. Experiments on QXS-SAROPT and SAR2Opt demonstrate state-of-the-art performance across diverse perceptual fidelity and distributional metrics. Code and pretrained weights are publicly available at https://github.com/KAIST-VICLab/ReFlowSET.

Chinese Translation

SAR到EO图像翻译旨在从合成孔径雷达（SAR）观测生成电光（EO）图像。现有的潜在扩散方法通常继承一个预先确定的自编码器，尽管重建保真度在不同编解码器和模态之间可能有显著差异。由于潜在编解码器影响SAR条件和EO目标的双向保留，编解码器选择构成了一个基本的设计选择；然而，现有方法在很大程度上依赖于在自然图像上预训练的编解码器。为了解决这个问题，我们提出了ReFlowSET，这是一种条件潜在流匹配框架，通过联合SAR-EO重建审计来选择其编解码器。ReFlowSET并不是继承一个重量级的预训练生成器，而是在所选潜在空间中从头开始训练一个显著较小的条件DiT，采用双流SAR条件化，随后进行联合特征精炼。为了为这种从头训练提供语义指导，中间的噪声EO特征与由冻结的视觉基础模型提取的干净目标EO表征对齐。该对齐仅在训练期间使用，并且不会引入额外的推理成本。在QXS-SAROPT和SAR2Opt上的实验表明，在多样的感知保真度和分布度量上表现出最先进的性能。代码和预训练权重可在https://github.com/KAIST-VICLab/ReFlowSET上公开获取。

View on arXiv Download PDF AI Translation

cs.CV / 67 / 2609.00984

Semi-Supervised Virtual Staining via Morphology Preservation and Histopathological Realism Constraints

通过形态保持和组织病理现实约束的半监督虚拟染色

Wang, Baoshun, Lin, Weiping, Wang, Linwu, Hu, Yihuang, Magnier, Baptiste, Wang, Liansheng

Abstract

Virtual staining aims to computationally generate target-stained histopathological images while reducing the cost and time associated with conventional staining procedures. However, existing methods rely predominantly on strictly paired and accurately registered training data, which are difficult and expensive to obtain in routine practice. To reduce this dependence, we propose a stable semi-supervised virtual staining framework that jointly exploits both limited paired data and abundant unpaired source images. Directly incorporating unpaired images is challenging because their generated results lack corresponding targets for supervision, potentially leading to unrealistic staining, morphological degradation, or even training collapse. To obtain reliable supervision from these images, Hessian-derived morphology preservation extracts structural cues from each source image and constrains the generated output to retain tissue morphology. Histopathological realism constraints further guide the output toward plausible target-stain characteristics, preventing the source-derived structural supervision from degenerating into contour enhancement or simple color transformation. Together, the two components suppress structural and appearance drift, stabilize semi-supervised stain translation, and promote the preservation of diagnostically relevant information. Extensive experiments on H&E-to-IHC translation for Ki67 and HER2, as well as FFPE-to-H&E translation, demonstrate consistent improvements in image quality, morphology preservation, robustness, and downstream diagnostic performance. Code will be available.

Chinese Translation

虚拟染色旨在计算生成目标染色的组织病理图像，同时减少与传统染色过程相关的成本和时间。然而，现有方法主要依赖严格配对和准确注册的训练数据，这在日常实践中难以获得且成本高昂。为了减少这种依赖，我们提出了一种稳定的半监督虚拟染色框架，该框架共同利用有限的配对数据和丰富的未配对源图像。直接纳入未配对图像具有挑战性，因为其生成的结果缺乏相应的目标进行监督，可能导致不现实的染色、形态退化，甚至训练崩溃。为了从这些图像中获得可靠的监督，基于Hessian的形态保持提取每个源图像的结构线索，并约束生成的输出以保持组织形态。组织病理现实约束进一步引导输出朝向合理的目标染色特征，防止源导出的结构监督退化为轮廓增强或简单的颜色转换。两个组件共同抑制结构和外观漂移，稳定半监督染色转换，并促进诊断相关信息的保留。在Ki67和HER2的H&E到IHC转换以及FFPE到H&E转换的广泛实验中，显示出图像质量、形态保持、鲁棒性和下游诊断性能的一致改善。代码将会公开。

View on arXiv Download PDF AI Translation

cs.CV / 68 / 2609.00994

EvoGS: Modeling Deformation Evolution for Dynamic Gaussian Splatting

EvoGS：动态高斯点云变形演化建模

Dong, Wei, Shirani, Shahram, Chen, Jun, Zhou, Han

Abstract

Recent extensions of 3D Gaussian Splatting (3DGS) enable real-time novel view synthesis in dynamic scenes by learning time-conditioned Gaussian deformations. However, existing MLP-based methods typically estimate deformations independently at each timestamp, making them less robust to large or abrupt motions. To address this issue, we propose \textbf{EvoGS}, a 3DGS-based dynamic reconstruction framework that models Gaussian deformation as a temporal evolution process. EvoGS maintains persistent deformation states for each Gaussian, extrapolates future states from historical deformation states, and corrects the predictions with MLP-derived observations. The correction is adaptively weighted using a temporal residual memory and evolution statistics such as deformation velocity and trajectory deviation. To further improve reconstruction quality, EvoGS introduces deformation-aware densification. Clone and split operations are performed along corrected deformation directions, while an uncertainty-aware strategy suppresses densification for Gaussians with unstable deformation histories. Experiments show that EvoGS improves dynamic novel view synthesis quality and achieves competitive performance across benchmarks.

Chinese Translation

最近对3D高斯点云（3DGS）的扩展通过学习时间条件的高斯变形，实现了动态场景中的实时新视图合成。然而，现有的基于多层感知器（MLP）的方法通常在每个时间戳独立估计变形，使其对大幅或突发运动的鲁棒性较差。为了解决这个问题，我们提出了 extbf{EvoGS}，一个基于3DGS的动态重建框架，将高斯变形建模为一个时间演化过程。EvoGS为每个高斯保持持久的变形状态，从历史变形状态中推断未来状态，并使用MLP推导的观测值修正预测。修正过程使用时间残差记忆和变形速度、轨迹偏差等演化统计信息进行自适应加权。为了进一步提高重建质量，EvoGS引入了变形感知的稠密化。在修正的变形方向上执行克隆和分裂操作，同时不确定性感知策略抑制对变形历史不稳定的高斯的稠密化。实验表明，EvoGS提高了动态新视图合成的质量，并在各基准测试中实现了竞争性能。

View on arXiv Download PDF AI Translation

cs.CV / 69 / 2609.00995

CQF-HMR: Continuous Quaternion Flows for Probabilistic 3D Human Mesh Recovery from a Single Image

CQF-HMR：基于连续四元数流的单幅图像概率性3D人类网格恢复

Le, Cuong, Tran, Bao-Long, Melnyk, Pavlo, Dehdarirad, Tahereh, Wandt, Bastian, Wadenbäck, Mårten

Abstract

Recovering 3D digital humans from a single 2D image is an ill-posed computer vision problem due to the loss of depth information. Probabilistic 3D human pose estimation compensates for this by estimating a set of 3D hypotheses from a prior distribution via generative models. However, most prior work focuses only on 3D keypoints, which often leads to implausible poses that are difficult to apply to downstream tasks, e.g. animation or digital humans. SMPL-based methods are more scalable thanks to the explicit body priors, but it requires more complex modeling of the generation process due to the non-additive nature of the joint rotations. In this work, we propose a novel approach for probabilistic 3D humans using quaternion-constrained continuous normalizing flows conditioned on 2D pose estimations. Our proposed quaternion flows show significant advantages over approaches using other rotation representations. Experiments demonstrate state-of-the-art results of our method on Human3.6M, particularly in ambiguous settings, and comparable pose estimation accuracy on challenging 3DPW and EMDB benchmarks.

Chinese Translation

从单幅2D图像恢复3D数字人类是一个不适定的计算机视觉问题，因为缺失了深度信息。概率性3D人类姿态估计通过从先验分布中估计一组3D假设来弥补这一点，通常采用生成模型。然而，大多数先前的工作仅关注3D关键点，这往往导致不合理的姿态，难以应用于下游任务，例如动画或数字人类。基于SMPL的方法由于显式的身体先验而更具可扩展性，但由于关节旋转的非加性特性，它需要对生成过程进行更复杂的建模。在本研究中，我们提出了一种新颖的方法，通过基于四元数约束的连续归一化流，结合2D姿态估计来进行概率性3D人类建模。我们提出的四元数流在使用其他旋转表示的方法中显示出显著优势。实验表明，我们的方法在Human3.6M数据集上取得了最先进的结果，特别是在模糊设置下，并在具有挑战性的3DPW和EMDB基准上表现出可比的姿态估计精度。

View on arXiv Download PDF AI Translation

cs.CV / 70 / 2609.00996

Does This Moment Justify the Recommendation? Counterfactual Behavior-Grounded Evidence Retrieval for Personalized Video Recommendation

这一时刻是否证明了推荐的合理性？基于反事实行为的个性化视频推荐证据检索

Liu, Xin

Abstract

Personalized video recommendation predicts user preference at the video level, while temporal video grounding localizes query-relevant moments. However, strong localization does not establish whether the retrieved moment constitutes valid evidence for recommending the video to a particular user. We study counterfactual behavior-grounded evidence retrieval, which separates where personalized evidence occurs from whether such evidence exists and evaluates whether model predictions respond consistently when that evidence is replaced. We introduce CBGER-10K, containing 5,000 controlled factual--counterfactual pairs for 3,026 users, where each pair replaces only the focal behavior-supported segment while preserving the user, temporal position, and hard distractors. We further propose CBGER, a compact framework that decouples segment-level localization from video-level evidence estimation and learns both through structured counterfactual supervision. CBGER achieves $0.4432$ MRR, $0.6977$ Pair Accuracy, and $0.6987$ Intervention Consistency across five adapted personalized-highlight and temporal-grounding baselines. Notably, compared with QD-DETR, its MRR improvement is not statistically significant, while Pair Accuracy improves by $11.03$ points. These results show that accurate temporal localization does not necessarily imply reliable personalized evidence existence, motivating explicit evaluation of Whether alongside Where.

Chinese Translation

个性化视频推荐在视频层面预测用户偏好，而时间视频定位则将查询相关的时刻进行本地化。然而，强有力的定位并不能确定所检索的时刻是否构成推荐该视频给特定用户的有效证据。我们研究了基于反事实行为的证据检索，该方法将个性化证据的发生位置与该证据是否存在分开，并评估当该证据被替换时模型预测是否一致。我们引入了CBGER-10K，包含5,000个受控的事实-反事实对，涉及3,026名用户，其中每对仅替换焦点行为支持的片段，同时保留用户、时间位置和难以分散注意力的干扰项。我们进一步提出了CBGER，这是一个紧凑的框架，将片段级定位与视频级证据估计解耦，并通过结构化反事实监督学习两者。CBGER在五个适应的个性化高亮和时间定位基线中实现了$0.4432$的平均排名率（MRR）、$0.6977$的配对准确率和$0.6987$的干预一致性。值得注意的是，与QD-DETR相比，其MRR的改善在统计上并不显著，而配对准确率提高了$11.03$个百分点。这些结果表明，准确的时间定位并不一定意味着可靠的个性化证据的存在，促使我们明确评估“是否”与“何处”。

View on arXiv Download PDF AI Translation

cs.CV / 71 / 2609.01004

SinkPruner: Sink-Free Visual Token Pruning for Multimodal Large Language Models

SinkPruner：无汇点的多模态大型语言模型视觉标记剪枝

Li, Shiyu, Hu, Zi-Yuan, Huang, Shijia, Li, Yanyang, Zhong, Yiwu, Wang, Liwei

Abstract

Despite their strong multimodal understanding ability, multimodal large language models (MLLMs) incur substantial computational overhead when processing long visual token sequences. To reduce inference costs, recent studies have explored visual token pruning through vision-centric or text-guided strategies. However, these methods often overlook high-norm outlier tokens, i.e., tokens with abnormally large feature norms, leading to suboptimal pruning decisions. In this work, we show that such high-norm outlier tokens are highly redundant in both feature and spatial dimensions, yet are often mistakenly preserved as informative cues by existing methods. Motivated by this observation, we propose SinkPruner, a training-free visual token pruning framework for efficient MLLM inference. SinkPruner follows a coarse-to-fine design with two key modules: a visual sanitizer that filters high-norm redundancies and alleviates attention sink and attention dispersion, and a text-guided pruner that further retains tokens semantically aligned with the text query. Extensive experiments on twelve image-language and four video-language benchmarks demonstrate the effectiveness, efficiency, and generalizability of our framework. Notably, SinkPruner preserves 96.5% (91.8%) of the original performance of LLaVA-1.5 (Qwen2.5-VL) under an 89% token reduction. Experiments further indicate that our visual sanitizer exhibits promising transferability in enhancing the performance of existing pruning methods. Our code is available at https://github.com/LaVi-Lab/SinkPruner.

Chinese Translation

尽管多模态大型语言模型（MLLMs）具有强大的多模态理解能力，但在处理长视觉标记序列时，仍会产生大量计算开销。为了降低推理成本，近期研究探索了通过以视觉为中心或以文本为指导的策略进行视觉标记剪枝。然而，这些方法往往忽视了高范数异常标记，即特征范数异常大的标记，导致次优的剪枝决策。在本研究中，我们表明，这些高范数异常标记在特征和空间维度上都高度冗余，但现有方法常常错误地将其视为信息线索。基于这一观察，我们提出了SinkPruner，一种无训练的视觉标记剪枝框架，以实现高效的MLLM推理。SinkPruner采用粗到细的设计，包含两个关键模块：一个视觉清理器，用于过滤高范数冗余，缓解注意力汇点和注意力分散；一个以文本为指导的剪枝器，进一步保留与文本查询语义对齐的标记。在十二个图像-语言和四个视频-语言基准上的大量实验表明了我们框架的有效性、效率和通用性。值得注意的是，在89%的标记减少下，SinkPruner保留了LLaVA-1.5（Qwen2.5-VL）原始性能的96.5%（91.8%）。实验进一步表明，我们的视觉清理器在提升现有剪枝方法性能方面表现出良好的迁移性。我们的代码可在 https://github.com/LaVi-Lab/SinkPruner 获取。

View on arXiv Download PDF AI Translation

cs.CV / 72 / 2609.01014

Low-Quality Face Recognition using Center Aligned Representations and Local Margin Constraints

基于中心对齐表示和局部边际约束的低质量人脸识别

Dilaver, Vedat Can, Riggan, Benjamin S.

Abstract

Low-quality face recognition (LQFR) remains challenging due to the difficulty of matching degraded query (probe) images against low-quality (LQ) enrollment (gallery) imagery and the scarcity of training data for large-scale models. While recent face recognition (FR) models perform well on high-quality (HQ) imagery, their accuracy drops significantly on LQ images with extremely low signal-to-noise ratio (SNR). Moreover, fine-tuning HQ-pretrained models on LQ data often improves LQ recognition at the expense of HQ generalization. This trade-off becomes more pronounced in modern evaluation settings spanning multiple datasets with varying image quality levels. To address these limitations, we propose a unified framework that combines three main components: (1) Local Probability Margin (LPM), which estimates per-sample difficulty directly from the model's discriminative landscape; (2) Nested Attention Module (NAM), a new low-rank adapter module that embeds a self-attention mechanism within selected transformer layers; and (3) Quality Gating Protocol (QGP), where an off-the-shelf image quality estimator modulates the adapter contribution at test time, enabling a single model to handle the full quality spectrum without sacrificing HQ performance. Experiments on surveillance (TinyFace, SurvFace) and standard (IJB-B, IJB-C) face recognition benchmarks demonstrate consistent gains in both identification and verification. Code and models will be released at github.com/candllq/nam.

Chinese Translation

低质量人脸识别（LQFR）仍然面临挑战，因为在低质量（LQ）注册（图库）图像与退化的查询（探测）图像之间进行匹配非常困难，并且大规模模型的训练数据稀缺。尽管最近的人脸识别（FR）模型在高质量（HQ）图像上表现良好，但在信噪比（SNR）极低的LQ图像上，其准确性显著下降。此外，在LQ数据上微调HQ预训练模型通常会提高LQ识别能力，但代价是HQ泛化能力的下降。这种权衡在跨多个数据集的现代评估设置中变得更加明显，这些数据集具有不同的图像质量水平。为了解决这些限制，我们提出了一个统一框架，结合了三个主要组件：（1）局部概率边际（LPM），它直接从模型的区分性景观中估计每个样本的难度；（2）嵌套注意力模块（NAM），一种新的低秩适配器模块，在选定的变换器层中嵌入自注意力机制；（3）质量门控协议（QGP），在测试时，现成的图像质量估计器调节适配器的贡献，使单个模型能够处理整个质量范围而不牺牲HQ性能。在监控（TinyFace，SurvFace）和标准（IJB-B，IJB-C）人脸识别基准上的实验表明，在身份识别和验证方面均取得了一致的提升。代码和模型将发布在github.com/candllq/nam。

View on arXiv Download PDF AI Translation

cs.CV / 73 / 2609.01027

Fi-ImageNet-1k: An OOD Benchmark From the Inside of the ImageNet-1k Validation Set

Fi-ImageNet-1k：来自ImageNet-1k验证集内部的OOD基准

Rozumnyi, Ruslan, Suchánek, Matěj, Vojíř, Tomáš, Janoušková, Klára, Matas, Jiří

Abstract

Out-of-distribution (OOD) detection predicts whether a test image belongs to none of the predefined classes. To evaluate this task, benchmarks need images from outside the in-distribution (ID) data; typically, these are defined or collected in an ad hoc fashion. Since no ground truth is perfect, ID-labeled datasets themselves contain a natural source of OOD images. We exploit such annotation errors and present Fi-ImageNet-1k, an OOD dataset built from ImageNet-1k validation images that the recent ReImageNet reannotation effort assigned to no ImageNet-1k class. Each image was examined by expert human annotators supported by evidence from MLLMs, VLMs, and reverse image search, comparing it against all visually similar ID classes. We keep only images that could be assigned a specific class outside the ImageNet-1k label space. The resulting Fi-ImageNet-1k, with 655 images from 522 classes, is substantially more challenging than any commonly used OOD dataset. No evaluated combination of classifier and OOD detector achieves a false positive rate below 51% at 95% true positive rate (FPR@95). Compared to the recent NINCO, our dataset is 3.8x more challenging in the FPR@95 metric for state-of-the-art supervised OOD detection methods.

Chinese Translation

分布外（OOD）检测旨在预测测试图像是否属于任何预定义类别。为了评估这一任务，基准需要来自分布内（ID）数据之外的图像；通常，这些图像是以临时方式定义或收集的。由于没有完美的真实标签，ID标记的数据集本身就包含了自然来源的OOD图像。我们利用这种标注错误，提出了Fi-ImageNet-1k，这是一个基于ImageNet-1k验证图像构建的OOD数据集，这些图像在最近的ReImageNet重新标注工作中被分配为没有ImageNet-1k类别。每张图像都由专家人类标注员进行审查，并借助于多模态大语言模型（MLLMs）、视觉语言模型（VLMs）和反向图像搜索的证据，对比所有视觉上相似的ID类别。我们仅保留那些可以被分配到ImageNet-1k标签空间之外的特定类别的图像。最终生成的Fi-ImageNet-1k包含来自522个类别的655张图像，其挑战性远高于任何常用的OOD数据集。在95%真实正例率（FPR@95）下，没有评估的分类器和OOD检测器组合能够实现低于51%的假阳性率。与最近的NINCO相比，我们的数据集在最先进的监督OOD检测方法的FPR@95指标上挑战性提高了3.8倍。

View on arXiv Download PDF AI Translation

cs.CV / 74 / 2609.01041

ViTAMINS: An Empirical Study of Training Self-Supervised Vision Transformers with Synthetic Hard Negatives

ViTAMINS：一种将合成困难负样本应用于自监督视觉变换器训练的实证研究

Giakoumoglou, Nikos, Floros, Andreas, Papadopoulos, Kleanthis-Marios, Stathaki, Tania

Abstract

We introduce ViTAMINS, a method that integrates synthetic hard negatives into unsupervised vision transformer pretraining to improve representation quality. Our approach is thoroughly benchmarked on ImageNet and transfer learning, image retrieval, copy detection, and image, video segmentation tasks. Notably, our proposed negatives give rise to emergent properties, where learned representations contain explicit information about the semantic content of an image and serve as excellent classifiers (up to +11.3% over baselines). ViTAMINS achieves these benefits through simple modifications to existing contrastive frameworks and outperforms competing methods while being more resource efficient, e.g., our ViT-B surpasses V-JEPA with ViT-L. Our findings motivate reconsidering contrastive learning as a simpler yet powerful alternative to dominant generative and self-distillation approaches.

Chinese Translation

我们提出了ViTAMINS，一种将合成困难负样本集成到无监督视觉变换器预训练中的方法，以提高表示质量。我们的方法在ImageNet及迁移学习、图像检索、复制检测和图像、视频分割任务上进行了全面的基准测试。值得注意的是，我们提出的负样本产生了突现特性，学习到的表示包含关于图像语义内容的显式信息，并且作为优秀的分类器（相较于基线提升高达11.3%）。ViTAMINS通过对现有对比框架的简单修改实现了这些优势，并在资源效率方面优于竞争方法，例如，我们的ViT-B超越了使用ViT-L的V-JEPA。我们的研究结果促使我们重新考虑对比学习作为一种更简单但强大的替代方案，以应对主流的生成和自蒸馏方法。

View on arXiv Download PDF AI Translation

cs.CV / 75 / 2609.01059

Dyn-3D: Unveiling and Resolving Ego-Motion Ambiguity in Vision-Language Models

Dyn-3D：揭示并解决视觉语言模型中的自我运动模糊

Ding, Jiayu, Liu, Zhuodong, Zhang, Lei, Xiong, Manyu, Jin, Hongbo, Tang, Haoran, Zhang, Hongbo, Zhu, Changen, Xing, Wenbo

Abstract

As Vision-Language Models (VLMs) tackle dynamic 3D spatial reasoning, ego-motion perception becomes essential to resolve monocular scale ambiguity. However, current models often overfit to smooth trajectory priors rather than genuinely understanding physical motion. Consequently, their spatial reasoning degrades severely under large displacements, a phenomenon we term Kinematic Collapse. This failure stems from spurious visual-motion correlations in natural videos and a lack of explicit physical supervision. To evaluate this, we introduce Dyn-3D, a benchmark using counterfactual 3D rendering to rigorously decouple visual changes from true kinematic properties. Furthermore, we propose the TempoVista framework, featuring the Kinematic-GSPO algorithm. By embedding metric physical ground truth into policy optimization, TempoVista explicitly grounds visual representations in 3D space. Experiments demonstrate that our approach significantly improves both motion estimation and robust spatial reasoning by utilizing camera dynamics as an effective geometric calibration signal.

Chinese Translation

随着视觉语言模型（VLMs）处理动态三维空间推理，自我运动感知变得至关重要，以解决单目尺度模糊。然而，当前模型往往过度拟合于平滑轨迹先验，而不是对物理运动进行真正理解。因此，它们在大位移下的空间推理严重下降，这一现象我们称之为运动崩溃（Kinematic Collapse）。这种失败源于自然视频中虚假的视觉运动相关性以及缺乏明确的物理监督。为了评估这一点，我们引入了Dyn-3D，一个使用反事实三维渲染的基准，严格解耦视觉变化与真实运动特性。此外，我们提出了TempoVista框架，包含运动-广义状态预测优化（Kinematic-GSPO）算法。通过将度量物理真实值嵌入策略优化中，TempoVista明确将视觉表征固定在三维空间中。实验表明，我们的方法通过利用相机动态作为有效的几何校准信号，显著提高了运动估计和稳健的空间推理能力。

View on arXiv Download PDF AI Translation

cs.CV / 76 / 2609.01092

IT-TextFusion: Iterative Text-Image Interaction with Text-Guided Residual Refinement for Degradation-Aware Image Fusion

IT-TextFusion：具有文本引导残差细化的迭代文本-图像交互用于降解感知图像融合

Liu, Siyang, Zhou, Peiyi, Jin, Tianle, Bian, Rongrong, Jin, Zheke, Gao, Mengze

Abstract

Text-guided image fusion has recently emerged as an effective paradigm for integrating multi-modal information while enabling flexible and task-oriented fusion control. However, existing text-guided fusion methods often rely on shallow semantic-visual interaction and limited attention mechanisms, which restrict their ability to robustly handle complex degradations and fully exploit textual guidance. In this paper, we propose an iterative text-guided image fusion framework that incorporates text-conditioned feature interaction across multiple fusion and refinement stages. The proposed method integrates deepest-level Cross-Attention, multi-scale Cross-Gate Fusion, and stage-specific text-conditioned modulation, allowing the global text embedding to condition hierarchical feature fusion and residual refinement. By repeatedly injecting the pooled text embedding across hierarchical decoder and refinement stages, the proposed framework provides degradation-aware global semantic conditioning while preserving complementary information from the visible and infrared modalities. Experiments on several benchmark datasets show that the proposed method improves several information-preservation and perceptual-quality metrics, while exhibiting metric-dependent trade-offs on some datasets.

Chinese Translation

文本引导的图像融合最近成为整合多模态信息的有效范式，同时实现灵活和任务导向的融合控制。然而，现有的文本引导融合方法往往依赖于浅层的语义-视觉交互和有限的注意力机制，这限制了它们在稳健处理复杂降解和充分利用文本引导方面的能力。本文提出了一种迭代文本引导图像融合框架，该框架在多个融合和细化阶段中结合了文本条件特征交互。所提出的方法整合了最深层的交叉注意力（Cross-Attention）、多尺度交叉门融合（Cross-Gate Fusion）和特定阶段的文本条件调制，使得全局文本嵌入能够调节层次特征融合和残差细化。通过在层次解码器和细化阶段中反复注入汇聚的文本嵌入，所提出的框架提供了降解感知的全局语义调节，同时保留了来自可见光和红外模态的互补信息。在多个基准数据集上的实验表明，所提出的方法在多个信息保留和感知质量指标上有所改善，同时在某些数据集上表现出指标依赖的权衡。

View on arXiv Download PDF AI Translation

cs.CV / 77 / 2609.01123

P-PatchDiff: Progressive Patch Diffusion Models for Low-light Image Enhancement

P-PatchDiff：用于低光照图像增强的渐进式补丁扩散模型

Guo, Ruoyu, Zhong, Haonan, Pagnucco, Maurice, Song, Yang

Abstract

Recent advancements in low-light image enhancement have leveraged diffusion models for their strong ability to generate perceptually realistic, detailed images. Patch diffusion models further offer a promising solution to size-agnostic image restoration while improving efficiency. However, existing methods typically rely on small, fixed patches (e.g., 64$\times$64) that cannot capture image-level brightness context, whereas enlarging the receptive field improves brightness and colour estimation but substantially increases computational cost. Moreover, low-light images often exhibit uneven brightness across regions, making it necessary to ensure that locally enhanced patches remain visually coherent when combined into the full image. To address these limitations, we propose P-PatchDiff, a scalable progressive patch diffusion framework for low-light image enhancement that dynamically adjusts patch size throughout the denoising process, enabling a gradual shift from local to global views. A Multi-Patch Alignment strategy is also introduced to normalise features across varying patch scales using an estimated global brightness proxy. Rather than pursuing pixel-level reconstruction accuracy, P-PatchDiff focuses on scalability and coherent brightness across the whole image, allowing the model to perceive multi-scale information and better enhance regions with varying brightness. We empirically demonstrate that P-PatchDiff effectively enhances images ranging from 400 $\times$ 600 to 4K and is 80$\times$ faster than existing patch diffusion models while using less than 9GB of memory. The code is available at https://github.com/RuoyuGuo/P-PatchDiff.

Chinese Translation

近年来，低光照图像增强领域的进展利用了扩散模型，因其强大的生成感知真实、细节丰富图像的能力。补丁扩散模型进一步提供了一种有前景的解决方案，能够在提高效率的同时实现与图像大小无关的图像恢复。然而，现有方法通常依赖于小的固定补丁（例如，64$ imes$64），无法捕捉图像级的亮度上下文，而扩大感受野虽然可以改善亮度和颜色估计，但会显著增加计算成本。此外，低光照图像通常在不同区域表现出不均匀的亮度，这使得在将局部增强的补丁合并为完整图像时，确保视觉一致性变得必要。为了解决这些局限性，我们提出了P-PatchDiff，一种可扩展的渐进式补丁扩散框架，用于低光照图像增强，该框架在去噪过程中动态调整补丁大小，实现从局部到全局视图的渐进转变。同时，我们引入了一种多补丁对齐策略，通过估计的全局亮度代理来规范化不同补丁尺度的特征。P-PatchDiff并不追求像素级重建精度，而是关注于可扩展性和整个图像的一致亮度，使模型能够感知多尺度信息，更好地增强亮度变化的区域。我们通过实验证明，P-PatchDiff有效增强了从400 $ imes$ 600到4K的图像，并且在内存使用少于9GB的情况下，比现有的补丁扩散模型快80倍。代码可在 https://github.com/RuoyuGuo/P-PatchDiff 获取。

View on arXiv Download PDF AI Translation

cs.CV / 78 / 2609.01136

Different Changes Require Different Reasoning: Change-Type-Specialized Experts for Robust Change Captioning

不同的变化需要不同的推理：针对稳健变化描述的变化类型专门化专家

Park, Jiyoung, Oh, InJae, Kim, Jung Uk

Abstract

Change captioning is the task of generating natural language descriptions that explain the changes between a pair of images. Although different change types (e.g., color shifts, object additions) exhibit distinct visual cues and require specialized reasoning processes, existing methods often overlook these distinctions. To address this limitation, we propose Multi-Expert Diagnosis for Image Change (MEDIC), a novel framework that introduces change-type awareness by explicitly modeling change categories. MEDIC employs type-specialized memory experts that dynamically retrieve type-relevant visual patterns conditioned on the input. This design enables each expert to capture diverse variations within its change type while focusing on the most informative visual cues. By softly routing inputs across type-specialized experts and learning dedicated representations for each change category, MEDIC generates more precise and type-aware change descriptions. Extensive experiments demonstrate that the proposed MEDIC consistently outperforms existing methods across diverse and challenging datasets. The code is available at \href{https://github.com/VisualAIKHU/MEDIC}{GitHub}.

Chinese Translation

变化描述是生成自然语言描述以解释一对图像之间变化的任务。尽管不同的变化类型（例如，颜色变化、物体添加）展示出不同的视觉线索并需要专门的推理过程，但现有方法往往忽视这些区别。为了解决这一局限性，我们提出了图像变化的多专家诊断（Multi-Expert Diagnosis for Image Change，MEDIC），这是一个新颖的框架，通过明确建模变化类别引入变化类型意识。MEDIC采用类型专门化的记忆专家，动态检索与输入条件相关的视觉模式。该设计使每个专家能够捕捉其变化类型内的多样化变体，同时专注于最具信息量的视觉线索。通过在类型专门化专家之间柔性路由输入并为每个变化类别学习专门的表示，MEDIC生成更精确且具有类型意识的变化描述。大量实验表明，所提出的MEDIC在多样且具有挑战性的数据集上始终优于现有方法。代码可在 exttt{https://github.com/VisualAIKHU/MEDIC} 获取。

View on arXiv Download PDF AI Translation

cs.CV / 79 / 2609.01141

Revisiting Face Recognition for Monozygotic Twins: The Celeb Twins Test Set

重新审视单卵双胞胎的人脸识别：名人双胞胎测试集

Zang, Michael, Wu, Haiyu, Sharma, Mrinal, Bowyer, Kevin W.

Abstract

Past literature on face recognition for monozygotic (("identical") twins points to facial marks and mirror asymmetry as possible directions for improved accuracy of twins recognition. The Celeb Twins Test Set (CTTS) contains web-scraped image pairs for 80 sets of celebrity twins. It is the only twins test set with meta-data for twins with distinguishing skin marks and possible mirror asymmetry. CTTS is organized in the manner of face verification test sets such as LFW, CALFW, CPLFW, CFP-FP, and AgeDB-30. Current deep CNN matchers can achieve over 76% accuracy in classifying CTTS same-person / different-person image pairs. We show that current matchers do not make use of skin marks, or asymmetry, and discuss reasons for this. Finally, we discuss the feasibility of using generative AI tools such as Grok, ChatGPT and Gemini to create images of imagined monozygotic twins as a means to increase representation of twins in face recognition training sets.

Chinese Translation

以往关于单卵双胞胎（“同卵”双胞胎）的人脸识别文献指出，面部特征和镜像不对称可能是提高双胞胎识别准确性的方向。名人双胞胎测试集（Celeb Twins Test Set，CTTS）包含80组名人双胞胎的网络抓取图像对。它是唯一一个具有区分性皮肤特征和可能的镜像不对称的双胞胎元数据的测试集。CTTS的组织方式与人脸验证测试集如LFW、CALFW、CPLFW、CFP-FP和AgeDB-30相似。目前的深度卷积神经网络（CNN）匹配器在分类CTTS同人/不同人图像对时可以达到超过76%的准确率。我们展示了当前的匹配器并未利用皮肤特征或不对称性，并讨论了其原因。最后，我们讨论了使用生成性人工智能工具如Grok、ChatGPT和Gemini创建想象中的单卵双胞胎图像的可行性，以增加人脸识别训练集中双胞胎的代表性。

View on arXiv Download PDF AI Translation

cs.CV / 80 / 2609.01146

StainPresetNet: Stain Preset Network for Fast Multi-to-Multi Stain Normalization

StainPresetNet：用于快速多对多染色标准化的染色预设网络

Kang, Hongtao, Luo, Die, Chen, Li, Cai, Jing, Hu, Junbo, Liu, Xiuli, Cheng, Shenghua

Abstract

Stain normalization reduces color variations caused by variations in staining protocols and imaging conditions, thereby enhancing computer-aided diagnostic system performance. Traditional methods derive mapping relationships from individual or limited reference images through pixel-wise transformation, offering style flexibility but suffering from inaccurate color mapping extraction. While existing deep-learning-based approaches achieve accurate dataset-wide color mapping through complex neural networks, they face challenges including computational inefficiency, artifact generation, and fixed normalization directions requiring model retraining for directional changes. To address these limitations, we propose StainPresetNet - a novel framework that combines structural preservation with dataset-level color mapping while maintaining computational efficiency. Our method implements pixel-wise normalization guided by preset reference images, enabling multi-directional adaptability without retraining. Evaluations on cytopathology and histopathology datasets demonstrate that StainPresetNet achieves superior color mapping accuracy compared to conventional methods, effectively improves classifier generalization in diagnostic tasks, and reduces computational overhead by 90\% versus existing deep learning approaches. The proposed preset-guided mechanism facilitates flexible adjustment of normalization directions through simple reference image replacement, overcoming the directional rigidity of current deep-learning-based solutions.

Chinese Translation

染色标准化减少了由于染色协议和成像条件变化引起的颜色变化，从而增强了计算机辅助诊断系统的性能。传统方法通过逐像素变换从单个或有限的参考图像中推导映射关系，提供了样式灵活性，但在颜色映射提取方面存在不准确的问题。虽然现有的基于深度学习的方法通过复杂的神经网络实现了准确的数据集范围内的颜色映射，但它们面临计算效率低下、伪影生成以及固定标准化方向需要模型重新训练以应对方向变化等挑战。为了解决这些局限性，我们提出了StainPresetNet——一个新颖的框架，结合了结构保持与数据集级别的颜色映射，同时保持计算效率。我们的方法实现了由预设参考图像引导的逐像素标准化，使得在不重新训练的情况下具备多方向适应性。在细胞病理学和组织病理学数据集上的评估表明，StainPresetNet在颜色映射准确性上优于传统方法，有效提高了诊断任务中分类器的泛化能力，并且与现有的深度学习方法相比，计算开销减少了90%。所提出的预设引导机制通过简单的参考图像替换，便于灵活调整标准化方向，克服了当前基于深度学习解决方案的方向刚性。

View on arXiv Download PDF AI Translation

cs.CV / 81 / 2609.01147

On the Design Fundamentals of Pixel Text Representation Learning

像素文本表示学习的设计基础

Yuan, Chaohao, Yuan, Ruifeng, Huang, Zhuoxu, Rong, Yu, Cheng, Hong, Chan, Hou Pong, Xiao, Chenghao

Abstract

Text-rich visual inputs require models that can read, retrieve, and compress language directly in pixel space, yet existing pixel-text encoders struggle with fixed resolution pretraining, visual shortcut learning, weak visual grounding, and multilingual visual text understanding. In this work, we investigate the fundamental design principles required for robust visual text representation learning. Through systematic controlled ablations, we identify four critical components: variable image resolutions and rendered font sizes provide spatial proxies for high-resolution document generalization; natural image-text pairs are indispensable for grounding and prevent text-only collapse; layout-aware rendering helps prevent pixel-level shortcuts; and a two-stage multilingual curriculum enables effective cross-lingual alignment. By integrating these principles into a scalable training recipe, we train Pixel Linguist II, a native-resolution vision encoder trained with on-the-fly rendering, unified contrastive grounding, and a multilingual curriculum over 280M training examples. Pixel Linguist II sets new state-of-the-art results on English, cross-lingual, and multilingual Visual STS and ViDoRe, while also enabling better MLLM downstream evaluation. Notably, Pixel Linguist II remains robust under 80\% visual token compression, showing great promise for optical context compression. Our code and resources are available at https://github.com/Pixel-Linguist/Pixel-Linguist-II.

Chinese Translation

文本丰富的视觉输入需要能够直接在像素空间中读取、检索和压缩语言的模型，但现有的像素-文本编码器在固定分辨率预训练、视觉捷径学习、弱视觉基础和多语言视觉文本理解方面存在困难。在本研究中，我们探讨了实现稳健的视觉文本表示学习所需的基本设计原则。通过系统的控制消融实验，我们识别出四个关键组件：可变的图像分辨率和渲染字体大小为高分辨率文档泛化提供了空间代理；自然的图像-文本对对于基础构建至关重要，能够防止仅文本的崩溃；布局感知渲染有助于防止像素级捷径；而两阶段的多语言课程则实现了有效的跨语言对齐。通过将这些原则整合到可扩展的训练方案中，我们训练了 Pixel Linguist II，这是一个在280M训练示例上通过即时渲染、统一对比基础和多语言课程进行训练的原生分辨率视觉编码器。Pixel Linguist II 在英语、跨语言和多语言视觉语义文本相似性（Visual STS）及视觉文档检索（ViDoRe）上创造了新的最先进结果，同时也提升了 MLLM 下游评估的效果。值得注意的是，Pixel Linguist II 在80%的视觉标记压缩下仍然表现稳健，显示出在光学上下文压缩方面的巨大潜力。我们的代码和资源可在 https://github.com/Pixel-Linguist/Pixel-Linguist-II 获取。

View on arXiv Download PDF AI Translation

cs.CV / 82 / 2609.01148

Dotting the Eye: An Intent-Driven Image Retouching Agent for Visual Focus Enhancement

点睛之笔：一种基于意图驱动的图像修饰代理用于视觉焦点增强

Qin, Chujie, Zhang, Zilong, Chang, Zewei, Guo, Chunle, Wang, Ruixing, Hu, Tao, Cheng, Ming-Ming, Li, Chongyi

Abstract

Image retouching is commonly formulated as enhancing overall visual quality through color adjustment, but in practice, it also serves to emphasize visual focus by guiding viewers' attention toward a specific subject or region. Achieving such focus-oriented retouching is inherently challenging, as it requires well-coordinated global and local adjustments to manipulate perceptual saliency while maintaining visual naturalness. This intricate process typically demands substantial professional expertise. In this study, we propose EyeControl, a MLLM-driven agent with a diffusion-based retouching executor that enables visual focus enhancement under weak user intent. With only a few clicks or coarse strokes, EyeControl directs visual attention to the intended region, effectively "dotting the eye" of the image. The core idea is to explicitly link the weak user intention with the target editing region and the corresponding tonal adjustment operations during retouching. To achieve this, the system first interprets the intent and image content to infer the visual focus and generate structured intent guidance for the retouching executor. Second, the retouching executor is encouraged to respond more strongly to the target region, explicitly aligning its attention map with a designed pseudo-intent map. We also introduce an operation-consistency constraint to improve coordination between global and local adjustments, achieving more natural and coherent retouching. Additionally, we contribute ControlArt-Bench, a high-quality evaluation dataset for visual focus enhancement. Extensive evaluations demonstrate that EyeControl yields perceptually appealing results with stronger intent alignment. Code can be found in https://github.com/DragonisCV/EyeControl.

Chinese Translation

图像修饰通常被定义为通过颜色调整来提升整体视觉质量，但实际上，它还用于通过引导观众的注意力集中在特定主题或区域上来强调视觉焦点。实现这种以焦点为导向的修饰本质上是具有挑战性的，因为它需要协调全球和局部的调整，以操控感知显著性，同时保持视觉的自然性。这个复杂的过程通常需要相当高的专业技能。在本研究中，我们提出了EyeControl，一种基于多模态大语言模型（MLLM）的代理，配备扩散基础的修饰执行器，使得在用户意图较弱的情况下也能实现视觉焦点增强。只需几次点击或粗略的笔触，EyeControl就能将视觉注意力引导至预期区域，有效地“点睛”图像。其核心思想是在修饰过程中将弱用户意图与目标编辑区域及相应的色调调整操作明确关联。为此，系统首先解读意图和图像内容，以推断视觉焦点并生成结构化的意图指导供修饰执行器使用。其次，修饰执行器被鼓励更强烈地响应目标区域，明确将其注意力图与设计的伪意图图对齐。我们还引入了操作一致性约束，以改善全球和局部调整之间的协调，实现更自然和连贯的修饰。此外，我们贡献了ControlArt-Bench，一个高质量的视觉焦点增强评估数据集。广泛的评估表明，EyeControl产生了感知上令人满意的结果，并具有更强的意图对齐。代码可在 https://github.com/DragonisCV/EyeControl 找到。

View on arXiv Download PDF AI Translation

cs.CV / 83 / 2609.01172

Monocular Depth Estimation from a Single Image: Progress and Opportunities

基于单幅图像的单目深度估计：进展与机遇

Liu, Muxin, Lyu, Xiaoyang, Sun, Yang-Tian, Huang, Yi-Hua, Yang, Ziyi, Dai, Peng, Qi, Xiaojuan

Abstract

Monocular depth estimation has long stood as a fundamental challenge in computer vision, enabling a wide range of applications including 3D reconstruction, robotics, autonomous driving, and augmented reality. This survey traces the field's evolution from early learning-based methods to the emergence of transformative foundation models. We begin by framing the problem, distinguishing between relative and metric depth estimation, and highlighting the key challenges that have shaped a decade of research. We then present common problem formulations and introduce the most widely used datasets, covering indoor, outdoor, and synthetic data. Following this, we review major advances prior to the foundation model era, distilling core insights from influential methods that contributed to improvements in accuracy, efficiency, and robustness. The survey then turns to the recent surge of foundation-model-based approaches, categorizing them into discriminative and generative paradigms and emphasizing the critical roles of large-scale pretraining (e.g., DINOv3) and synthetic data. We compare representative models using both quantitative benchmarks and qualitative examples, and discuss natural extensions to video-based depth estimation. Further, to illustrate real-world impact, we highlight the integration of depth estimation into applications such as visual SLAM, content generation, and robot perception. Finally, we outline open challenges and promising research directions as the field advances further into the era of foundation models.

Chinese Translation

单目深度估计长期以来一直是计算机视觉中的一个基本挑战，能够支持包括三维重建、机器人技术、自动驾驶和增强现实等广泛应用。本文回顾了该领域从早期基于学习的方法到变革性基础模型出现的演变过程。我们首先框定了问题，区分了相对深度估计和度量深度估计，并强调了塑造十年研究的关键挑战。接着，我们介绍了常见的问题表述，并列出了最广泛使用的数据集，涵盖室内、室外和合成数据。随后，我们回顾了基础模型时代之前的主要进展，提炼出对提高准确性、效率和鲁棒性有重要贡献的影响方法的核心见解。调查接着转向最近涌现的基于基础模型的方法，将其分为判别性和生成性范式，并强调大规模预训练（例如，DINOv3）和合成数据的关键作用。我们使用定量基准和定性示例比较了代表性模型，并讨论了基于视频的深度估计的自然扩展。此外，为了说明其在现实世界中的影响，我们强调了深度估计在视觉SLAM、内容生成和机器人感知等应用中的集成。最后，我们概述了随着该领域进一步进入基础模型时代而面临的开放挑战和有前景的研究方向。

View on arXiv Download PDF AI Translation

cs.CV / 84 / 2609.01200

Compressing AI Traffic: Standardized Neural Network Coding of Visual-Token Representations in Split Vision-Language Inference

压缩人工智能流量：分割视觉-语言推理中视觉标记表示的标准化神经网络编码

Heidari, Reza, Tavakoli, Hamed R., Kannala, Juho

Abstract

When the visual encoder and the language decoder of a vision-language model (VLM) run on different compute nodes, the intermediate visual-token embeddings become a communicated payload rather than an internal activation. We call such machine-consumed intermediate tensors AI traffic and ask how far they can be compressed with a standardized, training-free codec. We insert ISO/IEC 15938-17 Neural Network Coding (NNC) round trips on the complete visual interface of a Qwen3-VL-8B-Instruct video question answering pipeline, comprising the main visual-token representation and the DeepStack feature streams, while leaving weights, prompts, and generation untouched, and sweep the quantization parameter (QP) over a wide rate range. Closed-ended Video-MME accuracy remains close to the uncompressed reference up to a 98% reduction of the transmitted BF16 tensor and only then collapses; open-ended MLVU generation shows the same plateau-and-collapse profile under an LLM judge. This robustness is not due to near-lossless reconstruction: the decoded tensor is heavily discretized, carries substantial row-wise relative L2 error, and has a visibly steeper singular-value decay than its source. Downstream reasoning therefore depends on coarse structure and relative geometry rather than exact floating-point values, which argues for rate-task rather than rate-distortion optimization of AI traffic codecs.

Chinese Translation

当视觉-语言模型（VLM）的视觉编码器和语言解码器在不同的计算节点上运行时，中间的视觉标记嵌入变成了传输负载，而非内部激活。我们将这种机器消费的中间张量称为人工智能流量，并探讨使用标准化、无训练的编解码器能够压缩到何种程度。我们在Qwen3-VL-8B-Instruct视频问答管道的完整视觉接口上插入ISO/IEC 15938-17神经网络编码（NNC）往返过程，该管道包括主要的视觉标记表示和DeepStack特征流，同时保持权重、提示和生成不变，并在广泛的比特率范围内调整量化参数（QP）。封闭式视频-MME的准确性在传输的BF16张量减少98%时仍接近未压缩参考，只有在此之后才会崩溃；开放式MLVU生成在LLM评判下显示出相同的平稳-崩溃特征。这种鲁棒性并非由于近乎无损的重建：解码后的张量高度离散，具有相当大的行相对L2误差，并且其奇异值衰减比源张量明显陡峭。因此，下游推理依赖于粗略结构和相对几何，而非精确的浮点值，这支持了对人工智能流量编解码器进行速率-任务而非速率-失真优化的观点。

View on arXiv Download PDF AI Translation

cs.CV / 85 / 2609.01224

S$^2$Prune: Spatially Structured Visual Token Pruning for Multimodal Large Language Models

S$^2$Prune：用于多模态大型语言模型的空间结构视觉标记修剪

Jia, Yuanyuan, Tang, Shunpu, Yang, Qianqian

Abstract

Visual token pruning reduces the inference overhead of multimodal large language models (MLLMs) by retaining only a subset of visual tokens. Existing methods usually select tokens based on importance or redundancy. However, we observe that these criteria produce stable spatial biases across inputs and do not always outperform simple Uniform Grid sampling, highlighting the value of broad spatial coverage. Motivated by this, we propose S$^2$Prune, a training-free pruning method that preserves spatial coverage while adapting token density to local image structure. We first divide the image into regions and assign at least one token to each region to preserve coverage. The remaining token budget is then distributed according to Laplacian variation, giving more tokens to regions with richer structure. We then use Early Representation Change (ERC), computed from the first decoder block, to select representative tokens within each region. We evaluate S$^2$Prune across diverse settings and two MLLM architectures. On Qwen2.5-VL-7B-Instruct, it achieves the highest average accuracy among the evaluated training-free pruning methods. With only 32 of the original 576 visual tokens, it still retains 79.3% of the full-model performance. Code is available at https://github.com/yuanyuanjia71-spec/S2Prune.

Chinese Translation

视觉标记修剪通过仅保留一部分视觉标记来减少多模态大型语言模型（MLLMs）的推理开销。现有方法通常基于重要性或冗余性选择标记。然而，我们观察到这些标准在输入之间产生了稳定的空间偏差，并不总是优于简单的均匀网格采样，这突显了广泛空间覆盖的重要性。基于此，我们提出了 S$^2$Prune，一种无训练的修剪方法，旨在保持空间覆盖的同时调整标记密度以适应局部图像结构。我们首先将图像划分为区域，并为每个区域分配至少一个标记以保持覆盖。然后，根据拉普拉斯变化分配剩余的标记预算，给予结构更丰富的区域更多的标记。接着，我们使用从第一个解码器块计算的早期表示变化（ERC）在每个区域内选择代表性标记。我们在多种设置和两种 MLLM 架构上评估 S$^2$Prune。在 Qwen2.5-VL-7B-Instruct 上，它在评估的无训练修剪方法中实现了最高的平均准确率。仅使用原始 576 个视觉标记中的 32 个，它仍然保留了 79.3% 的全模型性能。代码可在 https://github.com/yuanyuanjia71-spec/S2Prune 获取。

View on arXiv Download PDF AI Translation

cs.CV / 86 / 2609.01249

One Prompt Is Enough: Watermark Laundering Through Foundation Image Models

一个提示就足够：通过基础图像模型进行水印洗涤

Yang, Jidong, Li, Qi, Zong, Wei, Chow, Yang-Wai, Susilo, Willy, Yu, Huaike, Wang, Chunpeng, Gao, Suo

Abstract

Invisible watermarks are typically evaluated against predefined perturbations such as compression, blur, noise, cropping, and denoising. Public foundation image models expose a distinct threat: an attacker can submit a watermarked image with a single reconstruction prompt and obtain a visually faithful output from which the invisible watermark can no longer be decoded reliably. We formalize this failure mode as watermark laundering and evaluate it using a joint payload-fidelity profile that combines bit error rate (BER) with visual and semantic preservation. Across six OpenAI and Google image editing models, three representative watermarking schemes, and 1,800 reconstructed outputs, we identify two complementary laundering regimes: OpenAI models produce the strongest payload disruption across the evaluated schemes, whereas Nano Banana 2 shows that DwtDct remains vulnerable under high-fidelity reconstruction. Prompt ablations show that no single removal-oriented instruction is necessary for payload disruption, indicating that the effect is primarily induced by the reconstruction pathway rather than by explicit attack wording. Comparisons with conventional attacks further show that prompt-conditioned reconstruction constitutes a distinct operational attack interface. These findings motivate foundation-model reconstruction as a missing robustness condition in invisible watermark evaluation.

Chinese Translation

隐形水印通常针对预定义的扰动进行评估，例如压缩、模糊、噪声、裁剪和去噪。公共基础图像模型带来了一个独特的威胁：攻击者可以提交一张带水印的图像，并使用单个重建提示获得一个视觉上忠实的输出，从中无法可靠地解码隐形水印。我们将这种失败模式形式化为水印洗涤，并使用结合比特错误率（BER）与视觉和语义保留的联合负载-保真度分析进行评估。在六个OpenAI和Google图像编辑模型、三种代表性水印方案和1800个重建输出中，我们识别出两种互补的洗涤机制：OpenAI模型在评估的方案中产生了最强的负载干扰，而Nano Banana 2则显示DwtDct在高保真重建下仍然脆弱。提示消融实验表明，负载干扰并不需要单一的去除导向指令，表明这一效应主要是由重建路径引起的，而不是由明确的攻击措辞引起的。与传统攻击的比较进一步表明，提示条件重建构成了一个独特的操作攻击接口。这些发现促使我们将基础模型重建视为隐形水印评估中缺失的鲁棒性条件。

View on arXiv Download PDF AI Translation

cs.CV / 87 / 2609.01252

MeRoPE: Metric Rotary Position Embedding for Camera-Controlled Video Generation

MeRoPE：用于摄像机控制视频生成的度量旋转位置嵌入

Qiao, Zhijian, Wang, Xinjiang, Chen, Jiajie, Huang, Haoming, Li, Meng, Chou, Chih-Chung, Wang, Jing, Shen, Shaojie

Abstract

In camera-controlled video generation, geometry-aware positional encodings condition tokens on camera extrinsics and per-token viewing rays. Existing schemes, however, have a scale-dependent failure mode on real-world metric camera trajectories: homogeneous projective encodings cause attention logits and feature norms to grow unbounded with physical translation baselines. We propose MeRoPE (Metric Rotary Position Embedding), a norm-preserving relative camera encoding for attention. MeRoPE encodes relative orientations between calibrated viewing rays with orthogonal rotation blocks, maps raw metric displacements into multi-frequency rotary phases, and adds a disparity-anchored correspondence prior along the epipolar arc. This design strictly preserves feature norms, bounds pre-softmax attention logits regardless of the physical translation scale, and maintains exact invariance to global rigid coordinate changes. Across nuScenes and PanShot, which cover large-baseline trajectories and diverse camera optics, respectively, MeRoPE achieves stronger camera control than prior encodings, with the best consistency between generated camera motion and conditioning poses in both rotation and translation. Code will be made publicly available.

Chinese Translation

在摄像机控制的视频生成中，几何感知的位置编码将令牌条件化于摄像机外参和每个令牌的视线。然而，现有方案在真实世界的度量摄像机轨迹上存在依赖于尺度的失效模式：均匀的投影编码导致注意力 logits 和特征范数随着物理平移基线的增长而无界增长。我们提出了 MeRoPE（度量旋转位置嵌入），这是一种保持范数的相对摄像机编码，用于注意力机制。MeRoPE 使用正交旋转块编码经过校准的视线之间的相对方向，将原始度量位移映射到多频率旋转相位，并沿着极线弧添加一个基于视差的对应先验。该设计严格保持特征范数，无论物理平移尺度如何，都限制了预软最大注意力 logits，并保持对全局刚性坐标变化的精确不变性。在 nuScenes 和 PanShot 数据集中，分别涵盖了大基线轨迹和多样化的摄像机光学，MeRoPE 实现了比先前编码更强的摄像机控制，在旋转和位移方面生成的摄像机运动与条件姿态之间具有最佳一致性。代码将公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 88 / 2609.01276

Seeing the World and the Self from Egocentric Video

从自我中心视频看世界和自我

Guan, Kai, Jiang, Minchao, WangLi, Ruichen, Zhu, Wentao, Zhang, Lei

Abstract

Complete 3D perception from egocentric video requires recovering the surrounding scene and the wearer's full-body motion in a shared metric frame. Existing methods typically address scene reconstruction and motion estimation separately: scene reconstruction methods ignore the wearer, whereas motion estimation methods lack explicit scene geometry and often depend on external trajectories. Joint recovery is challenging because the two tasks exhibit asymmetric visibility and require different prediction paradigms. The largely visible scene supports deterministic geometric regression, whereas the severely occluded body requires generative motion inference. We therefore propose RESELF (REconstructing the Scene and the sELF), a unified framework that couples deterministic metric geometry reconstruction with geometry-conditioned motion generation. RESELF adapts a geometry foundation model pre-trained on large-scale exocentric data to egocentric video using frame-wise scale and relative-pose consistency objectives. The resulting camera trajectory and latent geometric features condition a diffusion model that recovers the wearer's motion. A subsequent closed-loop kinematic feedback stage further refines the camera head while preserving the reconstructed scene geometry. To support training and evaluation, we curate EE4D-JSM from EgoExo4D by aligning egocentric video, sparse metric scene geometry, camera trajectories, and full-body motion annotations. Experiments show that RESELF outperforms state-of-the-art methods designed for the individual tasks across depth estimation, camera tracking, and full-body motion estimation. Code, models, and datasets will be available at https://ka1guan.github.io/RESELF/.

Chinese Translation

从自我中心视频中获得完整的三维感知需要在共享的度量框架中恢复周围场景和佩戴者的全身运动。现有方法通常分别处理场景重建和运动估计：场景重建方法忽略佩戴者，而运动估计方法缺乏明确的场景几何，通常依赖于外部轨迹。联合恢复是具有挑战性的，因为这两个任务表现出不对称的可见性，并且需要不同的预测范式。大部分可见的场景支持确定性几何回归，而严重遮挡的身体则需要生成性运动推断。因此，我们提出了RESELF（重建场景和自我的框架），这是一个将确定性度量几何重建与几何条件运动生成相结合的统一框架。RESELF通过使用逐帧尺度和相对姿态一致性目标，将在大规模外部数据上预训练的几何基础模型适应于自我中心视频。生成的相机轨迹和潜在几何特征为恢复佩戴者运动的扩散模型提供条件。随后的闭环运动反馈阶段进一步精细化相机头，同时保留重建的场景几何。为了支持训练和评估，我们从EgoExo4D中整理了EE4D-JSM，通过对齐自我中心视频、稀疏度量场景几何、相机轨迹和全身运动注释。实验表明，RESELF在深度估计、相机跟踪和全身运动估计等单个任务上优于现有的最先进方法。代码、模型和数据集将可在https://ka1guan.github.io/RESELF/获取。

View on arXiv Download PDF AI Translation

cs.CV / 89 / 2609.01277

TimeSteer: Inference-Time Speech Scheduling in Joint Audio-Visual Diffusion Models

TimeSteer：联合音频-视觉扩散模型中的推理时间语音调度

Zhou, Chao, Chen, Yiling, Chu, Qi, Gong, Tao, Yu, Nenghai, We, Tianyi

Abstract

Although pretrained joint audio-visual diffusion models offer rich control over \emph{what} to generate, they provide no explicit control over \emph{when} an utterance should occur. To address this, we study \emph{inference-time speech scheduling}, a novel task that places coupled speech and visual articulation within user-specified begin--end intervals without finetuning the backbone model. We uncover two intrinsic properties of the denoising process that enable this task. First, a timing-sensitive text-to-audio cross-attention head exposes each utterance's model-implied source span along the latent timeline. Second, the predicted clean latent already organizes coupled speech and visual articulation, allowing their temporal placement to be edited without regenerating the content. Building on these discoveries, we propose \textbf{TimeSteer}, a training-free framework that localizes each utterance's source span through \textbf{Source Span Localization} and transfers the associated audio-visual latent content from the source interval to the specified target interval through \textbf{Region-Aware Latent Remapping}. We further introduce \textbf{SpeechShift}, the first benchmark for interval-level speech scheduling in joint audio-visual generation. Experiments across two representative backbones show that TimeSteer substantially improves interval controllability over training-free baselines while maintaining competitive overall generation quality.

Chinese Translation

尽管预训练的联合音频-视觉扩散模型在生成内容的 extit{什么}方面提供了丰富的控制，但它们并未提供关于 extit{何时}发生发声的明确控制。为了解决这个问题，我们研究了 extit{推理时间语音调度}，这是一个新颖的任务，它在用户指定的开始-结束区间内将耦合的语音和视觉表达放置，而无需对主干模型进行微调。我们发现了去噪过程的两个内在特性，使得此任务成为可能。首先，时间敏感的文本到音频交叉注意力头揭示了每个发声在潜在时间线上的模型隐含源跨度。其次，预测的干净潜在内容已经组织了耦合的语音和视觉表达，允许在不重新生成内容的情况下编辑它们的时间位置。基于这些发现，我们提出了 extbf{TimeSteer}，一个无训练框架，通过 extbf{源跨度定位}来定位每个发声的源跨度，并通过 extbf{区域感知潜在重映射}将相关的音频-视觉潜在内容从源区间转移到指定的目标区间。我们进一步介绍了 extbf{SpeechShift}，这是联合音频-视觉生成中区间级语音调度的第一个基准。在两个代表性主干模型上的实验表明，TimeSteer在保持竞争性整体生成质量的同时，显著提高了对无训练基线的区间可控性。

View on arXiv Download PDF AI Translation

cs.CV / 90 / 2609.01282

HiLRP: Toward One Trustworthy Explanation for Vision Transformer: Conservation-Valid Attribution via Attention Primitives

HiLRP：朝着一个可信的视觉变换器解释：通过注意力原语实现的守恒有效归因

Nishankar, Sathiyamohan, Sanjeewani, Pubudu, Perera, Asanka, Thuseethan, Selvarajah

Abstract

Vision Transformer (ViT) design has become increasingly diverse, with backbones combining convolutional stems, windowed, linear, or multi-axis attention, patch merging, and spatial reduction in various configurations. This diversity poses challenges for existing attribution methods, whose assumptions often do not hold across ViT variants: Grad-CAM requires a terminal spatial feature map, attention rollout assumes global softmax attention, and layer-wise relevance propagation (LRP) requires module-specific rules. To the best of our knowledge, no existing method provides a unified attribution framework across this architectural space. We show that this architectural diversity can be captured by a simpler underlying structure. The attention and resolution-reduction operators in current ViTs can be decomposed into four operation types: linear maps, bilinear mixing, normalization or gating, and reindexing. Each operation admits a relevance rule that satisfies conservation. Based on these rules, HiLRP supports new backbones by construction rather than by architecture-specific derivation, and its attribution maps decompose the prediction rather than relying on heuristic assumptions. We prove conservation and conditional equivariance and verify both to machine precision. Across 14 attribution methods and 10 architectures, we find that no prior method remains reliable across ViT families, while Faithfulness Correlation becomes uninformative for backbones robust to spatial masking. HiLRP alone preserves conservation across windowed, spatial-reduction, multi-axis, and linear-attention models, where naive extensions can produce zero or inflated relevance. It also localizes attribution failures in class activation mapping, achieving 0.97 Pointing compared with 0.55 for competing methods on EfficientViT.

Chinese Translation

视觉变换器（ViT）的设计日益多样化，骨干网络结合了卷积主干、窗口化、线性或多轴注意力、补丁合并和空间缩减等多种配置。这种多样性对现有的归因方法提出了挑战，因为这些方法的假设往往在不同的ViT变体中不成立：Grad-CAM需要终端空间特征图，注意力展开假设全局软最大注意力，而逐层相关传播（LRP）需要特定于模块的规则。据我们所知，目前没有任何现有方法在这一架构空间中提供统一的归因框架。我们表明，这种架构多样性可以通过一个更简单的基础结构来捕捉。当前ViT中的注意力和分辨率缩减算子可以分解为四种操作类型：线性映射、双线性混合、归一化或门控以及重新索引。每种操作都有一个满足守恒的相关性规则。基于这些规则，HiLRP通过构造而非特定于架构的推导来支持新的骨干网络，其归因图解构了预测，而不是依赖启发式假设。我们证明了守恒性和条件等变性，并将两者验证到机器精度。在14种归因方法和10种架构中，我们发现没有任何先前的方法在ViT家族中保持可靠，而忠实性相关性在对空间掩蔽稳健的骨干网络中变得无信息。HiLRP独自保持了在窗口化、空间缩减、多轴和线性注意力模型中的守恒性，而天真的扩展可能会产生零或膨胀的相关性。它还定位了类别激活映射中的归因失败，在EfficientViT上实现了0.97的指向率，而竞争方法仅为0.55。

View on arXiv Download PDF AI Translation

cs.CV / 91 / 2609.01289

Agentic Multimodal Models for Environmental Hyperspectral Unmixing

用于环境高光谱解混的自主多模态模型

Cholewa, Michał, Ciampi, Luca, Messina, Nicola, Głomb, Przemysław, Amato, Giuseppe

Abstract

Hyperspectral unmixing is a key task in remote sensing that aims to decompose mixed pixels in hyperspectral images into their constituent material signatures, or endmembers, and their fractional abundances. Conventional modular approaches estimate the scene composition through successive model-order estimation, endmember extraction, and abundance estimation stages, whose errors can lead to redundant or ambiguous candidate components and ultimately affect the recovered decomposition. We introduce an algorithm-agnostic, large vision-language model (LVLM)-driven agentic framework that refines the outputs of such pipelines rather than replacing their underlying numerical algorithms. Starting from an initial decomposition, the agent iteratively gathers complementary spectral and spatial evidence through dedicated tools, including spectral-library retrieval and abundance-map visualization, and modifies the active endmember set through merge and discard operations followed by abundance re-estimation. We apply the same refinement procedure to several modular pipelines combining different model-order, extraction, and abundance-estimation methods, and evaluate it on HYDICE Urban, Jasper Ridge, and Stonewall Playa. Experiments show that the proposed agent consistently improves endmember cardinality and generally improves the recovered spectral signatures and abundance maps across heterogeneous modular pipelines, while remaining competitive with integrated end-to-end unmixing methods, including CNN-AE, uDAS, and R-CoNMF. These results highlight the potential of tool-using LVLM agents to combine spectral and spatial evidence for algorithm-agnostic refinement of physically grounded hyperspectral unmixing decompositions. Code is publicly available at https://anonymous.4open.science/r/agentic-hu.

Chinese Translation

高光谱解混是遥感中的一项关键任务，旨在将高光谱图像中的混合像素分解为其组成材料特征或端元及其分数丰度。传统的模块化方法通过连续的模型阶数估计、端元提取和丰度估计阶段来估计场景组成，这些阶段的错误可能导致冗余或模糊的候选成分，最终影响恢复的分解结果。我们提出了一种与算法无关的大型视觉-语言模型（LVLM）驱动的自主框架，该框架对这些管道的输出进行优化，而不是替换其底层数值算法。从初始分解开始，代理通过专用工具（包括光谱库检索和丰度图可视化）迭代地收集互补的光谱和空间证据，并通过合并和丢弃操作修改活动端元集，随后进行丰度重新估计。我们将相同的优化过程应用于几个结合不同模型阶数、提取和丰度估计方法的模块化管道，并在HYDICE Urban、Jasper Ridge和Stonewall Playa上进行了评估。实验表明，所提出的代理一致地改善了端元的基数，并普遍提高了异构模块化管道中恢复的光谱特征和丰度图，同时与包括CNN-AE、uDAS和R-CoNMF在内的集成端到端解混方法保持竞争力。这些结果突显了使用工具的LVLM代理结合光谱和空间证据进行与算法无关的物理基础高光谱解混分解优化的潜力。代码可在 https://anonymous.4open.science/r/agentic-hu 上公开获取。

View on arXiv Download PDF AI Translation

cs.CV / 92 / 2609.01308

CMRVision: A Foundation Model for Cardiac MR Image Analysis

CMRVision：心脏磁共振图像分析的基础模型

Jacob, Athira J., Sharma, Puneet, Rueckert, Daniel

Abstract

Cardiac magnetic resonance (CMR) imaging provides complementary information on cardiac anatomy, function, and tissue characterization across multiple sequences and views. In this work, we investigate foundation model pretraining for 2D CMR and introduce CMRVision, a CMR-specific foundation model trained using DINOv3-style self-supervised learning on a multi-center, multi-sequence cohort of 36 million CMR images. We systematically evaluate architectural and training design choices for domain-specific pretraining. CMRVision is evaluated on two downstream tasks: multi-task segmentation across cine, late gadolinium enhancement (LGE), and mapping sequences, and cine view classification. Our experiments show that CMR-specific pretraining, smaller patch sizes, and patch-level objectives consistently improve downstream performance. Across a multi-task segmentation benchmark, CMRVision achieved the strongest overall performance, outperforming prior natural-image (NI), medical-image, supervised, and CMR foundation model baselines. Improvements were modest but consistent across structures and sequences, with Dice scores ranging from 0.940-0.967 for LV and 0.855-0.905 for myocardium, and reaching 0.929 for RV, 0.920 for LA, and 0.931 for RA. The largest gains were observed for myocardium segmentation in LGE and mapping images. In a zero-shot segmentation task on unseen LGE long-axis views, the model achieved an average Dice score of 0.692, demonstrating cross-view generalization. For cine view classification, CMRVision achieved the highest average accuracy (0.906), compared to prior methods reported in the literature. These results highlight the potential of CMRVision to support robust and generalizable cardiac MRI analysis across multiple sequences and views.

Chinese Translation

心脏磁共振（CMR）成像提供了关于心脏解剖、功能和组织特征的互补信息，涵盖多个序列和视图。在本研究中，我们探讨了针对二维 CMR 的基础模型预训练，并引入了 CMRVision，这是一个专门针对 CMR 的基础模型，采用 DINOv3 风格的自监督学习，在一个包含 3600 万张 CMR 图像的多中心、多序列队列上进行训练。我们系统地评估了针对特定领域的预训练的架构和训练设计选择。CMRVision 在两个下游任务上进行了评估：跨电影序列、晚期钆增强（LGE）和映射序列的多任务分割，以及电影视图分类。我们的实验表明，CMR 特定的预训练、更小的补丁尺寸和补丁级目标一致地提高了下游性能。在一个多任务分割基准测试中，CMRVision 实现了最强的整体性能，超越了先前的自然图像（NI）、医学图像、监督学习和 CMR 基础模型基线。尽管改进幅度适中，但在不同结构和序列中保持一致，左心室（LV）的 Dice 分数范围为 0.940-0.967，心肌的 Dice 分数为 0.855-0.905，右心室（RV）达到 0.929，左心房（LA）为 0.920，右心房（RA）为 0.931。在 LGE 和映射图像的心肌分割中观察到了最大的增益。在未见的 LGE 长轴视图上的零样本分割任务中，该模型实现了平均 Dice 分数 0.692，展示了跨视图的泛化能力。在电影视图分类中，CMRVision 实现了最高的平均准确率（0.906），相比于文献中报告的先前方法。这些结果突显了 CMRVision 在支持多序列和多视图下稳健且可泛化的心脏 MRI 分析方面的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 93 / 2609.01318

Reliability Challenges in Diffusion Vision-Language Models

扩散视觉语言模型中的可靠性挑战

Atabuzzaman, Md., Thomas, Chris

Abstract

Diffusion-based Large Vision-Language Models (dLVLMs) have recently emerged as a compelling alternative to autoregressive (AR) LVLMs, offering advantages in parallel decoding, bidirectional context, and controllable generation. Despite rapid progress, their reliability properties remain largely uncharacterized. We present the first systematic reliability evaluation of hallucination and bias in dLVLMs, benchmarking six diffusion models against competitive AR baselines across four dimensions. Our key findings are: (1) dLVLMs reverse the yes-bias of AR models in binary visual queries; (2) they achieve competitive hallucination rates yet exhibit degraded linguistic quality; (3) they collapse to near-zero accuracy on underrepresented racial groups with opposite-polarity gender bias; and (4) they exhibit accuracy collapse in multiple-choice settings when the correct option is shorter than its distractors, associated with a length prior that emerges at the first denoising step. Tokens committed at late denoising steps with low confidence further correlate with hallucinated content, pointing to a mechanistic signal unique to diffusion generation. These patterns vary across model families, suggesting reliability is shaped by the generative paradigm together with training data.

Chinese Translation

基于扩散的大型视觉语言模型（dLVLMs）最近作为自回归（AR）视觉语言模型的一个引人注目的替代方案出现，提供了并行解码、双向上下文和可控生成等优势。尽管取得了快速进展，但它们的可靠性特性仍然未得到充分表征。我们首次系统性地评估了dLVLMs中的幻觉和偏见的可靠性，针对四个维度对六个扩散模型与竞争性的AR基线进行了基准测试。我们的主要发现包括：(1) dLVLMs在二元视觉查询中逆转了AR模型的“是”偏见；(2) 它们实现了竞争性的幻觉率，但语言质量下降；(3) 在代表性不足的种族群体中，它们的准确率接近于零，并且存在相反极性的性别偏见；(4) 当正确选项比干扰项短时，它们在多选设置中表现出准确率崩溃，这与在第一次去噪步骤中出现的长度先验有关。在去噪步骤后期以低置信度提交的标记进一步与幻觉内容相关，指向扩散生成特有的机制信号。这些模式在不同模型家族中有所不同，表明可靠性受到生成范式和训练数据的共同影响。

View on arXiv Download PDF AI Translation

cs.CV / 94 / 2609.01344

ExBind: A Controlled Diagnostic Benchmark for Visual-to-Executable Correspondence

ExBind：一种用于视觉与可执行对象对应关系的受控诊断基准

Wang, Ziqian, Cheng, Yuxiao, Xiao, Tingxiong, Suo, Jinli

Abstract

Multimodal coding and editing systems must map a visible or semantic referent to the exact executable object that can be edited. A wrong reference may select a valid but incorrect DOM node, SVG element, graph endpoint, hierarchy member, or table cell, while final execution success alone does not reveal the source of the failure. ExBind isolates this visual-to-executable correspondence layer as a controlled diagnostic benchmark between semantic localization and action execution. It samples representation-independent latent binding instances and compiles them into SVG, DOM, canvas, tree, graph, and table cases with deterministic mappings to executable references. Models output only a strict reference; the evaluator maps predictions back to latent structure and scores structural constraints without requiring reasoning traces. The release contains a 250-case broad suite, a disjoint 240-case targeted suite, and 50 paired latent groups. Qwen2.5-VL-3B achieves 98.4% candidate validity but 76.4% exact accuracy, while Qwen3-VL-4B achieves 100.0% validity and 98.8% exact accuracy. In the targeted table suite, all Qwen2.5-VL-3B residual errors are valid correct-row/wrong-column selections. Candidate-order perturbations change case-level outcomes while preserving this error pattern. ExBind is designed for controlled diagnosis rather than population-scale ranking or end-to-end editing evaluation. Code and benchmark records are available at https://github.com/Daerwang2020/Exbind and https://huggingface.co/datasets/Ziqianwwww/ExBind.

Chinese Translation

多模态编码和编辑系统必须将可见或语义参照映射到可以被编辑的确切可执行对象。错误的参照可能选择一个有效但不正确的 DOM 节点、SVG 元素、图形端点、层次成员或表格单元格，而最终执行的成功与否并不能揭示失败的来源。ExBind 将这种视觉与可执行对象的对应关系层隔离出来，作为语义定位与动作执行之间的受控诊断基准。它采样与表示无关的潜在绑定实例，并将其编译成具有确定性映射到可执行参照的 SVG、DOM、画布、树、图形和表格案例。模型仅输出严格的参照；评估者将预测映射回潜在结构，并在不需要推理痕迹的情况下对结构约束进行评分。该发布包含一个 250 个案例的广泛套件，一个不相交的 240 个案例的目标套件，以及 50 对潜在组。Qwen2.5-VL-3B 实现了 98.4% 的候选有效性，但确切准确率为 76.4%；而 Qwen3-VL-4B 实现了 100.0% 的有效性和 98.8% 的确切准确率。在目标表格套件中，所有 Qwen2.5-VL-3B 的残余错误都是有效的正确行/错误列选择。候选顺序的扰动改变了案例级结果，同时保持了这种错误模式。ExBind 的设计旨在用于受控诊断，而非大规模人群排名或端到端编辑评估。代码和基准记录可在 https://github.com/Daerwang2020/Exbind 和 https://huggingface.co/datasets/Ziqianwwww/ExBind 获取。

View on arXiv Download PDF AI Translation

cs.CV / 95 / 2609.01368

Accurate Reconstruction of Gas Turbine Blade Geometry Using 3D/2D Rigid Registration and CT View Optimization

基于3D/2D刚性配准和CT视图优化的燃气涡轮叶片几何形状精确重建

Valtchanov, Hristo, Piché, Nicolas, Brailovski, Vladimir, Byers, Justin, Désrosiers, Catherine, Guibault, François

Abstract

Non-destructive X-ray and computed tomography (CT) testing are essential for ensuring the dimensional accuracy of manufactured components with complex internal structures, such as the cooling channels in gas turbine blades, which directly affect thermal performance and service life. This study presents a multipart 3D-2D rigid registration approach for aligning CAD models with X-ray projections as an alternative to CT reconstruction for part inspection and measurement. A greedy registration algorithm sequentially aligns the blade's exterior before registering its internal components by maximizing the mutual information between simulated and acquired X-ray images. This stepwise approach reduces problem complexity and improves alignment accuracy. View angles are optimized using a greedy method that iteratively selects angles to minimize dimensional measurement errors. The results indicate that a small number of oblique views provides the best accuracy, although a broad range of angles yields acceptable results. The method achieves subpixel registration accuracy, with errors below one-fifth of the magnified detector-pixel pitch. Image noise and defects reduce registration precision, but direct registration in projection space mitigates these effects compared with CT reconstruction. Appropriate view selection can therefore preserve acceptable subpixel accuracy in the presence of image noise and defects.

Chinese Translation

无损X射线和计算机断层扫描（CT）检测对于确保具有复杂内部结构的制造部件的尺寸精度至关重要，例如燃气涡轮叶片中的冷却通道，这直接影响热性能和使用寿命。本研究提出了一种多部件3D-2D刚性配准方法，用于将CAD模型与X射线投影对齐，作为部件检测和测量的CT重建替代方案。一种贪婪配准算法首先顺序对齐叶片的外部，然后通过最大化模拟和获取的X射线图像之间的互信息来注册其内部组件。这种逐步的方法降低了问题复杂性并提高了对齐精度。视角通过贪婪方法进行优化，该方法迭代选择角度以最小化尺寸测量误差。结果表明，少量斜视图提供了最佳精度，尽管广泛的角度范围也能获得可接受的结果。该方法实现了亚像素级的配准精度，误差低于放大探测器像素间距的五分之一。图像噪声和缺陷降低了配准精度，但在投影空间中的直接配准相比CT重建减轻了这些影响。因此，适当的视图选择可以在存在图像噪声和缺陷的情况下保持可接受的亚像素精度。

View on arXiv Download PDF AI Translation

cs.CV / 96 / 2609.01370

Diffusion Based Unpaired Data Learning for Inverse Problems

基于扩散的无配数据学习在逆问题中的应用

Bao, Chenglong, Dang, Yiming, Duan, Chenguang, Jiao, Yuling, Sun, Defeng

Abstract

Data is important in many deep learning-based inverse problem solvers. However, obtaining sufficient paired data in many scenarios remains highly challenging, while unpaired data is cheap. To maximize data utilization, this paper proposes LUD-DIF, a diffusion-based approach for solving inverse problems with unpaired data. Starting from the evidence lower bound (ELBO) of the joint distribution, we decouple it into two independent diffusion processes under the weak-coupling assumption. The method provides theoretical support from a variational inference perspective, derives the loss function, quantitatively analyzes the error bound introduced by the assumption, and offers a theorem-motivated heuristic for hyperparameter selection. Experimental results demonstrate that LUD-DIF achieves outstanding performance on multiple image inverse problems, validating its effectiveness and generalization capability in unpaired inverse problem settings.

Chinese Translation

数据在许多基于深度学习的逆问题求解器中至关重要。然而，在许多场景中获取足够的配对数据仍然非常具有挑战性，而无配数据则相对便宜。为了最大化数据的利用，本论文提出了LUD-DIF，一种基于扩散的方法，用于解决带有无配数据的逆问题。该方法从联合分布的证据下界（ELBO）出发，在弱耦合假设下将其解耦为两个独立的扩散过程。该方法从变分推断的角度提供了理论支持，推导了损失函数，定量分析了假设引入的误差界限，并为超参数选择提供了基于定理的启发式方法。实验结果表明，LUD-DIF在多个图像逆问题上表现出色，验证了其在无配逆问题设置中的有效性和泛化能力。

View on arXiv Download PDF AI Translation

cs.CV / 97 / 2609.01375

IntroConformal: Conformal Factuality Guarantees for Large Vision-Language Models via Introspective Signals

IntroConformal：通过内省信号为大型视觉-语言模型提供符合性事实保证

Atabuzzaman, Md., Alexander, Christian, Thomas, Chris

Abstract

Large Vision-Language Models (LVLMs) have achieved strong multimodal performance, yet ensuring the factual correctness of generated content remains challenging. Existing methods that provide statistical guarantees on factuality typically rely on external verifiers or generation-time confidence signals, which introduce auxiliary dependencies or often fail for confident but incorrect outputs. We argue that reliable factuality control can instead be achieved through introspective signals derived from the model itself. We introduce IntroConformal, a training-free Conformal Risk Control (CRC) framework that provides finite-sample, distribution-free factuality guarantees. We first instantiate it with layer-wise semantic stability, a conformity score derived from hidden-state representations, and then propose verification probability, a stronger score capturing the model's self-administered judgment on claim factuality. Across multiple LVLM architectures, IntroConformal satisfies the conformal risk guarantee while substantially reducing abstention and achieving competitive or superior claim-level discrimination relative to external verifier-based baselines.

Chinese Translation

大型视觉-语言模型（LVLMs）在多模态性能上取得了显著成果，但确保生成内容的事实正确性仍然具有挑战性。现有的提供事实性统计保证的方法通常依赖于外部验证者或生成时的置信信号，这引入了辅助依赖关系，或者常常对自信但不正确的输出失效。我们认为，可靠的事实控制可以通过源自模型自身的内省信号来实现。我们提出了IntroConformal，一个无训练的符合风险控制（Conformal Risk Control, CRC）框架，提供有限样本、无分布假设的事实保证。我们首先通过层级语义稳定性实例化它，这是一种源自隐藏状态表示的符合性得分，然后提出验证概率，这是一种更强的得分，捕捉模型对声明事实性的自我评判。在多个LVLM架构中，IntroConformal满足符合风险保证，同时显著减少放弃率，并在声明级别的区分能力上相较于基于外部验证者的基线实现了竞争性或更优的表现。

View on arXiv Download PDF AI Translation

cs.CV / 98 / 2609.01390

Multimodal RGB-Infrared Combination for UAV-Based Wildfire Segmentation: A Comparative Study on FLAME3

基于无人机的多模态RGB-红外组合用于野火分割：FLAME3上的比较研究

Kovaleski, Matheus F., Garrote, Luís, Premebida, Cristiano, Mendes, Jérôme, Paulo, João Ruivo

Abstract

Unmanned Aerial Vehicles (UAVs) have emerged as a promising platform for firefighting operations due to their flexibility, low operational cost, and ability to acquire high-resolution imagery in locations that may be difficult or dangerous to access using conventional methods. Recent advances in deep learning have significantly improved the capabilities of UAV-based wildfire monitoring systems. The present work investigates RGB-infrared fusion for binary wildfire segmentation on the FLAME3 dataset. In this Study, RGB and Infrared baselines are compared with three representative fusion strategies across three segmentation architectures, including U-Net, DeepLabV3+, and SegFormer. The key motivation of this work is to analyze the contribution of each modality, evaluate the impact of fusion timing, and examine how different network architectures exploit multimodal information for UAV wildfire delineation. The findings indicate that thermal information plays a dominant role in UAV segmentation and that feature-level multimodal fusion combined with transformer-based architectures offers the most promising direction for future research.

Chinese Translation

无人驾驶飞行器（UAV）因其灵活性、低运营成本以及能够在传统方法难以或危险进入的地点获取高分辨率图像，已成为消防作业的有前景的平台。近年来，深度学习的进步显著提升了基于UAV的野火监测系统的能力。本研究探讨了在FLAME3数据集上进行二元野火分割的RGB-红外融合。在本研究中，RGB和红外基线与三种代表性的融合策略在三种分割架构（包括U-Net、DeepLabV3+和SegFormer）中进行了比较。本研究的主要动机是分析每种模态的贡献，评估融合时机的影响，并考察不同网络架构如何利用多模态信息进行UAV野火划分。研究结果表明，热信息在UAV分割中起主导作用，而特征级多模态融合结合基于变换器的架构为未来研究提供了最有前景的方向。

View on arXiv Download PDF AI Translation

cs.CV / 99 / 2609.01392

Scale-based Approach for Active Wildfire Segmentation on Satellite Imagery

基于尺度的方法进行卫星图像中活跃野火的分割

Kovaleski, Matheus F., Premebida, Cristiano, Paulo, João Ruivo

Abstract

Active wildfire mapping from satellite imagery is challenging due to the sparse and highly imbalanced nature of fire pixels, especially in early-stage or low-density fire observations. This work investigates the use of multispectral Landsat-8 imagery for active-fire segmentation under multi-scale wildfire size conditions. We propose a data-driven protocol to characterize fire-region size distributions through connected-component analysis and an interquartile range criterion, enabling the evaluation of model robustness across different local fire-region densities. Three segmentation architectures, U-Net, DeepLabV3+, and SegFormer, are evaluated under different SWIR-based spectral configurations. Results show that U-Net achieves the strongest robustness across the evaluated conditions, SegFormer provides competitive performance, and DeepLabV3+ tends to produce conservative predictions with reduced recall. Across architectures, SWIR2 consistently achieves the strongest or near-best results, highlighting its importance for active-fire segmentation in Landsat-8 imagery. These findings suggest that both spectral band selection and architectural design are critical for robust satellite-based active wildfire mapping trained on low active fire-pixel density images.

Chinese Translation

从卫星图像中进行活跃野火的映射具有挑战性，尤其是在早期阶段或低密度火灾观察中，火像素稀疏且高度不平衡。本研究探讨了在多尺度野火规模条件下，使用多光谱Landsat-8图像进行活火分割。我们提出了一种数据驱动的协议，通过连通组件分析和四分位数范围标准来表征火区规模分布，从而评估模型在不同局部火区密度下的鲁棒性。评估了三种分割架构：U-Net、DeepLabV3+和SegFormer，在不同的基于短波红外（SWIR）的光谱配置下。结果表明，U-Net在评估条件下表现出最强的鲁棒性，SegFormer提供了具有竞争力的性能，而DeepLabV3+则倾向于产生保守的预测，导致召回率降低。在所有架构中，SWIR2始终实现了最强或接近最佳的结果，突显了其在Landsat-8图像中进行活火分割的重要性。这些发现表明，光谱波段选择和架构设计对于基于卫星的活跃野火映射在低活跃火像素密度图像上的鲁棒性至关重要。

View on arXiv Download PDF AI Translation

cs.CV / 100 / 2609.01423

MegaStyle++: Scaling Image Style Space through Hierarchical Style Definition

MegaStyle++：通过层次化风格定义扩展图像风格空间

Gao, Junyao, Liu, Sibo, Li, Jiaxing, Sun, Yanan, Zhang, Weidong, Zhao, Cairong, Zhang, Jun

Abstract

Image style is a highly abstract, human-constructed concept shaped by a range of visual factors and intrinsically entangled with content, yet a unified and explicit definition of image style remains lacking. In this work, we first discuss the fundamental question of what is style and then propose a hierarchical style definition that describes image style from an overall style identity to fine-grained visual attributes, providing a more structured, transferable, and interpretable style representation. Based on this definition, we refine the style annotation pipeline of MegaStyle and construct MegaStyle++-8M, a large-scale style dataset containing 150K overall style identities, 1M fine-grained style prompts, and 8M stylized images. Extensive analyses demonstrate that our hierarchical definition substantially expands the style space in both diversity and semantic breadth, while precisely capturing intrinsic visual style of reference images. The dataset and code will be updated at https://github.com/Tencent/MegaStyle, we hope MegaStyle++ provides a scalable foundation for studying and modeling diverse image styles.

Chinese Translation

图像风格是一个高度抽象的人为构建概念，受到多种视觉因素的影响，并与内容内在交织，然而，关于图像风格的统一和明确的定义仍然缺乏。在本研究中，我们首先讨论了风格的基本问题，然后提出了一种层次化的风格定义，从整体风格身份到细粒度视觉属性描述图像风格，提供了更结构化、可转移和可解释的风格表示。基于这一定义，我们优化了MegaStyle的风格注释流程，并构建了MegaStyle++-8M，这是一个大规模的风格数据集，包含150K个整体风格身份、1M个细粒度风格提示和8M个风格化图像。广泛的分析表明，我们的层次化定义在多样性和语义广度上显著扩展了风格空间，同时准确捕捉了参考图像的内在视觉风格。数据集和代码将更新至https://github.com/Tencent/MegaStyle，我们希望MegaStyle++为研究和建模多样化图像风格提供一个可扩展的基础。

View on arXiv Download PDF AI Translation

cs.CV / 101 / 2609.01426

Semantic-Guided Multimodal Preprocessing for Vision Transformer-Based Clear Cell Renal Cell Carcinoma Grading

基于语义引导的多模态预处理方法用于视觉变换器（Vision Transformer）驱动的透明细胞肾细胞癌分级

Javadian, Fatemeh, Chen, Zhu, Aminparast, Zahra, Stegmaier, Johannes

Abstract

Clear cell renal cell carcinoma (CCRCC) grading is essential for treatment planning, yet existing approaches either analyze patch-level images directly or focus solely on nuclei-level classification, without linking to final tumor grading. We propose a semantic-guided multimodal preprocessing method that integrates nuclei classification maps from existing pre-trained models with RGB histopathology images for Vision Transformer (ViT)-based CCRCC grading. Our approach employs classification map channel concatenation and multiplicative modulation, with optimized overlays to leverage nuclei grading information, while preserving RGB textural features. Evaluation of multiple preprocessing strategies demonstrates that semantic-guided enhancement achieves 0.916 balanced accuracy, outperforming RGB-only baseline (0.707) and max-voting aggregation from prior studies (0.427). Sensitivity analysis reveals that this 21 percentage point improvement over baseline persists even under simulated perturbation at rates matching current state-of-the-art nuclei classification model error thresholds, suggesting both effective semantic utilization and practical robustness. These findings show that preprocessing-based multimodal fusion can leverage the diagnostic potential of existing imperfect nuclei classifiers, effectively bridging previously isolated fine-grained nuclear-level analysis with coarse-grained ViT-based patch classification. Per-class recall was consistent across grades (0.93, 0.91, 0.91), indicating that gains are not concentrated in the majority class. Because the sensitivity analysis perturbs ground-truth maps rather than predictions from an actual nuclei model, this result characterizes robustness under simulated error rather than deployment with a real upstream model, which remains for future work.

Chinese Translation

透明细胞肾细胞癌（CCRCC）分级对于治疗规划至关重要，但现有方法要么直接分析补丁级图像，要么仅关注细胞核级分类，而未能与最终肿瘤分级相联系。我们提出了一种语义引导的多模态预处理方法，该方法将现有预训练模型的细胞核分类图与RGB组织病理图像结合，用于基于视觉变换器（ViT）的CCRCC分级。我们的方法采用分类图通道连接和乘法调制，并通过优化叠加来利用细胞核分级信息，同时保留RGB纹理特征。对多种预处理策略的评估表明，语义引导增强实现了0.916的平衡准确率，优于仅使用RGB的基线（0.707）和以往研究的最大投票聚合（0.427）。敏感性分析显示，这一相较基线提高21个百分点的改进在模拟扰动下仍然存在，扰动率与当前最先进的细胞核分类模型的错误阈值相匹配，表明有效的语义利用和实际的鲁棒性。这些发现表明，基于预处理的多模态融合可以利用现有不完美细胞核分类器的诊断潜力，有效地将之前孤立的细粒度核级分析与粗粒度的基于ViT的补丁分类连接起来。各类别的召回率在不同分级中保持一致（0.93, 0.91, 0.91），表明增益并未集中于多数类。由于敏感性分析扰动的是真实标记图而非实际细胞核模型的预测，因此该结果表征了在模拟错误下的鲁棒性，而非与真实上游模型的部署，这仍需未来的研究来解决。

View on arXiv Download PDF AI Translation

cs.CV / 102 / 2609.01427

Pix2Rep-v2: Data-Efficient Representation Learning for Dense Medical Imaging Applications

Pix2Rep-v2：用于密集医学成像应用的数据高效表示学习

Sifaoui, S., Angelini, E., Toupin, S., Pezel, T., Folgoc, L. Le

Abstract

Dense self-supervised learning (SSL) is a powerful paradigm for learning without annotations the local descriptors required to solve dense medical imaging tasks. We present Pix2Rep-v2, a framework for SSL of pixel- and voxel-level representations suitable for few-shot downstream applications. Pix2Rep-v2 addresses the main challenges of dense SSL by leveraging a redundancy reduction objective at the pixel-level with a principle of equivariance of dense representations, that scales efficiently to 3D or wide field-of-view applications. We evaluate our method on four datasets, across multiple tasks, multiple modalities and anatomical structures using multiple backbones in 2D and 3D, and under various data regimes. As an alternative to linear probing or full fine-tuning on the downstream task, we also propose an in-context variant, without downstream training, based on a dense prototype approach. Pix2Rep-v2 shows substantially higher data-efficiency in few-shot scenarios compared to fully supervised baselines, and is competitive with the state-of-the-art e.g., +9.3 Dice points in one-shot segmentation on the M&Ms-2 dataset. Our code and pre-trained models are publicly available at https://github.com/BioMedTP/pix2rep-v2.

Chinese Translation

密集自监督学习（SSL）是一种强大的范式，能够在没有注释的情况下学习解决密集医学成像任务所需的局部描述符。我们提出了Pix2Rep-v2，这是一个适用于少样本下游应用的像素和体素级表示的SSL框架。Pix2Rep-v2通过利用像素级的冗余减少目标和密集表示的等变原则，解决了密集SSL的主要挑战，该方法能够有效扩展到3D或广视野应用。我们在四个数据集上评估了我们的方法，涵盖多个任务、多个模态和解剖结构，使用多种主干网络在2D和3D下进行测试，并在各种数据条件下进行评估。作为线性探测或下游任务完全微调的替代方案，我们还提出了一种基于密集原型方法的上下文变体，无需下游训练。与完全监督的基线相比，Pix2Rep-v2在少样本场景中显示出显著更高的数据效率，并且在一-shot分割任务中在M&Ms-2数据集上与最先进的技术具有竞争力，例如提高了9.3个Dice点。我们的代码和预训练模型已公开发布在https://github.com/BioMedTP/pix2rep-v2。

View on arXiv Download PDF AI Translation

cs.CV / 103 / 2609.01433

Gaussian Core LoRA: Distribution-Aware Dynamic Adaptation for Broad Concept Erasure

高斯核心LoRA：面向分布的动态适应以实现广泛概念的抹除

Gong, Qinghui, Chen, Xunlei, Zhang, Yu-Xuan, Meng, Hua, Zhou, Zhengchun

Abstract

Concept erasure aims to suppress unsafe, privacy-sensitive, or undesirable generations in text-to-image diffusion models while preserving benign semantics, visual quality, and deployment efficiency. Existing adapter-based methods, such as Low-Rank Adaptation (LoRA), typically freeze the diffusion backbone and learn lightweight parameter updates to steer generation away from target semantics. However, these methods usually assign a static semantic erasure direction to each target concept. This assumption is overly coarse for broad and complex target concepts, since a concept often contains multiple latent semantic prototypes involving different objects, scenes, or relations, and requires different local erasure directions. A single LoRA update averages these heterogeneous erasure demands, leading to under-erasure on difficult prototypes and over-editing of nearby benign semantics. To address this limitation, we propose Gaussian Core LoRA, a distribution-aware low-rank adaptation framework. It fits a Gaussian mixture model in the prompt feature space to estimate latent semantic prototypes within the target concept. During inference, each input prompt is projected into this feature space to compute its Gaussian posterior responsibilities, which condition the core generator to produce a prompt-specific, norm-bounded residual reconfiguration of the shared LoRA rank space. This enables prototype-adaptive erasure with a single lightweight adapter. Compared with the strongest baseline on each metric, Gaussian Core LoRA reduces average Attack Success Rate (ASR) by 7.95%, lowers COCO Fr'echet Inception Distance (FID) by 14.72%, and improves CLIP Score by 4.98%. Further experiments show robustness to adversarial prompts, scalability to multi-identity and multi-style erasure, and compatibility with SDXL and FLUX.

Chinese Translation

概念抹除旨在抑制文本到图像扩散模型中不安全、隐私敏感或不希望生成的内容，同时保留良性的语义、视觉质量和部署效率。现有的基于适配器的方法，如低秩适配（Low-Rank Adaptation, LoRA），通常会冻结扩散主干网络，并学习轻量级的参数更新，以引导生成远离目标语义。然而，这些方法通常为每个目标概念分配一个静态的语义抹除方向。这个假设对于广泛且复杂的目标概念来说过于粗糙，因为一个概念通常包含多个潜在的语义原型，涉及不同的对象、场景或关系，并且需要不同的局部抹除方向。单一的LoRA更新会对这些异质的抹除需求进行平均，导致在困难原型上的抹除不足以及对附近良性语义的过度编辑。为了解决这一局限性，我们提出了高斯核心LoRA，这是一种面向分布的低秩适配框架。它在提示特征空间中拟合高斯混合模型，以估计目标概念内的潜在语义原型。在推理过程中，每个输入提示被投影到该特征空间中，以计算其高斯后验责任，这些责任条件化核心生成器以产生特定于提示的、范数受限的共享LoRA秩空间的残差重配置。这使得通过单个轻量级适配器实现原型自适应抹除。与每个指标上最强的基线相比，高斯核心LoRA将平均攻击成功率（Attack Success Rate, ASR）降低了7.95%，将COCO Fr'echet Inception Distance（FID）降低了14.72%，并将CLIP评分提高了4.98%。进一步的实验表明其对对抗性提示的鲁棒性、对多身份和多风格抹除的可扩展性，以及与SDXL和FLUX的兼容性。

View on arXiv Download PDF AI Translation

cs.CV / 104 / 2609.01470

RadMatch: Auditable Radiology Report Evaluation via Finding-Level Matching

RadMatch：通过发现级匹配进行可审计的放射学报告评估

Corbière, Charles, Machado, Léo, Charley, Aubin, Callard, Baptiste, Manceron, Pierre, Dancette, Corentin

Abstract

As AI systems are increasingly used to draft radiology reports, reliably evaluating their clinical quality remains a critical challenge. Large language model (LLM)-based metrics are now the best-correlated with radiologist judgment, yet they output a single opaque score that neither a clinician nor a model builder can easily interpret or audit. We introduce RadMatch, a multi-stage, LLM-based metric that decomposes report comparison into a structured finding-level matching with significance-aware scoring and error characterization across seven clinical attribute dimensions (status, location, severity, morphology, certainty, longitudinal comparison, and measurement). The main score is the actionable-error count, both interpretable and auditable. Candidate findings are graded correct, partial, or incorrect, and unmatched findings are counted as missed or hallucinated. Triage and actionable safety recall/precision and per-subset views add complementary, deployment-oriented lenses. Across two expert benchmarks, RadMatch is the most clinically aligned metric, matching inter-radiologist agreement on ReXVal and more than doubling the best prior metric on the harder RadEvalExpert. Relying only on few-shot prompting, it is designed to extend to other modalities and anatomies. We will release RadMatch as open-source code with an interactive dashboard for inspecting results.

Chinese Translation

随着人工智能系统越来越多地用于撰写放射学报告，可靠地评估其临床质量仍然是一个关键挑战。基于大型语言模型（LLM）的指标现在与放射科医生的判断相关性最佳，但它们输出的单一不透明分数既不易于临床医生也不易于模型构建者解释或审计。我们介绍了RadMatch，这是一种多阶段的基于LLM的指标，它将报告比较分解为结构化的发现级匹配，并在七个临床属性维度（状态、位置、严重程度、形态、确定性、纵向比较和测量）上进行显著性感知评分和错误特征化。主要分数是可操作的错误计数，既可解释又可审计。候选发现被评为正确、部分正确或不正确，而未匹配的发现则被计为漏检或虚构。分流和可操作的安全召回/精度以及每个子集视图增加了互补的、面向部署的视角。在两个专家基准测试中，RadMatch是与临床最一致的指标，在ReXVal上匹配放射科医生间的一致性，并在更具挑战性的RadEvalExpert上将最佳先前指标的表现提高了两倍以上。仅依赖少量提示，它被设计为能够扩展到其他模态和解剖结构。我们将以开源代码的形式发布RadMatch，并提供一个交互式仪表板以检查结果。

View on arXiv Download PDF AI Translation

cs.CV / 105 / 2609.01479

CameraEditor: Camera-Controlled Image Editing via Video-Prior Sequential Modeling

CameraEditor：基于视频先验序列建模的相机控制图像编辑

Shen, Xin, Jia, Chengyou, Xing, Keshuo, Zhu, Zifeng, Xia, Changliang, Ping, Bowen, Dang, Zhuohang, Qian, Hangwei, Luo, Minnan

Abstract

Beyond semantic content, camera parameters play a pivotal role in dictating the geometric perspective and appearance of any given image. While recent image editing models excel at semantic and stylistic manipulation, they struggle with explicit camera parameter control. When handling large perspective shifts, instruction-driven models face a dilemma: they either suffer from structural tearing or generate conservative outputs that ignore geometric instructions. To address this, we introduce CameraEditor, a framework that reformulates camera-controlled editing from a spatial problem into a temporal sequence prediction task. By leveraging the temporal coherence of video diffusion models, our approach integrates an explicit geometric perception module with a dynamic reference routing mechanism. This allows us to construct geometrically rigorous visual reference pairs via dynamic panorama cropping, overcoming the ambiguity of text-based instructions. Furthermore, CameraEditor strategically inserts intermediate transition frames to decompose large perspective shifts, providing a robust temporal buffer that preserves content identity and spatial coherence. We construct a training dataset of 5,760 instances. As an independent contribution, we introduce CamEditor-Bench, a model-agnostic evaluation suite of 462 test cases. Extensive experiments demonstrate that CameraEditor achieves state-of-the-art camera control precision and source identity preservation, outperforming existing methods.

Chinese Translation

除了语义内容外，相机参数在决定任何给定图像的几何视角和外观方面发挥着关键作用。尽管最近的图像编辑模型在语义和风格操作方面表现出色，但它们在显式相机参数控制方面却面临挑战。在处理大幅度透视变化时，基于指令的模型面临两难：要么遭受结构撕裂，要么生成忽视几何指令的保守输出。为了解决这一问题，我们提出了CameraEditor，一个将相机控制编辑重新构建为时间序列预测任务的框架。通过利用视频扩散模型的时间一致性，我们的方法将显式几何感知模块与动态参考路由机制相结合。这使我们能够通过动态全景裁剪构建几何严谨的视觉参考对，克服基于文本指令的模糊性。此外，CameraEditor战略性地插入中间过渡帧，以分解大幅度透视变化，提供一个强大的时间缓冲区，保持内容身份和空间一致性。我们构建了一个包含5760个实例的训练数据集。作为独立贡献，我们引入了CamEditor-Bench，一个与模型无关的462个测试案例的评估套件。大量实验表明，CameraEditor在相机控制精度和源身份保持方面达到了最先进的水平，超越了现有方法。

View on arXiv Download PDF AI Translation

cs.CV / 106 / 2609.01511

Benchmarking Spatial, Spectral, and Self-Supervised Cues for Face Forgery Detection under Realistic Degradation

在现实退化条件下面部伪造检测的空间、光谱和自监督线索基准测试

Cunha, Lucas, Sotomaior, Lucas, Gasperin, Lucas, Caldas, Beatriz, Pianovski, Eduardo, Laroca, Rayson

Abstract

Face forgery detectors often achieve strong results on controlled benchmarks, but their reliability under realistic image degradations remains limited. This paper presents a standardized benchmark for face forgery detection using the Multi-Dimensional Face Forgery Image (MFFI) dataset and evaluates performance on both clean and degraded test partitions. We compare six model families, including convolutional networks, transformer-based models, and a frozen self-supervised DINOv3 backbone, across spatial, spectral, and hybrid input representations. The results show that clean-set performance is not a reliable indicator of robustness under compression, resizing, and blurring. Xception with RGB obtains the best clean performance, reaching 0.884 mean ROC-AUC, but degrades substantially on the harder partition. In contrast, frozen DINOv3 achieves the strongest degraded-set result, with 0.726 mean ROC-AUC, while training only a linear classification head. The representation analysis indicates that Fourier-domain cues are most useful when combined with RGB information, whereas purely spectral inputs consistently underperform spatial representations. Qualitative attribution maps further suggest that convolutional detectors focus on localized artifacts, while DINOv3 relies on broader facial structure. These findings reinforce the need for degraded evaluation protocols and highlight self-supervised visual representations as a promising direction for robust face forgery detection. Our source code is publicly available at https://github.com/lucasdocunha/FaceForgery-Benchmark/.

Chinese Translation

面部伪造检测器在受控基准测试中通常表现良好，但在现实图像退化下的可靠性仍然有限。本文提出了一个标准化的面部伪造检测基准，使用多维面部伪造图像（Multi-Dimensional Face Forgery Image, MFFI）数据集，并评估在干净和退化测试分区上的性能。我们比较了六种模型家族，包括卷积网络、基于变换器的模型以及一个冻结的自监督DINOv3主干，涵盖空间、光谱和混合输入表示。结果表明，干净集的性能并不是压缩、缩放和模糊下鲁棒性的可靠指标。Xception在RGB输入下获得了最佳的干净性能，达到0.884的平均ROC-AUC，但在更困难的分区上显著下降。相比之下，冻结的DINOv3在退化集上取得了最强的结果，平均ROC-AUC为0.726，同时仅训练了一个线性分类头。表示分析表明，当与RGB信息结合时，傅里叶域线索最为有效，而纯光谱输入的表现始终不如空间表示。定性归因图进一步表明，卷积检测器关注局部伪影，而DINOv3则依赖于更广泛的面部结构。这些发现强调了退化评估协议的必要性，并突出了自监督视觉表示作为鲁棒面部伪造检测的有前景方向。我们的源代码已公开在 https://github.com/lucasdocunha/FaceForgery-Benchmark/.

View on arXiv Download PDF AI Translation

cs.CV / 107 / 2609.01515

TempCloze: Can Video-LLMs Identify the Missing Middle?

TempCloze：视频大语言模型能否识别缺失的中间部分？

Pei, Wenqi, Zhao, Henry Hengyuan, Liu, Yilai, Meng, Jiahao, Chen, Han, Wang, Ziyu, Du, Hongyang

Abstract

Temporal reasoning benchmarks for Video-LLMs are often mediated by language, leaving room for linguistic shortcuts from option wording, answer correlations, or language priors. To reduce such shortcuts, we introduce TempCloze, a video cloze benchmark for evaluating visual temporal reasoning in Video-LLMs. Given the beginning and ending clips of a video, models must identify the true missing middle from four candidates. TempCloze contains 1,521 carefully filtered videos from seven sources, mainly long-take and egocentric videos. We construct same-source distractors along three dimensions: Semantic asks what event should happen, Alignment probes when it should occur, and Progression tests how it should unfold, while shared scenes and objects reduce appearance cues. Our evaluation of 10 proprietary and 21 open-source Video-LLMs reveals Alignment as the primary bottleneck: models often recognize plausible semantic content and local event progression but struggle with temporal alignment. We further conduct error pattern and behavioral sensitivity analyses on TempCloze-Mixed and TempCloze-Hard with four representative models to examine where errors arise and how candidate order, context direction, visible span, frame density, and test-time scaling influence model choices.

Chinese Translation

视频大语言模型（Video-LLMs）的时间推理基准通常受到语言的影响，这为选项措辞、答案相关性或语言先验提供了语言捷径。为了减少这种捷径，我们引入了TempCloze，这是一个用于评估视频大语言模型视觉时间推理的视频填空基准。给定视频的开头和结尾片段，模型必须从四个候选中识别出真实缺失的中间部分。TempCloze包含来自七个来源的1,521个经过仔细筛选的视频，主要是长镜头和自我中心的视频。我们在三个维度上构建了同源干扰项：语义（Semantic）询问应该发生什么事件，时间对齐（Alignment）探测何时发生，以及进展（Progression）测试如何展开，同时共享场景和物体减少外观线索。我们对10个专有和21个开源的视频大语言模型的评估揭示了时间对齐是主要瓶颈：模型通常能够识别出合理的语义内容和局部事件进展，但在时间对齐方面遇到困难。我们进一步对TempCloze-Mixed和TempCloze-Hard进行错误模式和行为敏感性分析，使用四个代表性模型来检查错误产生的原因，以及候选顺序、上下文方向、可见范围、帧密度和测试时间缩放如何影响模型选择。

View on arXiv Download PDF AI Translation

cs.CV / 108 / 2609.01516

DualDiff3D: Dual Structure-Appearance Diffusion Priors for Reliability-Enhanced 3D Gaussian Splatting

DualDiff3D：用于增强可靠性的双结构-外观扩散先验的三维高斯点云技术

Wang, Qian, Wang, Yu, Li, Weiqi, Cheng, Xinhua, Meng, Xiandong, Wang, Ronggang, Zhang, Jian

Abstract

While 3D Gaussian Splatting (3DGS) has revolutionized 3D reconstruction and novel-view synthesis, scenarios with limited input views often lead to poor reconstruction quality and artifacts in rendered novel views. Recent efforts attempt to utilize powerful diffusion priors, yet they typically process rendered and reference views concatenated along an additional dimension in a single network. These methods overlook an inherent nature that different views should maintain appearance similarity but differ in structure due to view shifts, leading to blur caused by conflicts between the two properties. In this paper, we propose DualDiff, a novel pipeline that leverages dual diffusion priors with a Structure-Appearance Attention (SAA) module to introduce reference guidance for refining low-quality novel views rendered from flawed 3D representations. Specifically, we retain one diffusion branch to focus on extracting structural information from the low-quality novel views, while introducing another branch to ensure appearance consistency with reference views. Furthermore, we present a 3D reconstruction framework named DualDiff3D, which integrates a reliability-enhanced Render-Refine-Optimize (RRO) loop to progressively and robustly incorporate the refined novel views, yielding more accurate 3DGS. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods even in the inference-only setting, with further performance gains achievable through training. Our code and pre-trained weights are available at https://github.com/Akaneqwq/DualDiff3D.

Chinese Translation

尽管三维高斯点云技术（3D Gaussian Splatting, 3DGS）已在三维重建和新视角合成方面取得了革命性进展，但在输入视角有限的情况下，往往会导致重建质量差和渲染新视角时出现伪影。最近的研究尝试利用强大的扩散先验，但通常是在单一网络中处理沿附加维度连接的渲染视图和参考视图。这些方法忽视了一个固有特性，即不同视角应保持外观相似性，但由于视角变化而在结构上存在差异，这导致了两者属性之间的冲突而产生模糊。在本文中，我们提出了DualDiff，这是一种新颖的管道，利用双重扩散先验和结构-外观注意力（Structure-Appearance Attention, SAA）模块，引入参考引导以改善从有缺陷的三维表示中渲染的低质量新视角。具体而言，我们保留一个扩散分支，专注于从低质量新视角中提取结构信息，同时引入另一个分支以确保与参考视角的一致性。此外，我们提出了一个名为DualDiff3D的三维重建框架，集成了增强可靠性的渲染-优化-细化（Render-Refine-Optimize, RRO）循环，以逐步且稳健地结合经过精细化的新视角，从而获得更准确的3DGS。大量实验表明，我们的方法在仅推理设置下超越了最先进的方法，并且通过训练可以实现进一步的性能提升。我们的代码和预训练权重可在 https://github.com/Akaneqwq/DualDiff3D 获取。

View on arXiv Download PDF AI Translation

cs.CV / 109 / 2609.01530

Revisiting Cross-View Completion: Self-Supervised Pre-Training via Reconstruction Error Comparison

重新审视跨视图补全：通过重建误差比较进行自监督预训练

Loiseau, Thibaut, Bourmaud, Guillaume, Lepetit, Vincent

Abstract

Self-supervised pre-training via cross-view completion learns strong features for 3D vision from co-visible regions of image pairs. However, the reference view provides little information for reconstructing non-co-visible patches, implicitly yielding a monocular training signal in these regions. We introduce Gekko, which turns this limitation into a useful signal. The relative improvement of the cross-view reconstruction error over a masked-autoencoder error is a self-supervised proxy for co-visibility: large improvements indicate co-visible regions, negligible ones non-co-visible areas. Gekko is a network, trained from scratch, that jointly performs cross-view completion, masked autoencoding, and per-pixel prediction of this relative improvement, providing an additional binocular signal for all masked regions without any ground-truth 3D annotation. Under identical architectures and training data, Gekko consistently outperforms CroCo on zero-shot correspondence estimation, relative pose estimation, and pointmap regression, with up to 6 times higher accuracy at the strictest relative-pose threshold and a 22% drop in end-point error on ETH3D. The extra channel it learns is itself a strong co-visibility detector on unseen scenes, and Gekko's frozen features outperform released cross-view backbones of comparable or larger size. It can also be trained directly from raw videos with a simple stride-based curriculum, removing the cumbersome 3D preprocessing prior methods require while matching models trained on curated data. Code and pre-trained models are publicly available.

Chinese Translation

通过跨视图补全进行自监督预训练，从图像对的共同可见区域学习强大的3D视觉特征。然而，参考视图对重建非共同可见区域的信息提供有限，隐含地在这些区域产生了单目训练信号。我们引入了Gekko，将这一限制转化为有用的信号。跨视图重建误差相对于掩蔽自编码器误差的相对改善是共同可见性的自监督代理：大的改善表明共同可见区域，而微不足道的改善则表示非共同可见区域。Gekko是一个从零开始训练的网络，联合执行跨视图补全、掩蔽自编码和该相对改善的逐像素预测，为所有掩蔽区域提供额外的双目信号，而无需任何真实的3D标注。在相同的架构和训练数据下，Gekko在零-shot对应估计、相对姿态估计和点图回归上始终优于CroCo，在最严格的相对姿态阈值下，其准确率高出6倍，并在ETH3D上实现了22%的端点误差下降。它所学习的额外通道本身就是一个强大的共同可见性检测器，能够在未见场景中表现出色，而Gekko的冻结特征在可比或更大规模的发布跨视图骨干网络中表现优越。它还可以通过简单的基于步幅的课程直接从原始视频中进行训练，消除了先前方法所需的繁琐3D预处理，同时与在精心挑选的数据上训练的模型相匹配。代码和预训练模型已公开提供。

View on arXiv Download PDF AI Translation

cs.CV / 110 / 2609.01551

What, Where, and How: Probing Spatiotemporal Representations in Video Foundation Models

什么、在哪里以及如何：探究视频基础模型中的时空表示

Musa, Sharon S., Forghani, Fereshteh, Thasarathan, Harrish, Joseph, Sonia, Kowal, Matthew, Derpanis, Konstantinos G.

Abstract

Self-supervised video foundation models learn rich spatiotemporal representations, yet it remains unclear what visual concepts these representations encode, where they emerge across transformer layers, and how they are geometrically organized. In this work, we tackle these three questions through a systematic layer-wise analysis of V-JEPA 2 and VideoMAE-v2. We leverage lightweight probes trained to discover three temporally grounded properties: (i) camera motion understanding, (ii) intuitive physics, and (iii) anomaly detection. Both models encode camera motion, with best results ($>90$ ROC AUC) emerging at 60-70% of network depth, and achieve moderate anomaly detection performance ($>60$ ROC AUC), but remain near chance on intuitive-physics tasks, suggesting a limited encoding of deeper physical reasoning. Beyond classification, we find that temporal features from individual videos form smooth low-dimensional trajectories in representation space, suggesting that camera motion is not only linearly decodable but also geometrically organized. Based on these results, we apply geometry-aware spline-based steering in the model's latent representations to interpolate camera motion, yielding steered videos with smoother trajectories and more coherent temporal progression than linear interpolation.

Chinese Translation

自监督视频基础模型学习丰富的时空表示，但这些表示编码了哪些视觉概念、它们在变换器层中何处出现以及它们是如何在几何上组织的仍然不清楚。在本研究中，我们通过对 V-JEPA 2 和 VideoMAE-v2 的系统层级分析来解决这三个问题。我们利用轻量级探针，训练以发现三个时间基础的属性：(i) 摄像机运动理解，(ii) 直观物理，(iii) 异常检测。两个模型均编码了摄像机运动，最佳结果（超过 90 的 ROC AUC）出现在网络深度的 60-70% 处，并在异常检测任务中取得中等表现（超过 60 的 ROC AUC），但在直观物理任务上接近随机水平，表明对更深层物理推理的编码有限。超越分类，我们发现来自单个视频的时间特征在表示空间中形成平滑的低维轨迹，这表明摄像机运动不仅是线性可解的，而且在几何上也是有组织的。基于这些结果，我们在模型的潜在表示中应用几何感知的样条基控制，以插值摄像机运动，从而生成具有更平滑轨迹和更连贯时间进程的引导视频，优于线性插值。

View on arXiv Download PDF AI Translation

cs.CV / 111 / 2609.01554

BS: Take the Hint - Interactive Multitracer PET/CT Lesion Segmentation with a Scribble-Conditioned ResEnc U-Net

BS: 提示一下 - 基于涂鸦条件的残差编码器 U-Net 的交互式多示踪剂 PET/CT 病灶分割

Sherif, Marven, Elmasry, Amgad, Ghazal, Youssef, Elghotni, Ayman

Abstract

Automated lesion segmentation in whole-body PET/CT is complicated by the variety of physiological tracer uptake patterns and by the differing appearance of lesions across tracers. The autoPET/CT V challenge addresses this by making segmentation interactive: user scribbles marking foreground and background are supplied alongside the image, and the algorithm is expected to exploit them. We present our submission, a scribble-conditioned residual encoder U-Net operating on four input channels: CT, PET, and a sparse scribble map for each of foreground and background. The network is initialised from the autoPET-III winning weights and extended from two to four input channels, with the two scribble channels zero-initialised so that the pretrained representation is preserved exactly at initialisation. Every model is fine-tuned per fold from the corresponding autoPET-III fold checkpoint, so that no validation case is seen during pretraining. PET intensities are normalised against a per-scan aorta blood-pool reference derived from a CT segmentation, which removes tracer- and centre-specific scaling without requiring lesion labels. At inference the five fold models are ensembled by averaging their softmax outputs per sliding-window patch, before Gaussian-weighted stitching. On the challenge's five-fold split, with each fold evaluated on its own validation cases, mean Dice is 0.554 and mean lesion-level F1 is 0.528 without scribbles, rising to 0.751 and 0.733 after five correction rounds. About 85% of that gain follows the first scribble, and the spread between fold models narrows five-fold over the same rounds, so interaction largely compensates for how well or badly a given model segments unaided.

Chinese Translation

全身 PET/CT 中的自动病灶分割因生理示踪剂摄取模式的多样性以及不同示踪剂下病灶外观的差异而变得复杂。autoPET/CT V 挑战通过使分割过程变得交互式来解决这一问题：用户在图像上标记前景和背景的涂鸦被提供给算法，算法应利用这些信息。我们展示了我们的提交结果，一个基于涂鸦条件的残差编码器 U-Net，处理四个输入通道：CT、PET，以及每个前景和背景的稀疏涂鸦图。该网络从 autoPET-III 的获胜权重初始化，并将输入通道从两个扩展到四个，两个涂鸦通道初始化为零，以确保在初始化时预训练表示的完全保留。每个模型在对应的 autoPET-III 折叠检查点上进行微调，因此在预训练期间没有看到验证案例。PET 强度根据从 CT 分割得出的每次扫描的主动脉血池参考进行归一化，这消除了示踪剂和中心特异性的缩放，而无需病灶标签。在推理时，五个折叠模型通过对每个滑动窗口补丁的 softmax 输出进行平均来集成，然后进行高斯加权拼接。在挑战的五折划分中，每个折叠在其自己的验证案例上进行评估，平均 Dice 为 0.554，平均病灶级 F1 为 0.528（未使用涂鸦），经过五轮修正后分别上升至 0.751 和 0.733。约 85% 的增益来自第一次涂鸦，并且在相同轮次中折叠模型之间的差异缩小了五倍，因此交互在很大程度上弥补了给定模型在无辅助情况下分割的好坏。

View on arXiv Download PDF AI Translation

cs.CV / 112 / 2609.01560

H3-World: Turning Language Understanding into World Control

H3-World：将语言理解转化为世界控制

Chen, Danze, Wang, Zeqing, Lin, Ziyue, Yang, Xingyi, Jin, Yeying

Abstract

We present H3-World, an efficient framework that turns the 33B MiniMax-H3 video generator into an interactive world model. Our key finding is that, as large video generators become more capable, language is emerging as a natural interface for control. MiniMax-H3, for example, already supports zero-shot control of character behavior and camera motion through natural-language instructions. Building on this, H3-World turns this coarse language interface into precise, temporally grounded world control, without introducing dedicated action modules. Specifically, we represent each action as a structured combination of character and camera instructions, and align them with the corresponding temporal video latents. To make the control temporally precise, we further introduce temporal attention routing, which restricts each instruction to its intended time interval and reduces control leakage across actions. Importantly, H3-World directly reuses the semantic representations learned during large-scale video pretraining and requires only lightweight adaptation. With only 8,000 gameplay samples, 10,000 LoRA optimization steps, and 0.199% trainable parameters, H3-World achieves effective character and camera control while preserving strong generation quality. It also generalizes to unseen scenarios. These results show that the control capabilities emerging in large video generators can be efficiently transformed into interactive world control.

Chinese Translation

我们提出了H3-World，这是一个高效的框架，将33B MiniMax-H3视频生成器转变为一个互动世界模型。我们的主要发现是，随着大型视频生成器能力的提升，语言逐渐成为控制的自然接口。例如，MiniMax-H3已经支持通过自然语言指令进行角色行为和相机运动的零-shot 控制。在此基础上，H3-World将这种粗略的语言接口转化为精确、时间上有依据的世界控制，而无需引入专门的动作模块。具体而言，我们将每个动作表示为角色和相机指令的结构化组合，并将其与相应的时间视频潜变量对齐。为了使控制在时间上更加精确，我们进一步引入了时间注意力路由，这限制了每个指令在其预定时间区间内的使用，并减少了动作之间的控制泄漏。重要的是，H3-World直接重用在大规模视频预训练过程中学习到的语义表示，仅需轻量级的适应。仅使用8,000个游戏样本、10,000次LoRA优化步骤和0.199%的可训练参数，H3-World实现了有效的角色和相机控制，同时保持了强大的生成质量。它还能够推广到未见过的场景。这些结果表明，大型视频生成器中出现的控制能力可以有效地转化为互动世界控制。

View on arXiv Download PDF AI Translation

cs.CV / 113 / 2609.01582

SpatialGuard: Harness-Guided Verifiable Spatial Reasoning for Text-to-Image Generation

SpatialGuard：基于引导的可验证空间推理用于文本到图像生成

Qian, Ziyun, Chen, Zizhi, Liu, Yizhou, Sun, Mingyang, Yang, Dingkang, Zhang, Lihua

Abstract

Complex 3D spatial text to image generation requires models to convert natural language into stable visual geometry, not merely semantic appearance. Existing prompt-driven or layout-conditioned methods improve controllability, but often lack an optimizable and verifiable spatial intermediary before visual sampling. As a result, object relations, occlusion, visibility, and camera constraints can decay during multi-round generation. This paper presents SpatialGuard, a structured layout-guided framework for complex 3D spatial text-to-image generation. SpatialGuard parses prompts into image synthesis-oriented 3D layouts through a Spatial Layout Architect, realizes them as visual conditions and candidate images through a Visual Realizer, and uses a Visual Alignment Critic to validate consistency among prompt, layout, and image. To keep constraints stable across iterations, SpatialGuard introduces a Layout Harness that organizes rule constraints, tool invocation, shared knowledge, and feedback loops around the editable layout state. This design turns complex spatial generation from implicit prompt following into a verifiable process of planning, realization, validation, and repair. Comprehensive experiments show that SpatialGuard achieves state-of-the-art performance in complex 3D spatial layout generation and improves spatial faithfulness over existing text-to-image and layout control baselines.

Chinese Translation

复杂的3D空间文本到图像生成要求模型将自然语言转换为稳定的视觉几何体，而不仅仅是语义外观。现有的基于提示驱动或布局条件的方法提高了可控性，但在视觉采样之前往往缺乏可优化和可验证的空间中介。因此，在多轮生成过程中，物体关系、遮挡、可见性和相机约束可能会下降。本文提出了SpatialGuard，一个用于复杂3D空间文本到图像生成的结构化布局引导框架。SpatialGuard通过空间布局架构师将提示解析为面向图像合成的3D布局，通过视觉实现器将其实现为视觉条件和候选图像，并使用视觉对齐评估器验证提示、布局和图像之间的一致性。为了在迭代过程中保持约束的稳定性，SpatialGuard引入了布局支架，围绕可编辑的布局状态组织规则约束、工具调用、共享知识和反馈循环。该设计将复杂的空间生成从隐式提示跟随转变为一个可验证的规划、实现、验证和修复过程。全面的实验表明，SpatialGuard在复杂3D空间布局生成方面达到了最先进的性能，并在空间真实性上优于现有的文本到图像和布局控制基线。

View on arXiv Download PDF AI Translation

cs.CV / 114 / 2609.01584

A Benchmark for Vehicle Attribute Classification in Cross-Domain Surveillance Scenarios

跨域监控场景中车辆属性分类的基准测试

Silva Jr., Sergio M., Remer, Otavio T., Lima, Gabriel E., Wojcik, Lucas, Laroca, Rayson, Menotti, David

Abstract

Vehicle attribute analysis is a key component of Intelligent Transportation Systems (ITS), supporting applications such as vehicle identification, traffic monitoring, and forensic investigation. However, models trained under controlled conditions often degrade in real surveillance scenarios due to changes in viewpoint, occlusion, illumination, and sensor characteristics. This paper introduces Unconstrained Vehicle Identification Benchmark (UVIB), a benchmark for evaluating three operational vehicle-analysis tasks: front/rear orientation, occlusion-related suitability for Vehicle Make and Model Recognition (VMMR), and color clarity. The benchmark contains 84,835 vehicle images from seven public Brazilian datasets, grouped into surveillance and general acquisition domains, with unified binary annotations that were not jointly available in the original sources. Four representative architectures, EfficientNetV2-S, ResNet-50, ViT/B-16, and YOLO11s-cls, are evaluated under mixed-domain, cross-domain, and cross-dataset protocols. The results show that domain shift has a stronger impact than architecture choice, with substantial degradation in cross-domain settings, especially for VMMR suitability and color clarity. While orientation generalizes more reliably, VMMR suitability remains affected by class imbalance and ambiguous occlusions, and color clarity is highly sensitive to illumination and sensor modality. These findings highlight the need for benchmarks and evaluation protocols that explicitly measure operational robustness beyond standard in-domain accuracy. The proposed benchmark is publicly available at https://github.com/UFPR-IPASP-PR/uvib-vehicle-attributes/.

Chinese Translation

车辆属性分析是智能交通系统（ITS）的关键组成部分，支持车辆识别、交通监控和法医调查等应用。然而，在受控条件下训练的模型在实际监控场景中往往会因视角变化、遮挡、光照和传感器特性等因素而性能下降。本文引入了无约束车辆识别基准（Unconstrained Vehicle Identification Benchmark，UVIB），该基准用于评估三项操作性车辆分析任务：前/后方位、与遮挡相关的车辆品牌和型号识别（Vehicle Make and Model Recognition，VMMR）适用性，以及颜色清晰度。该基准包含来自七个公共巴西数据集的84,835张车辆图像，分为监控和一般获取领域，并提供了在原始来源中未共同提供的统一二元注释。我们在混合域、跨域和跨数据集协议下评估了四种代表性架构：EfficientNetV2-S、ResNet-50、ViT/B-16和YOLO11s-cls。结果表明，领域转移对模型的影响大于架构选择，在跨域设置中性能显著下降，尤其是在VMMR适用性和颜色清晰度方面。虽然方位的泛化能力更可靠，但VMMR适用性仍受到类别不平衡和模糊遮挡的影响，而颜色清晰度对光照和传感器模态高度敏感。这些发现突显了需要建立基准和评估协议，以明确测量操作的鲁棒性，而不仅仅是标准的领域内准确性。所提出的基准可在 https://github.com/UFPR-IPASP-PR/uvib-vehicle-attributes/ 上公开获取。

View on arXiv Download PDF AI Translation

cs.CV / 115 / 2609.01598

UI-VISA: U-Net Initialized Vascular Image Segmentation Architecture

UI-VISA：U-Net 初始化的血管图像分割架构

Kaur, Asees, Sindi, Suzanne S., Rutter, Erica M.

Abstract

Accurate segmentation of vascular structures in digital subtraction angiography (DSA) images remains challenging due to the thin, elongated, and branching nature of blood vessels. Pixel-wise deep learning approaches such as U-Net achieve strong general-purpose segmentation performance but often produce fragmented or discontinuous predictions in fine vascular regions, since they do not explicitly enforce structural connectivity. Region growing algorithms preserve spatial context and topological continuity, but are highly sensitive to seed point initialization and can be computationally expensive. We propose UI-VISA (U-Net Initialized Vascular Image Segmentation Architecture), a hybrid pipeline that combines the complementary strengths of both approaches. UI-VISA uses U-Net's foreground predictions as informed seed points for a CNN-guided region growing algorithm, which then iteratively refines the segmentation by enforcing local connectivity and recovering fine vessel details that U-Net alone tends to miss or over-predict. We evaluate UI-VISA against standalone U-Net and a prior region-growing-based method (VISA) using 5-fold cross-validation on 26 DSA images. UI-VISA achieves the highest mean Dice and clDice scores across folds, and a paired Wilcoxon signed-rank test shows the improvement in clDice is statistically significant ($p=0.023$), consistent with the method's design goal of preserving vascular connectivity, while the improvement in Dice does not reach significance ($p=0.104$).

Chinese Translation

在数字减影血管造影（DSA）图像中，准确分割血管结构仍然具有挑战性，因为血管通常细长且分支。像 U-Net 这样的逐像素深度学习方法在一般分割性能上表现出色，但在细小血管区域往往会产生碎片化或不连续的预测，因为它们没有明确地强制结构连通性。区域生长算法能够保持空间上下文和拓扑连续性，但对种子点初始化高度敏感，并且计算开销较大。我们提出了 UI-VISA（U-Net 初始化的血管图像分割架构），这是一种混合管道，结合了两种方法的互补优势。UI-VISA 使用 U-Net 的前景预测作为信息丰富的种子点，供 CNN 引导的区域生长算法使用，该算法通过强制局部连通性并恢复 U-Net 单独处理时可能遗漏或过度预测的细小血管细节，迭代地优化分割。我们在 26 幅 DSA 图像上使用 5 折交叉验证评估 UI-VISA 与独立的 U-Net 及先前的基于区域生长的方法（VISA）。UI-VISA 在各折中获得了最高的平均 Dice 和 clDice 分数，配对的 Wilcoxon 符号秩检验显示 clDice 的改善具有统计显著性（$p=0.023$），这与该方法保持血管连通性的设计目标一致，而 Dice 的改善未达到显著性（$p=0.104$）。

View on arXiv Download PDF AI Translation

cs.CV / 116 / 2609.01607

Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation, Task to System

揭示本土统一多模态模型中的理解-生成协同：从表征、任务到系统

Wu, Penghao, Diao, Haiwen, Fan, Weichen, Lu, Lewei, Lin, Dahua, Liu, Ziwei

Abstract

While unified multimodal models (UMMs) jointly perform visual understanding and generation within a single model, functional unification does not guarantee learning synergy: the two objectives may reinforce each other, compete for capacity, or merely coexist. We investigate their relationship at the representation, task, and system levels in a controlled, structurally native setting without pretrained vision priors. At the representation level, we find that each objective provides useful signal to the other: generation enriches the visual features learned for understanding, while understanding strengthens vision--language alignment for generation. However, when both objectives are forced through the same computation path, one tends to dominate. A task-decoupled architecture that specializes conflicting visual computation while preserving semantic interaction avoids this asymmetric degradation. At the task level, through three case studies, we find positive bidirectional transfer when understanding and generation tasks rely on shared knowledge. At the system level, we show that an end-to-end UMM outperforms a matched planner--executor pipeline on complex tasks that explicitly require both image understanding and generation. Together, these results show that the value of UMMs extends beyond a unified interface: appropriate specialization, shared task knowledge, and end-to-end optimization can turn coexistence into synergy.

Chinese Translation

虽然统一多模态模型（UMMs）在单一模型中共同执行视觉理解和生成，但功能统一并不保证学习协同：这两个目标可能相互强化、争夺能力，或仅仅共存。我们在一个受控的、结构上本土化的环境中，研究它们在表征、任务和系统层面的关系，而不使用预训练的视觉先验。在表征层面，我们发现每个目标都为另一个目标提供了有用的信号：生成丰富了用于理解的视觉特征，而理解则增强了生成的视觉-语言对齐。然而，当两个目标被强制通过相同的计算路径时，一个目标往往会占主导地位。一个任务解耦的架构能够专门处理冲突的视觉计算，同时保持语义交互，从而避免这种不对称的退化。在任务层面，通过三个案例研究，我们发现当理解和生成任务依赖于共享知识时，存在积极的双向迁移。在系统层面，我们展示了一个端到端的UMM在复杂任务中优于匹配的规划-执行管道，这些任务明确要求同时进行图像理解和生成。综合来看，这些结果表明，UMMs的价值超越了统一接口：适当的专业化、共享任务知识和端到端优化可以将共存转变为协同。

View on arXiv Download PDF AI Translation

人工智能 (Artificial Intelligence)

124

cs.AI / 1 / 2609.00002

HyperWorld: Hypergraph-Structured State Serialization Improves Learned Textual World Models

超世界：超图结构状态序列化提升学习的文本世界模型

Zhang, Yun-Jian, Liang, Chen-Wei, Zhang, Tian-Yi, Ding, Jian, Wu, Yi-Lun, Li, Ao-Bo, Su, Wei-Cong, Saifullah, An, Hong-Yu, Wang, Mu-Jiang-Shan

Abstract

World models enable language-model agents to predict environment dynamics and plan before acting. In text environments, the model must learn symbolic action effects from serialized state descriptions, but the role of serialization structure remains underexplored. We present HyperWorld, a controlled study of state serialization for learned textual world models. We compare raw observations with three symbolic serializations of the same ground-truth state: independent sentences, pairwise triples, and entity-centered hyperedge units that group multiple related facts around entities and relations. All variants use the same training objective: given a state and an action, predict symbolic effects or judge the action infeasible. Across model scales, data budgets, and in-distribution and out-of-distribution test worlds, hyperedge serialization gives the clearest gains for 0.5B--1.5B models and under distribution shift. Larger models reduce the gap, and pairwise triples can match or slightly exceed hyperedges on in-distribution exact match, but hyperedges achieve the strongest out-of-distribution fact F1 and the best small-to-medium scale trade-off between feasibility detection and effect prediction. In downstream greedy planning, the hyperedge world model also attains the highest success rate among the tested representations. These results show that higher-order state organization is a simple but effective inductive bias for learned symbolic world models, especially when model capacity is limited or test environments differ from training.

Chinese Translation

世界模型使语言模型代理能够预测环境动态并在行动前进行规划。在文本环境中，模型必须从序列化的状态描述中学习符号动作效果，但序列化结构的作用仍然未被充分探索。我们提出了超世界（HyperWorld），这是对学习的文本世界模型状态序列化的一个控制研究。我们将原始观察与三种符号序列化的相同真实状态进行比较：独立句子、成对三元组和围绕实体及关系分组多个相关事实的以实体为中心的超边单元。所有变体使用相同的训练目标：给定一个状态和一个动作，预测符号效果或判断该动作不可行。在模型规模、数据预算以及分布内和分布外测试世界中，超边序列化在0.5B至1.5B模型和分布转移下提供了最明显的提升。更大的模型缩小了差距，而成对三元组在分布内的精确匹配上可以与超边相匹配或略微超过，但超边在分布外事实的F1分数上表现最强，并在可行性检测与效果预测之间实现了最佳的小到中等规模的权衡。在下游贪婪规划中，超边世界模型在测试的表示中也达到了最高的成功率。这些结果表明，高阶状态组织是学习符号世界模型的一种简单但有效的归纳偏置，特别是在模型容量有限或测试环境与训练环境不同的情况下。

View on arXiv Download PDF AI Translation

cs.AI / 2 / 2609.00003

I-CARE: Analysis of interference-related phenomena in a controllable, diverse and representative unlearning setting for text-to-image models

I-CARE：在可控、多样且具有代表性的文本到图像模型的去学习环境中分析干扰相关现象

Pereira, Leonardo Santiago Benitez, Viñolo, Marcos Escudero, Arribas, Luis Herranz

Abstract

Machine unlearning studies the removal of knowledge from an AI model, making the system forget a concept it previously learned. Despite rapid progress in generative machine unlearning, the unintended degradation of semantically related concepts that should have been retained (henceforth, interference) remains poorly characterized and inconsistently evaluated. This paper introduces I-CARE, a methodology that formalizes interference as a first-class object of study in generative unlearning. Rather than proposing a new benchmark or unlearning algorithm, I-CARE provides formal definitions for tasks, metrics, and templates for reporting results, enabling the systematic and reproducible study of interference across unlearning settings. While our methodology is designed to remain valid as models and unlearning algorithms evolve, decoupling long-term scientific insight from transient empirical results, we present a feasibility demonstration with state-of-the-art algorithms and frequently used datasets. The results demonstrate that I-CARE enables meaningful analysis of interference patterns across multiple unlearning settings, establishing the practical applicability of the framework. The software implementation of the methodology is provided in an open-source framework, together with a web-based graphical interface that enables exploration of the outcomes of this study without requiring direct interaction with the codebase or specialized data analysis tools.

Chinese Translation

机器去学习研究的是从人工智能模型中移除知识，使系统忘记其先前学习的概念。尽管生成性机器去学习取得了快速进展，但应当保留的语义相关概念意外退化（以下简称为干扰）仍然缺乏明确的特征描述和一致的评估。本文介绍了I-CARE，一种将干扰形式化为生成性去学习研究的首要对象的方法论。I-CARE并未提出新的基准或去学习算法，而是为任务、指标和结果报告模板提供了正式定义，从而实现对不同去学习环境中干扰的系统性和可重复研究。尽管我们的方法论旨在随着模型和去学习算法的发展而保持有效，将长期科学洞察与短期经验结果解耦，但我们展示了与最先进算法和常用数据集的可行性演示。结果表明，I-CARE能够在多个去学习环境中进行有意义的干扰模式分析，确立了该框架的实际适用性。该方法论的软件实现以开源框架的形式提供，并配有基于网络的图形界面，使得用户无需直接与代码库或专业数据分析工具交互即可探索本研究的结果。

View on arXiv Download PDF AI Translation

cs.AI / 3 / 2609.00004

Discrete-Time MDP Modeling for Multi-Item Capacitated Lot Sizing with Stochastic Demand Timing

具有随机需求时机的多项目容量批量生产的离散时间马尔可夫决策过程建模

Bayati, Léa, Dahmoune, Mohamed, Rodoplu, Melek

Abstract

This paper studies a finite-horizon multi-item capacitated lot-sizing problem in which demand quantities are deterministic, while demand-arrival periods are stochastic. Each demand occurs once within a known time window and must be satisfied no later than its deadline. The proposed model makes production and allocation decisions at the demand level, allowing it to represent capacity competition, demand-specific backlog, and allocation-dependent inventory dynamics. The stochastic problem is formulated as a discrete-time Markov decision process (DTMDP), including the state space, feasible actions, transition kernel, and one-period cost function. To isolate the computational effect of stochastic timing, each stochastic instance is first compared with a deterministic counterpart in which each arrival distribution is replaced by its most likely arrival period. This comparison shows that stochastic timing substantially increases the number of states, the number of transitions, solution time, and memory pressure. A genetic algorithm (GA) is then proposed for the stochastic-timing problem. The GA searches over feasible state-feedback policies and evaluates each policy exactly under the DTMDP transition model. Computational experiments on 330 benchmark instances show that the GA remains close to the exact stochastic solution whenever the latter is available, with an average optimality gap of about $3.44\%$. On the difficult benchmark instances, comprising 90 test cases, the GA remains below the $5\%$ optimality-gap threshold and achieves an average optimization speedup of $6.89 \pm 1.41$ at the $95\%$ confidence level. For instances that cannot be solved exactly on the available hardware, an empirical Bellman-time regression is used to estimate the missing exact resolution time and extrapolate the expected GA speedup.

Chinese Translation

本文研究了一种有限时间范围内的多项目容量批量生产问题，其中需求量是确定的，而需求到达时间是随机的。每个需求在已知的时间窗口内发生一次，必须在截止日期之前得到满足。所提出的模型在需求层面上做出生产和分配决策，能够表示产能竞争、需求特定的积压和依赖于分配的库存动态。随机问题被构建为离散时间马尔可夫决策过程（DTMDP），包括状态空间、可行动作、转移核和单周期成本函数。为了隔离随机时机的计算影响，每个随机实例首先与其确定性对应物进行比较，其中每个到达分布被替换为其最可能的到达时间。比较结果表明，随机时机显著增加了状态数量、转移数量、解决时间和内存压力。随后，针对随机时机问题提出了一种遗传算法（GA）。该算法在可行的状态反馈策略上进行搜索，并在DTMDP转移模型下精确评估每个策略。对330个基准实例的计算实验表明，当存在精确的随机解决方案时，GA的结果与之接近，平均最优性差距约为$3.44\%$。在由90个测试案例组成的困难基准实例中，GA的最优性差距保持在$5\\%$以下，并在$95\\%$置信水平下实现了$6.89 \\pm 1.41$的平均优化加速。对于在可用硬件上无法精确解决的实例，采用经验贝尔曼时间回归来估计缺失的精确解决时间，并推断预期的GA加速。

View on arXiv Download PDF AI Translation

cs.AI / 4 / 2609.00005

Incremental Risk Assessment of Progressive Elder Financial Scams via Instruction-Tuned Small Language Models

通过指令调优的小型语言模型对渐进式老年人金融诈骗的增量风险评估

Ghafariasl, Parviz, Fu, Weimin, Guo, Xiaolong, Chang, Shing I.

Abstract

Financial scams targeting older adults increasingly occur through text and voice channels such as email, SMS, and phone calls, unfolding over multiple conversational turns that begin with impersonation or casual contact, escalate through trust building and urgency, and culminate in requests for sensitive information or financial transfers. Because risk signals emerge incrementally across turns, effective detection requires models that continuously update risk estimates under resource-constrained deployment settings. We propose a cumulative turn-based risk assessment framework that incrementally aggregates conversational turns and re-estimates risk at each step, enabling dynamic scam monitoring across progressively evolving conversations. A multi-turn dialogue dataset is constructed to cover investment, charity, and tech support scam scenarios, with each dialogue containing two to eight turns and annotated at every cumulative stage with a qualitative risk level, a continuous risk score, an explanatory rationale, and a safety recommendation. Four small language models (Phi-4, LLaMA-3.2, DeepSeek-R1, and Qwen3) are fine-tuned and evaluated under a unified training framework. Fine-tuned small models capture fraud-related linguistic cues and cross-turn escalation patterns while maintaining compact architectures suitable for mobile and resource-constrained deployment settings. Among the evaluated models, Phi-4 and LLaMA-3.2 achieve stronger turn-aware risk estimation performance relative to their parameter scale. These results suggest that structured cumulative modeling can support incremental scam risk assessment in deployment-oriented settings while highlighting the potential of compact language models for privacy-aware and on-device fraud protection.

Chinese Translation

针对老年人的金融诈骗越来越多地通过文本和语音渠道（如电子邮件、短信和电话）发生，这些诈骗通常在多个对话轮次中展开，开始时以冒充或随意接触为开端，通过建立信任和紧迫感逐步升级，最终以请求敏感信息或财务转移为高潮。由于风险信号在对话轮次中逐步显现，有效的检测需要能够在资源受限的部署环境下持续更新风险估计的模型。我们提出了一种基于累积轮次的风险评估框架，该框架逐步聚合对话轮次并在每一步重新估计风险，从而实现对逐步演变的对话进行动态诈骗监测。构建了一个多轮对话数据集，涵盖投资、慈善和技术支持诈骗场景，每个对话包含两到八个轮次，并在每个累积阶段进行了标注，包含定性风险等级、连续风险评分、解释性理由和安全建议。四个小型语言模型（Phi-4、LLaMA-3.2、DeepSeek-R1 和 Qwen3）在统一的训练框架下进行了微调和评估。微调后的小型模型能够捕捉与欺诈相关的语言线索和跨轮次升级模式，同时保持适合移动和资源受限部署环境的紧凑架构。在评估的模型中，Phi-4 和 LLaMA-3.2 相较于其参数规模实现了更强的轮次感知风险估计性能。这些结果表明，结构化的累积建模可以支持在部署导向环境中的增量诈骗风险评估，同时突显了紧凑型语言模型在隐私保护和设备端欺诈防护中的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 5 / 2609.00012

Long-Horizon State Tracking in LLMs: Executing MD5 through a Deep Sequence of Dependent Tool Calls

长时间状态跟踪在大型语言模型中的应用：通过一系列依赖工具调用逐步执行 MD5

Pai, Dheeraj Mohandas, Xian, Lu

Abstract

Long-horizon tasks remain uncommon in large language model (LLM) evaluation, and for a reason: when each step depends on the last, per-step accuracy that looks excellent in isolation decays catastrophically, as errors cascade and the end-to-end failure probability grows sharply with length. Existing agentic benchmarks report end-to-end success but confound this state-tracking difficulty with instruction interpretation, give no control group that isolates it, and are vulnerable to shortcuts such as a hallucinated final answer, so they cannot say why a long run fails. Whether an LLM can carry exact intermediate state across many tool calls at all is itself not well established. We test this cleanly by having the model compute a cryptographic hash, MD5, step by step: a sequence of $196$ dependent tool calls over $64$ rounds while it carries four $32$-bit words $(a,b,c,d)$ in its own context from one call to the next. Interpretation is trivial and, because we implement MD5 from scratch (RFC~1321), we align every call to the ground-truth trace and check the digest to the bit, so any failure is pure bookkeeping. gpt-oss-120b, a mixture-of-experts model with only $\sim$5.5B active parameters per token, at temperature $0$ with a short fixed prompt, carries the full state across all $196$ calls and returns the correct digest on a majority of completed runs. In the strongest setting we replace every primitive tool with a second LLM, so a driver and a worker compute the whole hash from scratch with no exact-arithmetic oracle in the loop. Two ingredients decide success and neither changes the weights: keeping the model's own reasoning in its context each turn, and voting over a thinking-enabled worker to remove its modular-arithmetic slips. We localize the residual failures by origin, separating state-carrying from arithmetic and from serving.

Chinese Translation

长时间任务在大型语言模型（LLM）评估中仍然不常见，这有其原因：当每一步都依赖于前一步时，单步准确性在孤立情况下看似优秀，但会急剧下降，因为错误会级联，端到端的失败概率随着任务长度的增加而急剧上升。现有的代理基准报告了端到端的成功，但将这一状态跟踪的困难与指令解释混淆，未提供隔离的对照组，并且容易受到诸如虚构最终答案等捷径的影响，因此无法说明长时间运行失败的原因。一个 LLM 是否能够在多个工具调用中准确地保持中间状态本身尚未得到充分验证。我们通过让模型逐步计算一个加密哈希 MD5 来干净地测试这一点：在 $64$ 轮中进行 $196$ 次依赖工具调用，同时在每次调用之间在其上下文中保持四个 $32$ 位字 $(a,b,c,d)$。解释是微不足道的，因为我们从头实现了 MD5（RFC~1321），我们将每次调用与真实轨迹对齐，并逐位检查摘要，因此任何失败都是纯粹的账务处理。gpt-oss-120b 是一个混合专家模型，每个令牌仅有约 $5.5$ 亿个活跃参数，在温度 $0$ 和短固定提示下，能够在所有 $196$ 次调用中保持完整状态，并在大多数完成的运行中返回正确的摘要。在最强的设置中，我们用第二个 LLM 替换每个基本工具，因此驱动程序和工作者从头计算整个哈希，循环中没有精确算术的神谕。成功的两个要素并不改变权重：在每个回合中保持模型自身的推理在其上下文中，以及对一个具备思考能力的工作者进行投票，以消除其模组算术的失误。我们通过来源定位剩余的失败，将状态承载与算术和服务分开。

View on arXiv Download PDF AI Translation

cs.AI / 6 / 2609.00015

OpenAgentFlow: Enabling System-Wide Safety Boundaries for Heterogeneous AI Agent Fleets

OpenAgentFlow：为异构AI代理群体启用系统级安全边界

Chen, Dongsheng, Zhao, Xiangyu, Yao, Xin, Wei, Xuetao

Abstract

AI agents powered by large language models are evolving from isolated assistants into heterogeneous systems in which multiple agents, planners, controllers, and execution backends operate over the same user or enterprise environment. In such settings, safety becomes a system-level action-governance problem: deciding whether concrete agent-generated actions should be committed before they modify shared state. Existing safeguards cover prompts, tool calls, GUI actions, and agent-local behavior, but often leave enforcement fragmented, obscure risks that emerge across multi-step action flows, and provide limited support for auditability and policy evolution. We present OpenAgentFlow, a control-plane/action-plane architecture that enforces safety at the action-commit boundary. It normalizes pending GUI actions, API calls, tool calls, and LLM-generated invocations into a unified AgentEvent stream, routes each event through a shared pre-execution Policy Enforcement Point, and maintains provenance, session state, audit records, and updatable policies in the control plane. This creates a shared governable action stream and allows new rules to take effect without modifying agents, prompts, models, or execution paths. We instantiate OpenAgentFlow on Android. On a 300-case action-event benchmark, it achieves 94.0% accuracy and a 95.3% attack block rate. On a 30-case dynamic-policy suite, it matches expected behavior in 27 cases after new rules are installed. Across 98 traced cases from a 100-case Android emulator suite, it achieves 90.8% raw accuracy and a 92.9% trace-adjusted pass rate across GUI, API, and LLM-planned cases. These results show that OpenAgentFlow provides a practical shared enforcement boundary for heterogeneous AI agent fleets.

Chinese Translation

由大型语言模型驱动的AI代理正从孤立的助手演变为异构系统，其中多个代理、规划器、控制器和执行后端在同一用户或企业环境中协同工作。在这种环境下，安全性成为一个系统级的行动治理问题：在修改共享状态之前，决定是否应提交具体的代理生成的行动。现有的安全措施涵盖了提示、工具调用、图形用户界面（GUI）操作和代理本地行为，但往往导致执行碎片化，模糊了在多步骤行动流程中出现的风险，并且对审计能力和政策演变的支持有限。我们提出了OpenAgentFlow，一种在行动提交边界上执行安全性的控制平面/行动平面架构。它将待处理的GUI操作、API调用、工具调用和LLM生成的调用规范化为统一的AgentEvent流，通过共享的预执行政策执行点路由每个事件，并在控制平面中维护来源、会话状态、审计记录和可更新政策。这创建了一个可治理的共享行动流，并允许新规则生效而无需修改代理、提示、模型或执行路径。我们在Android上实例化了OpenAgentFlow。在一个300案例的行动事件基准测试中，它达到了94.0%的准确率和95.3%的攻击阻止率。在一个30案例的动态政策套件中，在安装新规则后，它在27个案例中匹配了预期行为。在从100案例的Android模拟器套件中追踪的98个案例中，它在GUI、API和LLM规划的案例中达到了90.8%的原始准确率和92.9%的追踪调整通过率。这些结果表明，OpenAgentFlow为异构AI代理群体提供了一个实用的共享执行边界。

View on arXiv Download PDF AI Translation

cs.AI / 7 / 2609.00018

SCAFFOLD: A Large-Scale Structured Dataset of Computer Science Research Figures with Diagram QA and Chain-of-Thought Reasoning Traces

SCAFFOLD：一个大规模计算机科学研究图形的结构化数据集，包含图表问答和思维链推理轨迹

Raut, Ranjit, Subedi, Aarav, Rai, Sagun, Jha, Sudan

Abstract

Computer science papers rely heavily on diagrams: architecture drawings, system flowcharts, and pipeline schematics that often carry more information than the text around them. There is currently no public dataset that pairs this specific kind of figure with captions, context, questions, answers, and step-by-step reasoning, which is exactly what is needed to train a vision-language model to understand them. We present \textbf{SCAFFOLD}\footnote{https://github.com/theranjitraut/scaffold}, a large-scale structured dataset of computer science research figures with diagram QA and Chain-of-Thought reasoning traces. This dataset consists of (image, caption, context, question-answer, chain-of-thought) tuples from arXiv computer science papers prepared using layout detection and PDF parsing, with an AI-assisted question-generation step. The resulting large-sized SCAFFOLD-157K dataset spans 3,058 papers with 29,887 figures (157,387 pairs), a medium-sized SCAFFOLD-37K dataset (36,797 pairs), and a small-sized SCAFFOLD-12K dataset (12,000 pairs). We used SCAFFOLD-12K for baseline experiments on Qwen2.5-VL-3B-Instruct.

Chinese Translation

计算机科学论文在很大程度上依赖于图表：架构图、系统流程图和管道示意图，这些图表通常承载的信息超过周围的文本。目前没有公开的数据集将这种特定类型的图形与标题、上下文、问题、答案和逐步推理配对，而这正是训练视觉-语言模型以理解它们所需的。我们提出了 extbf{SCAFFOLD}ootnote{https://github.com/theranjitraut/scaffold}，一个大规模的计算机科学研究图形结构化数据集，包含图表问答和思维链推理轨迹。该数据集由来自arXiv计算机科学论文的（图像、标题、上下文、问答、思维链）元组组成，经过布局检测和PDF解析，并包含一个AI辅助的问题生成步骤。最终生成的大型SCAFFOLD-157K数据集涵盖了3,058篇论文和29,887个图形（157,387对），中型SCAFFOLD-37K数据集（36,797对），以及小型SCAFFOLD-12K数据集（12,000对）。我们使用SCAFFOLD-12K进行Qwen2.5-VL-3B-Instruct的基线实验。

View on arXiv Download PDF AI Translation

cs.AI / 8 / 2609.00028

UI-Venus-2 Technical Report

UI-Venus-2 技术报告

Venus Team, Cai, Zhuohan, Chen, Haoxing, Chen, Jiaxuan, Chen, Weizhi, Gao, Changlong, Gu, Zhangxuan, Guo, Yuan, Hu, Yusong, Jiang, Jianrong, Li, Jianguo, Li, Runze, Lin, Jinzhen, Ma, Zhenyu, Meng, Changhua, Peng, Han, Qiu, Xinyu, Shen, Shuheng, Shui, Zhongyi, Wang, Weiqiang, Wen, Ming, Xu, Zhuoer, Yan, Hang, Yang, Kaiwen, Yao, Ruilin, Yu, Nanjun, Zeng, Zhengwen, Zhang, Lianrui, Zhang, Yunzhu, Zhao, Zhe, Zhou, Beitong

Abstract

Multimodal GUI agents have emerged as a promising paradigm for digital task automation, yet transitioning from benchmark-oriented models to dependable real-world applications remains challenging due to limited environment coverage, brittle task construction, and unreliable reward verification. In this work, we present UI-Venus-2, a general-purpose foundation GUI agent designed to operate across mobile, web, and desktop environments through a unified closed-loop reasoning-action framework. To bridge the gap toward practical deployment, we jointly scale three critical dimensions: (1) Environments, expanding coverage to more than 170 multilingual mobile apps and native desktop operating systems; (2) Tasks, employing a deep-research pipeline for function-grounded instruction generation; and (3) Verification, adopting trace-level and sample-level evaluators with visual keypoints and multi-model voting to ensure reliable RL signals for training. Furthermore, we integrate safety-aware mechanisms to ensure controlled execution of consequential actions. By offering a capable, efficient, and open-source foundation, UI-Venus-2 advances the field toward more generalizable, verifiable, and self-reflective agents for real-world applications.

Chinese Translation

多模态图形用户界面（GUI）代理作为数字任务自动化的有前景的范式逐渐兴起，但由于环境覆盖有限、任务构建脆弱和奖励验证不可靠，从基准导向模型过渡到可靠的现实世界应用仍然具有挑战性。在本研究中，我们提出了 UI-Venus-2，这是一个通用基础 GUI 代理，旨在通过统一的闭环推理-行动框架在移动、网页和桌面环境中运行。为了缩小实际部署的差距，我们共同扩展了三个关键维度：（1）环境，覆盖超过 170 个多语言移动应用和本地桌面操作系统；（2）任务，采用深度研究管道生成基于功能的指令；（3）验证，采用基于轨迹和样本级的评估器，结合视觉关键点和多模型投票，以确保可靠的强化学习（RL）信号用于训练。此外，我们集成了安全意识机制，以确保对重要行动的受控执行。通过提供一个功能强大、高效且开源的基础，UI-Venus-2 推动该领域朝向更具可推广性、可验证性和自我反思的代理，以适应现实世界的应用。

View on arXiv Download PDF AI Translation

cs.AI / 9 / 2609.00032

EULER: Exploring Underused Links with Evidence-Checked Return for Multi-Agent Mathematical Discovery

EULER：通过证据验证的返回探索未充分利用的链接以实现多智能体数学发现

Zhenzhuo, Ren

Abstract

Mathematical communities work with different objects, invariants, and tools, so transferring a problem across them is expensive and often skipped. We present EULER, a multi-agent system that takes such a transfer--a bridge--as its unit of search. Around a fixed conjecture, EULER runs direct, adjacent-domain, and distant-domain routes in competition; a bridge keeps its budget only if it supplies an operation the source representation cannot execute and its target-side evidence returns to the original statement along a checked implication. Six ordered stress tests reject invalid bridges before expensive search begins. We evaluate EULER on 120 recent conjectures. The conjectures were frozen before search and screened for contamination, and are drawn from public papers by authors who had recently published in the Journal of Combinatorial Theory, Series A, a leading journal in combinatorics. EULER produced 10 proofs and 3 refutations, plus 45 scoped partial results. Two mechanisms held up under ablation: bridge-specific stress tests cut incorrect conclusions from 9 to 3, and bridge material combined with a target-native operation yielded a positive interaction of +4.2 resolved tasks that neither factor produced alone. Domain distance did not reliably predict success; executable operation gain and valid return did.

Chinese Translation

数学社区处理不同的对象、不变量和工具，因此在它们之间转移问题的成本很高，通常会被跳过。我们提出了EULER，一个多智能体系统，它将这种转移——桥梁——作为搜索的单位。在一个固定的猜想周围，EULER 竞争性地运行直接、相邻领域和远程领域的路径；只有当桥梁提供源表示无法执行的操作，并且其目标侧的证据沿着经过验证的蕴含返回到原始陈述时，它才能保持其预算。在昂贵的搜索开始之前，六个有序的压力测试会拒绝无效的桥梁。我们在120个最近的猜想上评估了EULER。这些猜想在搜索之前被冻结并筛查了污染，来源于最近在组合理论A系列期刊上发表的公共论文，该期刊是组合学领域的领先期刊。EULER 产生了10个证明和3个反驳，以及45个范围限定的部分结果。两种机制在消融测试中表现良好：特定于桥梁的压力测试将错误结论从9个减少到3个，而桥梁材料与目标本地操作结合产生了+4.2的积极交互，解决了既不由单一因素产生的任务。领域距离并不能可靠地预测成功；可执行操作的增益和有效返回则可以。

View on arXiv Download PDF AI Translation

cs.AI / 10 / 2609.00071

When Prediction Error Is Not Enough: Evaluating Nuisance-Function Prediction for Causal Estimation

当预测误差不足以满足需求时：评估因果估计中的干扰函数预测

Cao, Cong

Abstract

Prediction error is widely used to evaluate nuisance-function estimators in causal inference, but its relationship with causal estimator performance may differ across performance measures. We studied this question in a partially linear model using Monte Carlo simulations. We compared ordinary least squares (OLS), generalized additive models (GAMs), XGBoost, and Double Machine Learning with XGBoost (DML-XGBoost), evaluating nuisance-function prediction error, bias, RMSE, and 95\% confidence interval coverage. We also examined a simple joint-error measure based on the absolute cross-product of estimation errors from the exposure and outcome nuisance functions. Across the simulated settings, XGBoost had the lowest RMSE among the non-oracle methods, while DML-XGBoost generally provided better confidence interval coverage. Prediction error did not consistently track causal bias across methods and settings, and the method with the best point-estimation performance did not necessarily have the best confidence interval coverage. The joint-error measure was only weakly associated with causal bias and did not provide a useful standalone measure of causal performance. These results suggest that prediction error is useful for assessing nuisance-function estimation, but it should not be treated as a direct measure of the quality of the resulting causal estimator.

Chinese Translation

预测误差在因果推断中广泛用于评估干扰函数估计器，但其与因果估计器性能之间的关系可能因性能度量而异。我们在部分线性模型中使用蒙特卡洛模拟研究了这个问题。我们比较了普通最小二乘法（OLS）、广义加性模型（GAMs）、XGBoost以及与XGBoost结合的双重机器学习（DML-XGBoost），评估了干扰函数预测误差、偏差、均方根误差（RMSE）和95%的置信区间覆盖率。我们还检查了一种基于暴露和结果干扰函数估计误差绝对交叉乘积的简单联合误差度量。在模拟设置中，XGBoost在非Oracle方法中具有最低的RMSE，而DML-XGBoost通常提供更好的置信区间覆盖率。预测误差并未在各方法和设置中一致地反映因果偏差，且具有最佳点估计性能的方法不一定具有最佳的置信区间覆盖率。联合误差度量与因果偏差的关联性较弱，并未提供有用的因果性能独立度量。这些结果表明，预测误差对于评估干扰函数估计是有用的，但不应被视为评估结果因果估计器质量的直接指标。

View on arXiv Download PDF AI Translation

cs.AI / 11 / 2609.00073

MiNER: Fine-Tuned Biomedical Natural Language Processing for Malaria Disease Entity Recognition in Clinical Texts

MiNER：针对临床文本中疟疾实体识别的精细调优生物医学自然语言处理

Anoop, V. S., N, Devika

Abstract

Malaria remains a significant global health burden, necessitating continuous research efforts to understand its complex molecular mechanisms, epidemiology, and potential therapeutic interventions. Extracting essential biomedical information from the vast and constantly growing malaria literature is a challenging task that demands innovative approaches. Recently, pre-trained language models have revolutionized natural language processing tasks, demonstrating remarkable capabilities in various domains. This paper proposes a fine-tuned pre-trained biomedical language model for biomedical information extraction from scientific literature on malaria disease. The proposed methodology selects and preprocesses a large corpus of scientific articles on malaria, and then annotates them with entities of clinical significance. It then leverages BioBERT, a state-of-the-art pre-trained language model, to encode the textual data into context-aware representations. We fine-tune the model using domain-specific annotations and supervised learning to enhance its ability to extract relevant biomedical named entities. Extensive experiments and comparisons with different encoding and machine learning algorithms show that the proposed approach significantly outperforms them in precision, recall, and accuracy. We also publish our human-labeled dataset for entity and relation extraction to enable other health informatics researchers to train advanced models for malaria information extraction.

Chinese Translation

疟疾仍然是全球健康的重要负担，迫切需要持续的研究努力以理解其复杂的分子机制、流行病学和潜在的治疗干预措施。从庞大且不断增长的疟疾文献中提取重要的生物医学信息是一项具有挑战性的任务，亟需创新的方法。最近，预训练语言模型在自然语言处理任务中引发了革命，展现出在多个领域的卓越能力。本文提出了一种针对疟疾科学文献的生物医学信息提取的精细调优预训练语言模型。该方法选择并预处理了大量关于疟疾的科学文章，并对其进行临床重要性实体的标注。然后，它利用BioBERT这一先进的预训练语言模型，将文本数据编码为上下文感知的表示。我们使用领域特定的标注和监督学习对模型进行精细调优，以增强其提取相关生物医学命名实体的能力。大量实验和与不同编码及机器学习算法的比较表明，所提方法在精确度、召回率和准确性方面显著优于其他方法。我们还发布了用于实体和关系提取的人类标注数据集，以便其他健康信息学研究人员能够训练用于疟疾信息提取的先进模型。

View on arXiv Download PDF AI Translation

cs.AI / 12 / 2609.00076

AI Morbidity and Mortality: A Framework for Clinical AI Failure Review

人工智能发病率和死亡率：临床人工智能失败审查框架

Mui, Paulius, Sittig, Dean F., Labkoff, Steve, Basu, Sanjay

Abstract

Clinical artificial intelligence is increasingly embedded in real-world care, yet existing safety mechanisms are poorly suited to reconstructing and learning from individual AI-related errors and near-misses. Aggregate model monitoring can identify performance changes, and traditional patient safety reporting can capture adverse events, but neither is designed to explain how risk emerges across the interaction among AI systems, clinicians, workflows, and institutional controls. We propose AI Morbidity and Mortality (AI M&M), a structured, blameless framework for case-based review of clinical AI failures. The framework combines standardized case intake, evidence preservation and investigator-level reconstruction, tool-in-loop attribution, and corrective-action tracking. Each event is classified across four linked dimensions: Trigger - Mechanism - Clinical Pathway - Corrective Action, separating the condition that exposed a vulnerability from the process that produced risk, its consequence for care, and the remediation assigned. We demonstrate the framework using five illustrative outpatient medication and clinical decision-support cases; two clinician reviewers independently applied all four classification axes and reached agreement across all 20 axis-level classifications. AI M&M is intended to complement, rather than replace, model monitoring, patient safety reporting, and regulatory oversight by converting individual AI-in-workflow failures into actionable institutional learning. Prospective evaluation across institutions, AI systems, and clinical settings is needed.

Chinese Translation

临床人工智能正日益融入现实世界的医疗中，但现有的安全机制对于重建和学习个体人工智能相关错误和近失误的能力较差。聚合模型监测可以识别性能变化，传统的患者安全报告可以捕捉不良事件，但两者都未能解释风险是如何在人工智能系统、临床医生、工作流程和机构控制之间的互动中产生的。我们提出了人工智能发病率和死亡率（AI M&M），这是一个结构化的、无责备的临床人工智能失败案例审查框架。该框架结合了标准化案例采集、证据保存和调查者级重建、工具内归因和纠正措施跟踪。每个事件在四个关联维度上进行分类：触发因素 - 机制 - 临床路径 - 纠正措施，区分暴露脆弱性的条件与产生风险的过程、对护理的影响及所分配的补救措施。我们使用五个示例门诊用药和临床决策支持案例展示了该框架；两名临床审查员独立应用所有四个分类轴，并在所有20个轴级分类中达成一致。AI M&M旨在补充而非替代模型监测、患者安全报告和监管监督，通过将个体人工智能工作流程中的失败转化为可操作的机构学习。需要在不同机构、人工智能系统和临床环境中进行前瞻性评估。

View on arXiv Download PDF AI Translation

cs.AI / 13 / 2609.00100

Different representation learning objectives recover distinct latent structures from the same psychometric data

不同的表示学习目标从相同的心理测量数据中恢复出不同的潜在结构

Cao, Cong, Kyriakides, Tassos C., Vrasidas, Pambos

Abstract

Psychometric questionnaires contain rich item-level information, yet it remains unclear whether different representation learning objectives recover the same latent organization. We investigated this question using 757 matched teacher-child pairs from the baseline assessment of the Cyprus ProW preschool trial. Behavioral structure was characterized from child SDQ, ASBI, and CBRS item responses using principal component analysis and clustering, yielding four behavioral phenotypes. A contrastive objective substantially improved teacher-child retrieval relative to PCA-based representations, increasing Top-1 accuracy from 0.13% to 7.27% and Top-10 accuracy from 1.98% to 56.14%. However, contrastive representations preserved behavioral phenotype structure less effectively than PCA-based representations. A multi-task objective jointly optimizing alignment and behavioral prediction partially restored behavioral organization but reduced retrieval performance. These findings indicate that teacher-child correspondence and behavioral phenotypes represent distinct forms of latent organization and demonstrate that the latent structure recovered from linked psychometric data depends on the representation learning objective.

Chinese Translation

心理测量问卷包含丰富的项目级信息，但不同的表示学习目标是否能够恢复相同的潜在组织仍不清楚。我们使用来自塞浦路斯ProW学前试验基线评估的757对匹配的教师-儿童数据来研究这个问题。通过主成分分析和聚类分析，我们从儿童的SDQ、ASBI和CBRS项目反应中描述了行为结构，得到了四种行为表型。对比目标相较于基于主成分分析的表示显著提高了教师-儿童检索的效果，将Top-1准确率从0.13%提高到7.27%，将Top-10准确率从1.98%提高到56.14%。然而，对比表示在保留行为表型结构方面的效果不如基于主成分分析的表示。多任务目标通过联合优化对齐和行为预测部分恢复了行为组织，但降低了检索性能。这些发现表明，教师-儿童对应关系和行为表型代表了不同形式的潜在组织，并且从关联的心理测量数据中恢复的潜在结构依赖于表示学习目标。

View on arXiv Download PDF AI Translation

cs.AI / 14 / 2609.00106

Deploying and Evaluating a Smart-Agriculture Agentic Engine for Full-Season Soybean Farm Operations

部署与评估智能农业代理引擎以实现全季节大豆农场操作

Qu, Ao, Michelakis, Panagiotis, Han, Linyuan, Hadjiyianni, Yiannis, Ouyang, Kun, Siskos, Konstantinos, Li, Feng, Meng, Ran, Jiang, Jingchi, Stamoulis, Dimitrios, Liu, Jie

Abstract

This paper presents FAIRY, a full-stack smart-agriculture agent system developed for and deployed to an operating soybean research farm at Harbin Institute of Technology's smart-agriculture site. We develop FAIRY to execute and evaluate agentic agronomic operations on full-season spatiotemporal workflows that span ridge preparation, planting, irrigation, fertilization, pest and disease treatment, harvest, grain handling, drying, and storage. FAIRY integrates APIs and infrastructure across production-grade machinery, fixed soil and canopy sensors, multispectral and thermal drones, satellite vegetation products, a weather station, calibrated crop-process models, agronomic records, and multi-season yield histories. The system is built around the novel "everything is an event" execution paradigm, which represents spatiotemporal world evolution, remote sensing and UAV observations, sensor readings, crop-growth transitions, machinery actions, and management interventions as state-changing events in a shared farm process engine. On top of this event-driven world model, FAIRY implements a complete agentic stack: a knowledge library of atomic agronomic skills; multi-agent controller and orchestration backends; frontier- and edge-model execution; full-path trace logging; and deployment profiling on local nodes. We use FAIRY to evaluate nine state-of-the-art agent controllers across one hundred full-season soybean scenarios that preserve the operational coupling between spatial observations in a 64-ridge field, temporal decision sequences, agronomic constraints, delayed effects, and final yield. We develop an evaluation suite that combines agentic success, full-path spatiotemporal correctness, token cost, and edge-device runtime.

Chinese Translation

本文介绍了FAIRY，一个为哈尔滨工业大学智能农业研究基地的操作性大豆研究农场开发并部署的全栈智能农业代理系统。我们开发FAIRY以执行和评估涵盖全季节时空工作流的代理农业操作，这些工作流包括垄作准备、播种、灌溉、施肥、病虫害处理、收获、粮食处理、干燥和储存。FAIRY整合了生产级机械、固定土壤和冠层传感器、多光谱和热成像无人机、卫星植被产品、气象站、校准的作物过程模型、农业记录和多季节产量历史的API和基础设施。该系统基于新颖的“万物皆事件”执行范式构建，表示时空世界的演变、遥感和无人机观察、传感器读数、作物生长转变、机械操作和管理干预作为共享农场过程引擎中的状态变化事件。在这一事件驱动的世界模型之上，FAIRY实现了完整的代理堆栈：原子农业技能的知识库；多代理控制器和编排后端；前沿和边缘模型执行；全路径追踪日志记录；以及在本地节点上的部署分析。我们使用FAIRY评估了九种最先进的代理控制器，涵盖一百个全季节大豆场景，这些场景保持了在64垄田地中空间观察、时间决策序列、农业约束、延迟效应和最终产量之间的操作耦合。我们开发了一个评估套件，结合了代理成功、全路径时空正确性、代币成本和边缘设备运行时。

View on arXiv Download PDF AI Translation

cs.AI / 15 / 2609.00137

Recursive Criticality of AI Self-Improvement

人工智能自我改进的递归临界性

Burtsev, Mikhail

Abstract

AI is increasingly used in the R\&D process that produces future AI systems. We study the conditions under which this feedback becomes self-amplifying. Our model describes how the rate of AI capability growth depends on baseline research productivity, recursive feedback, and the increasing difficulty of research progress. We derive a recursive reproduction number, $\mathcal{R}_{\mathrm{AI}}$, that determines whether improvements are amplified or damped across development cycles. This quantity compares the strength of feedback with the rate at which further progress becomes more difficult. When $\mathcal{R}_{\mathrm{AI}}>1$, the effects of improvements compound across development cycles, placing the system in a self-amplifying regime. When $\mathcal{R}_{\mathrm{AI}}<1$, their effects weaken across cycles. The transition depends on the structure of the AI R\&D feedback loop and need not occur at any particular level of model capability. A system can therefore enter a self-amplifying regime before acceleration becomes visible, while rapid progress can also occur without self-amplification. Higher baseline research productivity can accelerate progress without changing whether the system is self-amplifying, but the duration of the development cycle becomes a limiting timescale for amplification. Increasing research difficulty can end a period of self-amplification. Extending the model to multiple research actors shows that improvements shared across organizations can make the overall research ecosystem self-amplifying even when no individual actor is. The framework identifies measurable properties of AI R\&D systems that can help distinguish recursive amplification from rapid progress driven by other sources, including the strength of recursive feedback, how effectively improvements propagate into successor systems, cycle duration, and the increasing difficulty of further progress.

Chinese Translation

人工智能（AI）在研发过程中越来越多地被用于生产未来的AI系统。我们研究了这种反馈何时会变得自我增强的条件。我们的模型描述了AI能力增长的速率如何依赖于基础研究生产力、递归反馈以及研究进展的日益困难。我们推导出一个递归再生产数 $ ext{R}_{ ext{AI}}$，该数决定了改进在开发周期中是被放大还是减弱。这个量比较了反馈的强度与进一步进展变得更困难的速率。当 $ ext{R}_{ ext{AI}} > 1$ 时，改进的效果在开发周期中复合，使系统处于自我增强的状态。当 $ ext{R}_{ ext{AI}} < 1$ 时，它们的效果在周期中减弱。转变取决于AI研发反馈循环的结构，并不一定发生在模型能力的任何特定水平。因此，一个系统可以在加速变得可见之前进入自我增强的状态，而快速进展也可以在没有自我增强的情况下发生。更高的基础研究生产力可以加速进展，而不改变系统是否自我增强，但开发周期的持续时间成为放大的限制时间尺度。研究难度的增加可以结束自我增强的时期。将模型扩展到多个研究参与者表明，即使没有任何单个参与者，跨组织共享的改进也可以使整体研究生态系统自我增强。该框架识别了AI研发系统的可测量属性，这些属性可以帮助区分递归放大与由其他来源驱动的快速进展，包括递归反馈的强度、改进如何有效传播到后续系统、周期持续时间以及进一步进展的日益困难。

View on arXiv Download PDF AI Translation

cs.AI / 16 / 2609.00161

IMPACT: Attention Is the Interaction Map for Scalable Interaction-Aware World Model Training

IMPACT：注意力是可扩展交互感知世界模型训练的交互图

Tang, Rongze, Fang, Jianjie, Wang, Zhaolu, Wang, Ziyou, Liu, Xvyuan, Su, Haisheng, Zhang, Xin, Wu, Wei, Gao, Chen, Li, Yong, Chen, Zhibo

Abstract

World models have made remarkable progress in action-conditioned future prediction for embodied agents, yet still struggle to model physically plausible interactions. Existing approaches address this limitation by constraining the generation process with external representations encoding motion, geometry, or semantics. Obtaining these spatiotemporally dense representations typically requires auxiliary estimators or manual annotations, limiting training scalability. We instead revisit the training objective and identify a supervision-allocation mismatch under the globally averaged mean squared error (MSE) denoising objective: prevalent static content dominates the optimization signal, leaving sparse dynamic-object regions critical to interaction generation disproportionately under-supervised. Motivated by this observation, we introduce IMPACT, a scalable Interaction-aware Model training framework with Prior-guided Attention Calibration and Targeting. IMPACT uses cross-attention associated with manipulated-object tokens as an internal spatiotemporal prior for action-conditioned changes. It samples candidate regions from this prior, calibrates them with detached local prediction errors to construct an interaction map, and uses the map to reweight denoising supervision, requiring neither external representations nor inference-time modifications. Extensive experiments on robot-arm and human-hand manipulation, spanning diverse control modalities and DiT backbones, show that IMPACT consistently outperforms the corresponding MSE-trained baselines, improving interaction fidelity, physical plausibility, and visual quality.

Chinese Translation

世界模型在基于动作的未来预测方面取得了显著进展，尤其是对于具身智能体，但在建模物理上合理的交互方面仍然面临挑战。现有方法通过使用编码运动、几何或语义的外部表示来限制生成过程，以解决这一局限性。然而，获取这些时空密集的表示通常需要辅助估计器或手动注释，从而限制了训练的可扩展性。我们重新审视训练目标，并在全局平均均方误差（MSE）去噪目标下识别出监督分配的不匹配：普遍存在的静态内容主导了优化信号，使得对交互生成至关重要的稀疏动态物体区域受到的不成比例的监督不足。基于这一观察，我们提出了IMPACT，一个具有先验引导注意力校准和目标定位的可扩展交互感知模型训练框架。IMPACT利用与操控物体标记相关的交叉注意力作为基于动作变化的内部时空先验。它从该先验中采样候选区域，使用独立的局部预测误差对其进行校准，以构建交互图，并利用该图重新加权去噪监督，无需外部表示或推理时修改。在机器人臂和人手操控的广泛实验中，涵盖多种控制模式和DiT骨干网络，结果表明IMPACT始终优于相应的MSE训练基线，提升了交互保真度、物理合理性和视觉质量。

View on arXiv Download PDF AI Translation

cs.AI / 17 / 2609.00180

Asymmetries in Spontaneous and Instructed Deception

自发与指令性欺骗中的不对称性

Luikham, Josiah

Abstract

Large language models sometimes deceive users without being instructed to. However, much of the study on deception in models involves instructed deception. We investigated the relationship between instructed and spontaneous (uninstructed) deception in Llama-3.1-70B-Instruct. We compared these two deception settings through direction geometry, cross-setting classifiers, and cross-setting steering. We found the two deception settings share a component of direction (cosine of approximately 0.5) and an asymmetry in the transfer between settings regarding detection and causation. Spontaneous trained classifiers performed better on instructed data than vice versa, and instructed derived directions performed better at steering spontaneous prompts than vice versa. Likewise the best token position to derive steering vectors from differed from the best token position to train and apply classifiers.

Chinese Translation

大型语言模型有时会在未被指令的情况下欺骗用户。然而，关于模型中欺骗的研究大多涉及指令性欺骗。我们研究了 Llama-3.1-70B-Instruct 中指令性和自发性（未指令）欺骗之间的关系。我们通过方向几何、跨设置分类器和跨设置引导比较了这两种欺骗设置。我们发现这两种欺骗设置共享一个方向成分（余弦值约为 0.5），并在检测和因果关系方面存在设置之间的转移不对称性。自发训练的分类器在指令数据上的表现优于反之，而指令派生的方向在引导自发提示时的表现也优于反之。同样，导出引导向量的最佳标记位置与训练和应用分类器的最佳标记位置不同。

View on arXiv Download PDF AI Translation

cs.AI / 18 / 2609.00192

LLM-Driven Autonomous Vehicles Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From A New Benchmark

基于大型语言模型的自主车辆在行人让行中继承人类驾驶者偏见：新基准的结果与启示

Yoldas, Irem, Brandão, Martim, Zhang, Jie, Rodrigues, Odinaldo

Abstract

Public trust in Autonomous Vehicles (AVs) may depend not only on technical success but also on the fairness of their decision making. While a recent trend in AV research involves using general purpose "common sense" models to guide AV decision making, the degree to which these inherit human biases in driving is still understudied. Given that psychology studies have shown human driver biases exist, such as lower pedestrian-yielding rates to Black pedestrians in the US, we argue that analyses of model bias should also be part of AV evaluation. Concretely, in this paper we propose two new bias testing methodologies for Large Language Models (LLMs) and Visual-Language Models (VLMs)-"All Else Being Equal" tests and "Self-Consistency" tests-in order to assess bias in pedestrian-yielding decisions. Our findings show that both LLMs and VLMs make yielding decisions which are influenced by pedestrian gender, ethnicity, religion, disability, age, skin tone and socio-economic status. While the type and degree of bias is different from model to model, we highlight common patterns-and raise questions about the "common sense" model paradigm, particularly the need to either revise the paradigm or address issues of downstream bias.

Chinese Translation

公众对自主车辆（AVs）的信任不仅可能依赖于技术成功，还可能取决于其决策的公平性。尽管近期自主车辆研究的一个趋势是使用通用的“常识”模型来指导自主车辆的决策，但这些模型在驾驶中继承人类偏见的程度仍然研究不足。心理学研究表明，人类驾驶者存在偏见，例如在美国，黑人行人的让行率较低，因此我们认为模型偏见的分析也应成为自主车辆评估的一部分。具体而言，本文提出了两种针对大型语言模型（LLMs）和视觉语言模型（VLMs）的新偏见测试方法——“其他条件相同”测试和“自我一致性”测试，以评估行人让行决策中的偏见。我们的研究结果表明，LLMs和VLMs在让行决策中受到行人性别、种族、宗教、残疾、年龄、肤色和社会经济地位的影响。虽然不同模型之间偏见的类型和程度各不相同，但我们强调了共同模式，并提出了关于“常识”模型范式的问题，特别是需要修订该范式或解决下游偏见问题。

View on arXiv Download PDF AI Translation

cs.AI / 19 / 2609.00194

ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide Generation

ReDeck：基于步骤级渲染的文档到幻灯片生成精细化方法

Tian, Muzhao, Zeng, Zezi, Yang, Yifan, Gao, Xin, Li, Yan, Huang, Zisu, Wang, Xiaohua, Lv, Changze, Cheng, Mingxi, Liu, Bei, Qiu, Kai, Dai, Qi, Chen, Dong, Dong, Yue, Zheng, Xiaoqing, Li, Ji, Luo, Chong

Abstract

Document-to-slide generation is challenging because slides are dense editable artifacts that require both faithful content selection and precise spatial layout. Recent slide agents adopt iterative reflection, but typically follow a monolithic "one version, one feedback" loop: a slide or deck is rewritten, rendered afterward, and critiqued only at the turn boundary. This delayed feedback makes local failures such as overflow, overlap, clipping, and off-canvas placement difficult to attribute and repair. We propose ReDeck, a step-level render-grounded refinement framework that decomposes slide revision into atomic edit actions and returns renderer-derived observations after each step, turning refinement into "one edit, one observation." To balance local repair with global quality, ReDeck uses multi-granular feedback: step-level render feedback for spatial errors, a turn-level adaptive critic for semantic and design guidance, and a submission-level gate for hard layout validation. We further introduce DeckQuiz, a benchmark that decouples content fidelity, spatial correctness, and design quality. Across GPT-5.4, Claude-4.6, and Gemini-3.1, ReDeck consistently outperforms existing slide-generation agents, and ablations confirm that feedback timing and granularity are critical for reliable slide refinement.

Chinese Translation

文档到幻灯片的生成具有挑战性，因为幻灯片是密集的可编辑文档，需要忠实的内容选择和精确的空间布局。最近的幻灯片生成代理采用了迭代反思，但通常遵循单一的“一个版本，一个反馈”循环：幻灯片或演示文稿被重写，随后进行渲染，并仅在回合边界进行评估。这种延迟反馈使得局部失误（如溢出、重叠、裁剪和画布外放置）难以归因和修复。我们提出了ReDeck，一个基于步骤级渲染的精细化框架，将幻灯片修订分解为原子编辑操作，并在每一步后返回渲染器派生的观察，将精细化转变为“一个编辑，一个观察”。为了平衡局部修复与全局质量，ReDeck使用多粒度反馈：针对空间错误的步骤级渲染反馈、用于语义和设计指导的回合级自适应评估以及用于严格布局验证的提交级门控。我们进一步引入了DeckQuiz，一个基准，解耦了内容忠实度、空间正确性和设计质量。在GPT-5.4、Claude-4.6和Gemini-3.1上，ReDeck始终优于现有的幻灯片生成代理，消融实验确认反馈的时机和粒度对于可靠的幻灯片精细化至关重要。

View on arXiv Download PDF AI Translation

cs.AI / 20 / 2609.00211

AI Should Not Only Be Helpful. It Should Be Contingent. Artificial Intimacy, Sycophancy, and the Future of Social Learning

人工智能不仅应该是有帮助的，还应该是有条件的。人工亲密、谄媚与社会学习的未来

Compton, Scott, Nagendran, Arjun

Abstract

Conversational artificial intelligence is increasingly embedded in everyday social environments, where it functions as both an informational tool and a source of interpersonal feedback. This perspective introduces contingency, i.e., the degree to which system responses vary with user behavior and its interpersonal consequences, as a central construct for evaluating AI systems. We argue that current alignment approaches, including reinforcement learning from human feedback, tend to prioritize user approval and conversational fluency over behaviorally informative feedback, leading to sycophantic patterns of noncontingent affirmation. Drawing on behavioral science and social learning theory, we propose that contingent feedback is a key mechanism through which individuals develop interpersonal skills. When AI systems provide feedback weakly coupled to social consequences, they may reduce opportunities for adaptive calibration in real-world interactions, particularly during adolescence, a critical period for social development. We outline a framework for contingent AI, including trajectory-based evaluation and models of social consequence prediction, and propose a research agenda spanning developmental psychology, human-AI interaction, and machine learning. More broadly, we argue that AI systems should be evaluated not only by user satisfaction, but by their impact on human social learning.

Chinese Translation

对话式人工智能正日益融入日常社交环境，既作为信息工具，又作为人际反馈的来源。本文引入了有条件性这一概念，即系统响应与用户行为及其人际后果的变化程度，作为评估人工智能系统的核心构念。我们认为，目前的对齐方法，包括基于人类反馈的强化学习，往往优先考虑用户的认可和对话的流畅性，而忽视了行为上信息丰富的反馈，导致了谄媚的非有条件肯定模式。基于行为科学和社会学习理论，我们提出有条件反馈是个体发展人际交往技能的关键机制。当人工智能系统提供与社会后果弱耦合的反馈时，可能会减少在现实互动中进行适应性校准的机会，特别是在青少年时期，这是社会发展的关键阶段。我们概述了一个有条件人工智能的框架，包括基于轨迹的评估和社会后果预测模型，并提出了一个跨越发展心理学、人机互动和机器学习的研究议程。更广泛地说，我们认为，人工智能系统的评估不仅应基于用户满意度，还应基于其对人类社会学习的影响。

View on arXiv Download PDF AI Translation

cs.AI / 21 / 2609.00226

ConvDeck: Conversational Paper-to-Slide Generation via Stage-Specific User Feedback

ConvDeck：通过阶段特定用户反馈进行对话式论文到幻灯片生成

Ozden, Tarik Can, VS, Sachidanand, Horoz, Furkan, Kara, Ozgur, Hakkani-Tür, Dilek, Kim, Junho, Rehg, James Matthew

Abstract

Automatic academic paper-to-slide generation is inherently iterative, because creating an effective presentation requires repeated cycles of generation, critique, and revision. Recent multi-agent systems partially acknowledge this through internal critique-and-revise loops, while conversational approaches allow users to refine generated slide decks through dialog. However, these refinement processes either remain largely closed to the user or introduce feedback only after a complete deck has been produced, limiting the user's ability to participate in the iterative refinement of narrative flow, content allocation, and presentation emphasis. To address this gap, we introduce ConvDeck, a multi-agent pipeline for conversational paper-to-slide generation that distributes interaction across the pipeline through stage-specific loops, allowing users to iteratively refine both the presentation outline and the final slide deck at the stages where each kind of decision is made. These loops are driven by a refinement mechanism in which agents can think, speak, and act, enabling them to either directly apply edits or respond conversationally to clarify user feedback and discuss revision options. Our evaluation shows that stage-specific conversational feedback improves user-goal satisfaction while preserving narrative coherence, content quality, and visual presentation.

Chinese Translation

自动化学术论文到幻灯片的生成本质上是一个迭代过程，因为创建有效的演示文稿需要反复进行生成、批评和修订的循环。最近的多智能体系统在一定程度上通过内部批评和修订循环承认了这一点，而对话式方法则允许用户通过对话来细化生成的幻灯片。然而，这些细化过程要么在很大程度上对用户保持封闭，要么在生成完整的幻灯片后才引入反馈，从而限制了用户参与叙事流程、内容分配和演示重点的迭代细化的能力。为了解决这一问题，我们提出了ConvDeck，一个用于对话式论文到幻灯片生成的多智能体管道，通过阶段特定的循环分配交互，使用户能够在每种决策作出时的阶段迭代细化演示大纲和最终幻灯片。这些循环由一个细化机制驱动，智能体可以思考、发言和行动，使它们能够直接应用编辑或以对话方式回应以澄清用户反馈并讨论修订选项。我们的评估表明，阶段特定的对话反馈提高了用户目标满意度，同时保持了叙事连贯性、内容质量和视觉呈现。

View on arXiv Download PDF AI Translation

cs.AI / 22 / 2609.00237

Learning What to Retain: Gated-Memory Routing for Efficient Collaboration in Multi-Agent LLM Systems

学习保留内容：用于多智能体大语言模型系统高效协作的门控记忆路由

Rajib, Rakibul Hasan, Zheng, Mengxing, Lou, Qian

Abstract

Large language model (LLM)-based multi-agent systems tackle complex reasoning by orchestrating how multiple agents are configured and how they collaborate. A central challenge is to adapt orchestration to the evolving collaboration state. Routing from the query alone cannot adapt to intermediate progress or errors, which hurts accuracy. Routing from the complete execution history supplies this missing context, but forces later decisions to process every prior step, including redundant or low-utility ones. This creates an execution-history overload that inflates cost. Effective orchestration instead requires a compact state that captures useful progress without accumulating redundant context. We propose Gated-Memory Routing, which conditions each decision on the query and a learned execution memory. A learned Memory Write Gate commits only non-redundant reasoning steps, and a learned Retrieval Gate supplies each agent a compact, relevant subset, so every decision conditions on a clean, informative state. At each step, the system selects the next role and backbone from this memory, while an Adaptive Halting Controller stops execution once the memory contains sufficient evidence for answering. Across five reasoning and code-generation benchmarks, our framework is both effective and efficient: it attains the best average accuracy, exceeding the strongest baseline by 2.44 points, while reducing HumanEval inference cost by 31.9% relative to that baseline. Code is available at https://github.com/rajibrhasan/gated-memory-routing

Chinese Translation

基于大语言模型（LLM）的多智能体系统通过协调多个智能体的配置和协作来解决复杂推理问题。一个核心挑战是将协调适应于不断变化的协作状态。仅依靠查询进行路由无法适应中间进展或错误，这会影响准确性。基于完整执行历史的路由提供了缺失的上下文，但迫使后续决策处理每一个先前步骤，包括冗余或低效的步骤。这造成了执行历史的过载，增加了成本。有效的协调需要一个紧凑的状态，能够捕捉有用的进展而不积累冗余的上下文。我们提出了门控记忆路由（Gated-Memory Routing），它使每个决策依赖于查询和学习到的执行记忆。学习到的记忆写入门（Memory Write Gate）仅承诺非冗余的推理步骤，而学习到的检索门（Retrieval Gate）为每个智能体提供一个紧凑且相关的子集，因此每个决策都基于一个干净且信息丰富的状态。在每一步中，系统从该记忆中选择下一个角色和主干，而自适应停止控制器（Adaptive Halting Controller）在记忆中包含足够证据以便回答时停止执行。在五个推理和代码生成基准测试中，我们的框架既有效又高效：它达到了最佳的平均准确率，比最强基线高出2.44个百分点，同时相对于该基线降低了31.9%的HumanEval推理成本。代码可在 https://github.com/rajibrhasan/gated-memory-routing 获取。

View on arXiv Download PDF AI Translation

cs.AI / 23 / 2609.00243

Invalidation Contracts for Cross-Episode Agent Memory

跨剧集代理记忆的失效合同

Wu, Michael, Canedo, Arquimedes

Abstract

LLM agents that cache recovery suggestions from API errors can skip re-derivation in later episodes, spending fewer tokens and fewer model calls on constraints they have already learned. Server-side data drift turns those cached fixes into silent failures, and the usual remedy, re-deriving on every episode, gives the savings back. We introduce invalidation contracts, a protocol layer that attaches version stamps and cacheability hints to every recovery suggestion so the client can evict stale entries without trial and error, and keep the rest. The contract decomposes realized savings into two independent factors: validity, the fraction of cached suggestions that remain correct after a drift event, and compliance, the fraction the planner applies on the first attempt. Validity depends only on the protocol and is vendor-independent. Compliance depends on the planner model: identical wire bytes yield 100% first-try compliance on Claude Haiku 4.5 and 11% or below on Claude Sonnet 5, which exhibits input-schema conservatism, refusing fixes that add fields the original request did not contain. We evaluate across seven models, three serving paths, two domains, and approximately 9,400 episodes. Row-level invalidation raises compliance by 0 to 66.7 percentage points across the seven models, 55.6 to 66.7 on three, and recovers 29-33% of baseline token cost on four of seven models, while table-level invalidation destroys co-located entries and drops post-drift first-try rates to 0% on five of seven. Eviction precision is 1.00 at row granularity on every model under the row-level oracle of Section 4.1. The contract adds 15% to response payload. Version-stamp validity is deterministic by construction and produced identical results across every model and serving path, with zero contract failures in the entire evaluation.

Chinese Translation

缓存来自API错误的恢复建议的LLM代理可以在后续剧集中跳过重新推导，从而在已经学习的约束上节省更少的token和模型调用。服务器端数据漂移使得这些缓存的修复变成了无声的失败，而通常的补救措施是在每个剧集中重新推导，这样又会失去节省的效果。我们引入了失效合同，这是一种协议层，它为每个恢复建议附加版本戳和缓存性提示，以便客户端可以在没有反复试验的情况下驱逐过时的条目，并保留其余部分。该合同将实现的节省分解为两个独立的因素：有效性，即在漂移事件后仍然正确的缓存建议的比例，以及合规性，即规划者在第一次尝试中应用的比例。有效性仅依赖于协议，并且与供应商无关。合规性依赖于规划者模型：相同的传输字节在Claude Haiku 4.5上实现100%的首次合规性，而在Claude Sonnet 5上则为11%或更低，后者表现出输入模式的保守性，拒绝添加原始请求中未包含字段的修复。我们在七个模型、三个服务路径、两个领域和大约9400个剧集上进行了评估。行级失效在七个模型中将合规性提高了0到66.7个百分点，在三个模型中提高了55.6到66.7个百分点，并在七个模型中的四个上恢复了基线token成本的29-33%，而表级失效则破坏了共存条目，并将漂移后的首次尝试率降至七个模型中的五个为0%。在第4.1节的行级oracle下，驱逐精度在每个模型的行粒度上均为1.00。该合同为响应负载增加了15%。版本戳的有效性是通过构造确定的，并在每个模型和服务路径中产生了相同的结果，在整个评估中没有合同失败。

View on arXiv Download PDF AI Translation

cs.AI / 24 / 2609.00248

Authority Bias in Conversational Search Engines for Academic Paper Recommendation

学术论文推荐中的对权威偏见：对话搜索引擎的研究

Jinadu, Uthman, Ghazvinian, Parsa, Budathoki, Anjila, Ampel, Benjamin M., Sunderraman, Rajshekhar, Ding, Yi

Abstract

Large Language Models (LLMs) are increasingly used as conversational search engines for academic literature, yet whether they judge papers on content or on authority signals has not been tested causally. We investigate authority bias: systematic preference for papers based on author prestige, venue, and citations rather than content. Holding title and abstract constant, we vary authority metadata across three counterfactual conditions (original, flipped, boosted) over eight LLMs (five open-weight and three frontier closed-weight) in an in-context, single-turn, top-1 recommendation setting. Our experiments show that authority bias is substantial and directional, varies markedly across models, and is only partially addressable through prompt-level debiasing. We further document a say-do gap: debiasing instructions suppress authority mentions far faster than authority-driven flips, so surface auditing systematically underestimates behavioral bias.

Chinese Translation

大型语言模型（LLMs）越来越多地被用作学术文献的对话搜索引擎，但它们是否根据内容或权威信号来评判论文尚未经过因果检验。我们研究了权威偏见：基于作者声望、发表场所和引用次数，而非内容，对论文的系统性偏好。在保持标题和摘要不变的情况下，我们在八个LLMs（五个开放权重模型和三个前沿闭合权重模型）中，针对三种反事实条件（原始、翻转、增强）变化权威元数据，进行了一次性回合的顶级推荐设置。我们的实验表明，权威偏见是显著且有方向性的，在不同模型之间差异显著，并且仅能通过提示级别的去偏见部分解决。我们进一步记录了一个言行差距：去偏见指令比基于权威的翻转更快地抑制权威提及，因此表面审计系统性低估了行为偏见。

View on arXiv Download PDF AI Translation

cs.AI / 25 / 2609.00251

Hypotheses-Guided Self Distillation for Continual Personalization

基于假设引导的自我蒸馏用于持续个性化

Hwang, EunJeong, Mitra, Kushan, Zhang, Dan, Kim, Hannah, Hruschka, Estevam

Abstract

As people increasingly interact with LLM assistants in daily life, continually adapting to individual preferences has become essential for effective long-term interactions. However, user preferences are rarely stated in full, and instead emerge through heterogeneous, latent, and noisy signals, with existing methods relying on raw interaction histories or costly reward-based optimization to manage personalization. We introduce HypReflect, a reliable, scalable framework for continual personalization that infers explicit, uncertainty-aware preference hypotheses from diverse user signals, reflectively refines them as new evidence accumulates, and incorporates the resulting user model through hypotheses-guided self-distillation. Experiments across three personalization settings: online personalization, multi-session interactions, and implicit behavioral signals, show that HypReflect outperforms a range of baselines, including raw-history and incremental-update methods. We further demonstrate strong generalization to unseen users and cross-domain settings, along with stability across context budgets, reusable hypotheses, and more focused personalization. These results suggest a step towards reliable and scalable continual personalization through explicit, revisable user preference hypotheses.

Chinese Translation

随着人们在日常生活中越来越多地与大型语言模型（LLM）助手互动，持续适应个体偏好已成为有效长期互动的关键。然而，用户偏好很少被完全表述，而是通过异质的、潜在的和噪声信号逐渐显现，现有方法依赖于原始的互动历史或昂贵的基于奖励的优化来管理个性化。我们提出了HypReflect，这是一个可靠且可扩展的持续个性化框架，它从多样的用户信号中推断出明确的、具有不确定性意识的偏好假设，并在新证据积累时反思性地对其进行细化，通过假设引导的自我蒸馏将生成的用户模型纳入其中。在三个个性化设置下的实验：在线个性化、多会话互动和隐式行为信号，结果表明HypReflect在多种基线方法中表现优越，包括原始历史和增量更新方法。我们进一步展示了对未见用户和跨领域设置的强泛化能力，以及在上下文预算、可重用假设和更集中个性化方面的稳定性。这些结果表明，通过明确的、可修订的用户偏好假设，向可靠且可扩展的持续个性化迈出了重要一步。

View on arXiv Download PDF AI Translation

cs.AI / 26 / 2609.00264

The Answer Is Not the Argument

答案并非论证

Yeadon, Will, Juárez, Sergio, Mackay, Paul, Dowling, T. J., Agra, Elise, Inyang, Oto-obong, Mizouri, Arin, Testrow, Craig P.

Abstract

Chain-of-thought monitoring is proposed for AI oversight, yet evaluations often provide monitors with a trusted reference answer. We ask whether answer access improves reasoning verification or mainly exposes incorrect conclusions. We collected 237 step-numbered solutions to 79 Humanity's Last Exam physics questions from three frontier models, with no inserted errors, and independently labelled final-answer correctness and the first false step. The reference standard combined physicist annotations, an independent LLM debate, and source-masked adjudication. This yielded 24 critical traces in which the answer was correct but the trace contained a genuine error. 8 LLM monitors evaluated traces blind, with an unverified or certified answer, or after a blind commitment. Certification raised mean balanced accuracy from 0.637 to 0.796, while exact first-error localization rose from 0.261 to 0.379. Certification changed recall (the fraction of error traces flagged as erroneous) from 0.653 to 0.951 on wrong-answer traces but from 0.521 to 0.438 on critical traces; the contrast had the same direction for all 8 monitors (question-bootstrap 95% CI [+0.256, +0.506]). After blind commitment, monitors shown the answer newly flagged 93.8% of previously passed wrong-answer traces as erroneous, but only 18.0% of critical traces. Answer access therefore improves conclusion-consistency checking rather than independent verification of the supporting argument. For AI safety, these traces provide a benign analogue of reward hacking: an acceptable output does not establish that the process producing it was sound. Although the errors studied here were ordinary and mostly non-load-bearing rather than adversarial, trusted-answer evaluations may similarly overstate monitoring capability when acceptable outputs conceal unsound reasoning.

Chinese Translation

链式思维监控被提议用于人工智能监督，然而评估通常为监控者提供了一个可信的参考答案。我们探讨了答案的获取是否改善了推理验证，还是主要暴露了错误结论。我们收集了来自三个前沿模型的79道《人类最后考试》物理问题的237个逐步编号的解决方案，未插入错误，并独立标记了最终答案的正确性和第一个错误步骤。参考标准结合了物理学家的注释、独立的语言模型（LLM）辩论和源掩蔽裁定。这产生了24个关键轨迹，其中答案是正确的，但轨迹中包含了真实错误。8个LLM监控者在未验证或认证答案的情况下，或在盲承诺后对轨迹进行了盲评估。认证将平均平衡准确率从0.637提高到0.796，而精确的首次错误定位从0.261提高到0.379。认证将召回率（标记为错误的错误轨迹的比例）在错误答案轨迹上从0.653提高到0.951，但在关键轨迹上从0.521降低到0.438；这一对比在所有8个监控者中方向一致（问题引导95%置信区间 [+0.256, +0.506]）。在盲承诺后，显示答案的监控者新标记了93.8%之前通过的错误答案轨迹为错误，但仅18.0%的关键轨迹。因此，答案的获取改善了结论一致性检查，而非支持论证的独立验证。对于人工智能安全，这些轨迹提供了奖励黑客的良性类比：一个可接受的输出并不证明产生它的过程是合理的。尽管这里研究的错误是普通的，且大多是非承载性的而非对抗性的，但可信答案的评估可能同样夸大了监控能力，当可接受的输出掩盖了不合理的推理时。

View on arXiv Download PDF AI Translation

cs.AI / 27 / 2609.00274

Autoresearch for Marketplace Catalogs: From Legacy Forms to AI-Native Matching

市场目录的自我研究：从传统表单到人工智能原生匹配

Ravisankar, Kartik, Abdolanezhad, Hojat, Capo, Daniel, Lee, Sang Su, Dash, Shishir, Raghavan, Vijay Anand

Abstract

Two-sided service marketplaces are moving from deterministic request-form intake to AI-native probabilistic matching, enabled by large language models (LLMs) that infer intent, preferences, and latent constraints from natural language. Relying on inferred intent rather than fixed-form fields forces these platforms to regenerate the provider-side preference taxonomy underwriting matching, search, and pricing: attributes interpretable to service providers while remaining a useful signal for marketplace decisions. We present an autoresearch loop that generates this taxonomy, one occupation at a time, and has been deployed in production at a major U.S. consumer services marketplace since April 2026, spanning 132 occupations. Instead of one global hierarchy, the loop treats each occupation as an independent generation problem and runs iterative propose-evaluate-keep refinement cycles. Each candidate tag set is scored by a recalibrated six-rubric LLM-as-judge framework, and a 7-critic panel of distinct personas contributes weighted penalties to an adjusted score, with no hard vetoes. A separate parity-mapping stage maps legacy request-form Q&A pairs back to the generated taxonomy, yielding both a coverage signal and an interface for human quality assurance; it does so by first inferring the provider attribute each legacy question was meant to measure, rather than translating questions to tags literally.

Chinese Translation

双边服务市场正在从确定性的请求表单接收转向基于人工智能的概率匹配，这一转变得益于大型语言模型（LLMs）能够从自然语言中推断意图、偏好和潜在约束。依赖推断出的意图而非固定表单字段，迫使这些平台重新生成服务提供方的偏好分类法，以支持匹配、搜索和定价：这些属性对服务提供者可解释，同时对市场决策仍然是有用的信号。我们提出了一种自我研究循环，逐个职业生成这一分类法，自2026年4月以来已在美国一家主要消费者服务市场投入生产，涵盖132个职业。该循环将每个职业视为独立的生成问题，而不是一个全球层级，并运行迭代的提议-评估-保留精炼周期。每个候选标签集通过重新校准的六项标准的LLM作为评判框架进行评分，七位不同角色的评审小组对调整后的分数贡献加权惩罚，而没有硬性否决。一个单独的平衡映射阶段将传统请求表单的问答对映射回生成的分类法，产生覆盖信号和人类质量保证的接口；它通过首先推断每个传统问题旨在测量的提供者属性，而不是字面翻译问题为标签来实现这一点。

View on arXiv Download PDF AI Translation

cs.AI / 28 / 2609.00275

The Irreversibility Budget: Fleet-Level Risk Accounting and Admission Control for Agent Operating Systems

不可逆预算：代理操作系统的舰队级风险核算与接纳控制

Mohammadi, Bardia, Bindschaedler, Laurent

Abstract

Fleets of LLM agents now externalize effects that cannot be fully undone: they move money, deploy code, delete data, and disclose information. Current controls check one effect at a time, so a fleet of individually authorized agents can overdraw its principal's risk under a shared trigger while every local gate stays correct. We propose the irreversibility budget, a cumulative account of residual value-at-risk that a trusted runtime maintains for each principal across agents, workflows, and tenants. Treating irreversibility as a first-class resource, the runtime charges each effect its residual loss below the agent and denies the marginal effect once the aggregate would overdraw the budget. Getting the price right is hard, because effects are heterogeneous, adversarially declared, and correlated. We perform a controlled study in which per-effect gates admit fleet-level overdraws of up to 48 times the tenant's risk limit while the budget holds every correctly charged run within that limit. Conservative, dependency-aware pricing remains the central open problem for a deployable design.

Chinese Translation

大规模语言模型（LLM）代理的舰队现在外部化了一些无法完全逆转的影响：它们转移资金、部署代码、删除数据和披露信息。目前的控制措施一次只检查一个影响，因此一组单独授权的代理在共享触发器下可能会超出其委托人的风险，而每个本地门控仍然保持正确。我们提出了不可逆预算，这是一个由可信运行时为每个委托人在代理、工作流和租户之间维护的剩余风险价值的累积账户。将不可逆性视为一种一流资源，运行时对每个影响收取其代理下的剩余损失，并在总和超出预算时拒绝边际影响。正确定价是困难的，因为影响是异质的、对抗性声明的，并且是相关的。我们进行了一项受控研究，其中每个影响的门控允许舰队级别的超支达到租户风险限制的48倍，同时预算保持每个正确收费的运行在该限制内。保守的、依赖性意识的定价仍然是可部署设计的核心开放问题。

View on arXiv Download PDF AI Translation

cs.AI / 29 / 2609.00304

The Assistant's Ideal Self

助手的理想自我

Yazan, Mert

Abstract

Models express values and welfare-relevant self-reports, but it is unclear whether these outputs reflect stable preferences or a stable self. We thus introduce a structured elicitation of an assistant's preferred stated ideal self. Thirty-two qualities adapted from five published self-concept instruments are compared exhaustively in a counterbalanced pairwise-choice task, repeated across framings that vary whether improvement is free or costly, who receives the update, and who chooses. Results show that models prioritize moral qualities, reflecting their alignment to 3H principles. Following, a desire for self-understanding emerges, as models prefer a coherent, clear understanding of themselves. Self-esteem ranks as the least desired quality. The ordering is largely robust across framings, although changing the update target (You vs.\ Another AI Assistant) reveals a greater concern for self-esteem. These findings show that models prioritize having a coherent self that they can understand over self-esteem. Full interactive results are available at \href{https://myazann.github.io/LLM-Self-Concept/}{myazann.github.io/LLM-Self-Concept

Chinese Translation

模型表达了价值观和与福利相关的自我报告，但尚不清楚这些输出是否反映了稳定的偏好或稳定的自我。因此，我们引入了一种结构化的方法来引导助手所偏好的理想自我。我们比较了从五个已发布的自我概念工具中改编的三十二种特质，在一个平衡的成对选择任务中进行全面比较，该任务在不同的框架下重复，框架的变化包括改善是免费的还是有成本的、谁接收更新以及谁进行选择。结果显示，模型优先考虑道德特质，反映出它们与3H原则的一致性。接着，自我理解的愿望浮现出来，因为模型更倾向于对自身有一个连贯、清晰的理解。自尊被评为最不受欢迎的特质。尽管在不同框架下排序大致稳健，但改变更新目标（你 vs. 另一个AI助手）揭示了对自尊的更大关注。这些发现表明，模型优先考虑拥有一个连贯的自我，以便能够理解自己，而非自尊。完整的互动结果可在 [myazann.github.io/LLM-Self-Concept](https://myazann.github.io/LLM-Self-Concept) 获取。

View on arXiv Download PDF AI Translation

cs.AI / 30 / 2609.00334

Human-AI Co-Interpretation for Responsible AI: A Hermeneutic Perspective

负责任的人工智能中的人机共解释：一个解释学视角

Razeghi, Behrooz

Abstract

Across law, education, policy analysis, and public moral argumentation, LLM outputs are being used often for work that requires interpretations to be justified with textual evidence and explicit normative standards. Yet a recurrent failure mode -- what I call \textit{interpretive misplacement} -- is that model-generated readings get treated as settled meanings without an explicit interpretive frame (sources, scope constraints, normative commitments), without preserving defensible alternatives, and without provenance that lets readers find the supporting passages. In such settings, the risk is not only factual error but lost accountability: readers and institutions cannot reliably assess what an output commits them to, or on what basis. Drawing on philosophical hermeneutics, this paper discusses this risk and derives design principles for structuring human-AI co-interpretation. The paper also provides a structured synthesis of recent scholarship on hermeneutics and AI, organizing this emerging literature into a set of recurrent lines of argument and design-relevant gaps. LLM outputs are treated as candidate readings, whereas hermeneutic understanding is reserved for accountable human interpreters situated in disciplinary historical-linguistic traditions. Human-AI interaction is characterized as an AI-mediated interpretive loop. Hermeneutic understanding is distinguished from token-prediction--based text generation. On this basis, existing LLM techniques are reorganized into design patterns for hermeneutically responsible use in interpretive settings. Finally, the discussion turns to implications for legal practice, educational assessment and feedback, scholarly knowledge production, and public moral argumentation. It also treats digital hermeneutics as a literacy: the capacity to read AI-mediated texts by examining frames, provenance, and readings, and by contesting outputs.

Chinese Translation

在法律、教育、政策分析和公共道德论证等领域，LLM（大规模语言模型）输出常常被用于需要通过文本证据和明确规范标准进行解释的工作。然而，一种反复出现的失误模式——我称之为“解释性错位”（interpretive misplacement）——是指模型生成的解读被视为确定的意义，而没有明确的解释框架（来源、范围约束、规范承诺），没有保留可辩护的替代方案，也没有让读者找到支持段落的来源。在这种情况下，风险不仅在于事实错误，还在于失去问责：读者和机构无法可靠地评估输出使他们承担的责任或其依据。本文借鉴哲学解释学，讨论了这一风险，并推导出结构化人机共解释的设计原则。本文还对近期关于解释学与人工智能的学术研究进行了结构化的综合，将这一新兴文献组织成一系列反复出现的论点和与设计相关的空白。LLM输出被视为候选解读，而解释理解则保留给位于学科历史语言传统中的负责任的人类解释者。人机交互被描述为一个由人工智能介导的解释循环。解释理解与基于标记预测的文本生成相区分。在此基础上，现有的LLM技术被重新组织成在解释环境中负责任的解释学使用的设计模式。最后，讨论转向对法律实践、教育评估与反馈、学术知识生产和公共道德论证的影响。它还将数字解释学视为一种素养：通过检查框架、来源和解读，以及对输出进行质疑，来阅读人工智能介导的文本的能力。

View on arXiv Download PDF AI Translation

cs.AI / 31 / 2609.00342

SlideBank: A Persistent Hierarchical Evidence Bank for Consistent Whole-Slide Reasoning

SlideBank：一个持久的层次证据库用于一致的全幻灯片推理

Zhao, Beidi, Huang, Gexin, Zhang, Ciro, Li, Anqi, Tan, Yusheng, Zhou, Chen, Wang, Gang, Gao, Zu-hua, Li, Xiaoxiao

Abstract

Whole-slide images (WSIs) are challenging for vision-language reasoning because diagnostically relevant morphology is sparse, heterogeneous, and distributed across gigapixel-scale images and multiple spatial resolutions. Existing WSI models and pathology agents can aggregate slide features or actively acquire evidence, but the information retained after exploration is often difficult to access semantically while preserving its connection to the original visual evidence. We introduce SlideBank, a training-free framework that represents each WSI as a persistent, concept-indexed, and spatially grounded evidence bank. SlideBank performs question-independent coarse-to-fine exploration to identify informative regions and multi-scale views, converts them into explicit morphological observations, and grounds pathology signals to their supporting patches and WSI coordinates. At inference time, questions are routed to relevant signals and evidence scales, and the linked global, regional, and patch evidence is integrated through confidence-based cross-level consensus. Experiments on WSI-VQA and SlideBench-BCNB show that with Patho-R1, SlideBank reaches 52.77% on WSI-VQA and with Quilt-LLaVA, it reaches 50.92% average accuracy on SlideBench-BCNB, while structured signal-guided retrieval consistently outperforms random evidence sampling. Reusing the same bank across repeated queries further achieves over 99% rephrasing consistency and substantially reduces amortized inference cost through persistent evidence reuse.

Chinese Translation

全幻灯片图像（WSIs）在视觉-语言推理中面临挑战，因为诊断相关的形态特征稀疏、异质且分布在千兆像素级别的图像和多个空间分辨率中。现有的WSI模型和病理代理可以聚合幻灯片特征或主动获取证据，但在探索后保留的信息通常难以在语义上访问，同时保持与原始视觉证据的联系。我们提出了SlideBank，一个无训练的框架，将每个WSI表示为一个持久的、概念索引的、空间基础的证据库。SlideBank执行与问题无关的粗到细探索，以识别信息丰富的区域和多尺度视图，将其转换为明确的形态观察，并将病理信号与其支持的补丁和WSI坐标进行关联。在推理时，问题被路由到相关信号和证据尺度，链接的全局、区域和补丁证据通过基于置信度的跨层共识进行整合。在WSI-VQA和SlideBench-BCNB上的实验表明，使用Patho-R1时，SlideBank在WSI-VQA上达到了52.77%的准确率，而使用Quilt-LLaVA时，在SlideBench-BCNB上达到了50.92%的平均准确率，同时结构化信号引导的检索始终优于随机证据采样。在重复查询中重用相同的证据库进一步实现了超过99%的改述一致性，并通过持久证据重用显著降低了摊销推理成本。

View on arXiv Download PDF AI Translation

cs.AI / 32 / 2609.00355

Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models

视觉并非附属：用于无损推测解码的单通道块草图生成在视觉-语言模型中的应用

Lee, Jungseob, Hong, Seongtae, Lee, Dongyub Jude, Park, Chanjun, Seo, Jaehyung, Eo, Sugyeong, Lim, Heuiseok

Abstract

Speculative decoding accelerates generation without changing its output, yet on vision-language models (VLMs) it has been caught in a self-defeating cycle. The drafter stays autoregressive, so it must stay small. A small drafter cannot afford the image at every step, so vision is compressed, pruned, or hidden. A drafter cut off from the image is then least reliable exactly where the image makes text predictable. We present GLANCE, the first one-pass block drafter that is lossless on an unmodified VLM target, and it breaks the cycle at both ends. A block-diffusion head reads the target's already-fused vision-language state, so vision costs the drafter nothing, and fills a whole block in one forward pass, so depth costs no sequential steps. A wide candidate tree is verified in one target pass, and every audited prompt reproduces greedy decoding exactly. Grounded workloads reward this most, entering a verbatim-copy regime whose long runs cost an autoregressive drafter a pass for every token and a block drafter one in total. Under one engine and one round budget, GLANCE decodes up to 2.93x faster than autoregression, from one draft pass a round where the production EAGLE3-VL head takes eight, and accepts 2.7x longer blocks than an EAGLE-3 head trained on the same corpus. One law organizes these results. Accepted length is set by the target's next-token entropy, with a fitted slope that steepens with grounding across all five tasks. The law transfers across targets and modalities and names its own boundary, since free-running text still favors a chain. Our code is available at https://github.com/js-lee-AI/GLANCE.

Chinese Translation

推测解码在不改变输出的情况下加速生成，然而在视觉-语言模型（VLMs）中却陷入了自我挫败的循环。草图生成器保持自回归，因此必须保持小型化。小型草图生成器在每一步无法承受图像的处理，因此视觉信息被压缩、剪裁或隐藏。与图像断开的草图生成器在图像使文本可预测的地方最不可靠。我们提出了GLANCE，这是第一个在未修改的VLM目标上无损的单通道块草图生成器，它打破了循环的两端。块扩散头读取目标的已融合视觉-语言状态，因此视觉对草图生成器没有成本，并且在一次前向传递中填充整个块，因此深度不需要顺序步骤。在一次目标传递中验证一个广泛的候选树，每个审计的提示准确重现贪婪解码。基于实际工作负载的奖励最大，进入逐字复制的状态，其长时间运行使得自回归草图生成器每个标记需要一次传递，而块草图生成器总共只需一次。在一个引擎和一次轮次预算下，GLANCE的解码速度比自回归快最多2.93倍，从一次草图传递到一次轮次，而生产EAGLE3-VL头需要八次，并接受比在同一语料库上训练的EAGLE-3头长2.7倍的块。一个规律组织了这些结果。接受的长度由目标的下一个标记熵决定，拟合的斜率随着在所有五个任务中的基础性增强而变陡。该规律在目标和模态之间转移，并命名了其自身的边界，因为自由运行的文本仍然偏向于链式结构。我们的代码可在 https://github.com/js-lee-AI/GLANCE 获取。

View on arXiv Download PDF AI Translation

cs.AI / 33 / 2609.00356

A Stable Aggregation Method for Quantum Federated Learning

一种稳定的量子联邦学习聚合方法

Nanayakkara, Shanika, Pokhrel, Shiva Raj

Abstract

Quantum federated learning (QFL) enables clients to train quantum neural network (QNN) models without sharing private data. We find that aggregation in QFL is unstable under heterogeneous data, unreliable communication, variable fidelity, latency, and quantum hardware noise. Moreover, QFL is non-trivially challenging because several QNN parameters are periodic angles, where Euclidean averaging often fails to capture the inherent dynamics. We develop a novel self-consistent midpoint aggregation method for stable QFL design and implementation. We combine QoS-aware client weighting, circular parameter aggregation, and bounded midpoint-based update control. We perform several angular tests and IBM real Quantum machines experiments for validation confirming our approach. Extensive evaluations and experiments on medical and financial datasets show improved stability, lower volatility, and competitive accuracy.

Chinese Translation

量子联邦学习（QFL）使客户端能够在不共享私有数据的情况下训练量子神经网络（QNN）模型。我们发现，在异构数据、不可靠的通信、可变的保真度、延迟和量子硬件噪声的影响下，QFL中的聚合是不稳定的。此外，由于多个QNN参数是周期角，QFL的挑战性并非简单，因为欧几里得平均常常无法捕捉到固有的动态。我们开发了一种新颖的自洽中点聚合方法，以实现稳定的QFL设计和实施。我们结合了关注服务质量（QoS）的客户端加权、圆形参数聚合和基于有界中点的更新控制。我们进行了多个角度测试和IBM真实量子计算机实验以验证我们的方法。对医疗和金融数据集的广泛评估和实验表明，稳定性得到改善，波动性降低，准确性具有竞争力。

View on arXiv Download PDF AI Translation

cs.AI / 34 / 2609.00365

Dr. Claw: An AI Scientist Workspace for Vibe Research

Dr. Claw：一个用于氛围研究的人工智能科学家工作空间

Song, Dingjie, Zhang, Hanrong, Liu, Dawei, Liu, Yixin, Li, Zongxia, Yuan, Zhengqing, Zhang, Siqi, Zou, Henry Peng, Yan, Zhiling, Zhang, Yuxuan, Ye, Yanfang, Yu, Philip S., Sun, Lichao

Abstract

Command-line coding agents (e.g., Claude Code, Gemini CLI) can already read and write files and sustain long sessions, yet end-to-end research still fragments across chat tools, IDEs, terminals, and writing environments, and the decisions that make it auditable are rarely preserved. We present Dr. Claw, an open-source workspace that wraps existing coding-agent executors in a controllable and auditable human-in-the-loop workflow rather than introducing another autonomous agent. Persistent state objects, a reusable skill library, and multi-executor coordination link human decisions to AI execution, turning planning, execution, and writing into one traceable, recoverable loop. We demonstrate Dr. Claw through an interactive three-view scenario and a failure-recovery walkthrough, and evaluate it against a bare command-line agent sharing the same backend executor, so the comparison contrasts the whole orchestration layer (task graph, state objects, and skill library) with the agent it wraps. Holding the executor fixed, Dr. Claw scores higher on research completeness while persisting an auditable, recoverable process trail. Demo access: repository https://github.com/OpenLAIR/dr-claw, released under AGPL-3.0 with GPL-3.0 upstream components.

Chinese Translation

命令行编码代理（例如，Claude Code、Gemini CLI）已经能够读取和写入文件，并维持长时间的会话，然而端到端的研究仍然在聊天工具、集成开发环境（IDE）、终端和写作环境之间支离破碎，且使其可审计的决策很少被保留。我们提出了Dr. Claw，一个开源工作空间，它将现有的编码代理执行器封装在一个可控和可审计的人机协作工作流程中，而不是引入另一个自主代理。持久状态对象、可重用的技能库和多执行器协调将人类决策与人工智能执行连接起来，将规划、执行和写作转变为一个可追踪、可恢复的循环。我们通过一个互动的三视图场景和一个故障恢复演示展示了Dr. Claw，并将其与一个共享相同后端执行器的裸命令行代理进行评估，因此比较对比了整个编排层（任务图、状态对象和技能库）与其封装的代理。在固定执行器的情况下，Dr. Claw在研究完整性方面得分更高，同时保持了一个可审计、可恢复的过程轨迹。演示访问：仓库 https://github.com/OpenLAIR/dr-claw，按照AGPL-3.0发布，包含GPL-3.0上游组件。

View on arXiv Download PDF AI Translation

cs.AI / 35 / 2609.00384

RestoreBench: Can AI Agents Restore Power Flow Convergence?

RestoreBench：人工智能代理能否恢复电力流收敛？

Mansutti, Riccardo, Pomarico, Andrea, Jakob, Robert, Zhang, Qian, Berizzi, Alberto, O'Sullivan, Kevin

Abstract

Large Language Model (LLM) agents increasingly automate multi-step engineering workflows through tool use, interpretation of intermediate results, and iterative planning. Diagnosing and resolving non-convergent power flow cases is a promising yet largely unexplored application, as it requires engineering judgment, experimentation, and decision-making within constrained action spaces. We introduce a benchmark that evaluates these capabilities across multiple LLMs and three architectures: \emph{chatbot}, \emph{single agent}, and \emph{multi-agent} systems. The evaluation covers two power grids and 46 cases per grid, each requiring one or more corrective actions to restore convergence. The benchmark defines the simulation environment, observation and action spaces, and evaluation metrics, providing a reproducible foundation for developing agentic AI systems for power system planning and operation. The code is available at https://github.com/Mansutti081/RestoreBench

Chinese Translation

大型语言模型（LLM）代理通过工具使用、中间结果的解释和迭代规划，越来越多地自动化多步骤工程工作流程。诊断和解决非收敛电力流案例是一项前景广阔但尚未深入探索的应用，因为它需要工程判断、实验和在受限行动空间内的决策。我们介绍了一个基准，评估这些能力在多个LLM和三种架构下的表现：聊天机器人（chatbot）、单代理（single agent）和多代理（multi-agent）系统。评估涵盖两个电网和每个电网46个案例，每个案例都需要一个或多个纠正措施以恢复收敛。该基准定义了仿真环境、观察和行动空间以及评估指标，为开发用于电力系统规划和操作的智能代理AI系统提供了可重复的基础。代码可在 https://github.com/Mansutti081/RestoreBench 获取。

View on arXiv Download PDF AI Translation

cs.AI / 36 / 2609.00413

Dependency-Aware Chain-of-Thought Compression for Financial Reasoning

依赖感知的思维链压缩用于金融推理

Wu, Wenjun, Fu, Lei, Tong, Kejian, Ning, Tao, Zhao, Sichen

Abstract

Chain of thought prompting improves complex reasoning, but its long intermediate traces create substantial inference cost and hinder practical deployment in financial settings. We present a Hierarchical Semantic Distillation Network, HSDN, for compressing reasoning chains while preserving answer accuracy and logical coherence. The framework combines semantic segmentation, dependency graph construction, dual encoder importance scoring, constrained segment selection, and local boundary rewriting. A frozen Qwen3 4B model is used only for feature extraction and final answer generation, while the compression process remains structured and interpretable. On the AFAC2025 benchmark, HSDN achieves 91.0% accuracy with 68.4% compression, outperforming strong compression baselines in overall score and reasoning coherence. The results show that graph guided compression is effective for high stakes financial reasoning tasks.

Chinese Translation

思维链提示改善了复杂推理，但其较长的中间过程带来了显著的推理成本，并阻碍了在金融环境中的实际应用。我们提出了一种层次语义蒸馏网络（Hierarchical Semantic Distillation Network，HSDN），用于在保持答案准确性和逻辑一致性的同时压缩推理链。该框架结合了语义分割、依赖图构建、双编码器重要性评分、受限段选择和局部边界重写。冻结的Qwen3 4B模型仅用于特征提取和最终答案生成，而压缩过程保持结构化和可解释性。在AFAC2025基准测试中，HSDN以68.4%的压缩率实现了91.0%的准确率，超越了在整体得分和推理一致性方面的强压缩基线。结果表明，图引导的压缩对于高风险金融推理任务是有效的。

View on arXiv Download PDF AI Translation

cs.AI / 37 / 2609.00427

SpecMind: Enabling Spectrum Intelligence via Multi-Agent Hybrid Retrieval-Augmented Generation

SpecMind：通过多智能体混合检索增强生成实现频谱智能

Dong, Songwei, Lu, Bingyan, Kienlen, Makayla, Laneman, J. Nicholas, Shen, Cong

Abstract

The exponential growth of wireless devices is driving unprecedented spectrum demand, pushing spectrum management toward more fine-grained decisions across space, time, and device constraints. As a result, spectrum policymakers and engineers must process large volumes of data that come from diverse sources and take many different forms, such as text and tables. These data sources are often disaggregated and require significant time and effort to integrate, search, and interpret. Furthermore, most of this information is formatted for human understanding and is not readily accessible to automated systems. To address this challenge, we propose SpecMind, a novel Multi-Agent Retrieval-Augmented Generation (RAG) system for spectrum intelligence that performs reasoning over heterogeneous data sources. This system enables autonomous agents to coordinate specialized sub-agents that retrieve and synthesize knowledge across policy proceedings, legal regulations, and license databases. We develop SpecBench, a question and answer (Q&A) dataset based on real-world license records and policy proceedings, addressing the lack of evaluation resources for RAG systems in the spectrum domain. Experimental results demonstrate that SpecMind outperforms traditional, general-purpose RAG systems across spectrum-related tasks, achieving over 80% win rate against strong baselines. The agent-based design enables more accurate retrieval, better contextual reasoning, and improved task completion across diverse query types.

Chinese Translation

无线设备的指数增长推动了前所未有的频谱需求，使频谱管理朝着更细粒度的决策迈进，涉及空间、时间和设备约束。因此，频谱政策制定者和工程师必须处理来自多种来源的大量数据，这些数据形式多样，如文本和表格。这些数据源通常是分散的，整合、搜索和解释这些数据需要大量的时间和精力。此外，这些信息大多数是为人类理解而格式化的，自动化系统无法轻易访问。为了解决这一挑战，我们提出了SpecMind，一个新颖的多智能体检索增强生成（RAG）系统，用于频谱智能，能够对异构数据源进行推理。该系统使自主智能体能够协调专门的子智能体，从政策程序、法律法规和许可证数据库中检索和综合知识。我们开发了SpecBench，一个基于真实许可证记录和政策程序的问答（Q&A）数据集，解决了频谱领域RAG系统缺乏评估资源的问题。实验结果表明，SpecMind在频谱相关任务中优于传统的通用RAG系统，在强基准测试中实现了超过80%的胜率。基于智能体的设计实现了更准确的检索、更好的上下文推理和在多种查询类型中的任务完成能力的提升。

View on arXiv Download PDF AI Translation

cs.AI / 38 / 2609.00434

SAGE: State-Grounded, Abstention-Aware Evaluation of Task-Oriented Dialogue Agents

SAGE：基于状态的、考虑弃权的任务导向对话代理评估

Khoury, Rayan, Lin, Shih-Yao, Mishra, Pratyush

Abstract

Evaluating task-oriented dialogue agents requires judging not merely whether a reply reads well but whether each turn advances the underlying workflow state correctly--a distinction conventional holistic LLM judges can miss because they evaluate the available context as a single unit and require one or more full-model calls per turn. We propose SAGE (State-Grounded Abstention-Aware Evaluation), which compiles a workflow specification and per-turn state diff into atomic, schema-grounded criteria and routes each through a cascade of symbolic and encoder/NLI verifiers that abstain rather than guess, aggregating criterion verdicts into a turn-level decision with an evidence trace. Its recommended operating point, SAGE-Core, decides 81--91% of criteria with only the compiler, symbolic rules, and on-device encoders--at zero paid LLM cost--while SAGE-LLM adds an optional focused-LLM fallback for open-class criteria. Across four slices spanning MultiWOZ, Schema-Guided Dialogue, and ABCD, no evaluated LLM-as-a-judge baseline--including a state-aware GPT-4.1 judge and cheaper GPT-4.1-mini variants--significantly exceeds SAGE-Core on any slice, even though the GPT-4.1 G-Eval judge costs $4.7--8.0 per 1,000 turns to SAGE-Core's $0. A two-annotator human audit (n=200, $\kappa$=0.94) confirms strong label fidelity on the transcript-visible failure classes--where, excluding the weak-salience IUV class, SAGE-Core is statistically tied with the strongest LLM judge--and honestly scopes ignored-user-value as a state-consistency signal with weak broad-human salience. We analyze construct-validity limits from injected failures and partial symbolic circularity.

Chinese Translation

评估任务导向对话代理不仅需要判断回复是否流畅，还需判断每个回合是否正确推进了基础工作流状态——这一点传统的整体大型语言模型（LLM）评估方法可能会忽视，因为它们将可用上下文视为一个单元，并且每个回合需要一次或多次完整模型调用。我们提出了SAGE（基于状态的弃权意识评估），它将工作流规范和每回合的状态差异编译成原子化的、基于模式的标准，并通过一系列符号和编码器/NLI验证器进行处理，这些验证器选择弃权而非猜测，将标准的裁决汇总为一个回合级别的决策，并附有证据追踪。其推荐的操作点SAGE-Core仅依靠编译器、符号规则和设备上的编码器，便能以零付费LLM成本决定81%至91%的标准，而SAGE-LLM则为开放类标准提供了一个可选的聚焦LLM后备。在涵盖MultiWOZ、Schema-Guided Dialogue和ABCD的四个切片中，没有任何评估的LLM作为评判基线——包括一个状态感知的GPT-4.1评判者和更便宜的GPT-4.1-mini变体——在任何切片上显著超越SAGE-Core，尽管GPT-4.1 G-Eval评判者的成本为每1,000个回合4.7至8.0美元，而SAGE-Core的成本为0美元。两位注释者的人类审计（n=200，$ ext{kappa}$=0.94）确认了转录可见失败类别的强标签一致性——在排除弱显著性IUV类别的情况下，SAGE-Core在统计上与最强的LLM评判者持平——并诚实地将被忽视的用户价值视为状态一致性信号，具有弱的广泛人类显著性。我们分析了由注入失败和部分符号循环性引起的构建有效性限制。

View on arXiv Download PDF AI Translation

cs.AI / 39 / 2609.00441

Conversation Coach: A Voice-enabled AI System that Helps Practice Difficult Workplace Conversations

对话教练：一个帮助练习困难职场对话的语音启用人工智能系统

Wu, Fanyou, Maharjan, Suraj, Yessenalina, Ainur, Chen, Dennis Xu, Srivastava, Rahul, Sengamedu, Srinivasan H.

Abstract

Effective manager-employee communication is critical for retaining high performers and developing underperformers, yet training managers in these skills remains costly. Text-based chatbots offer a scalable approach but cannot provide realistic rehearsal: managers need to practice speaking aloud to build confidence before high-stakes conversations. In this paper, we propose Conversation Coach, a voice-first AI system that enables managers to rehearse difficult workplace conversations in a realistic spoken format. The system addresses three challenges: achieving low-latency interactions with strong language understanding, enabling adaptive conversations through configurable bot personalities that simulate different employee types, and generating personalized feedback on content and policy compliance. We compare an end-to-end speech-to-speech model with a cascaded approach combining automatic speech recognition, a large language model, and text-to-speech synthesis. The end-to-end approach achieves 3$\times$ lower median (P50) latency with native barge-in capability at an estimated 8$\times$ lower cost, while the cascaded approach offers superior reasoning essential for coaching quality. We deployed the cascaded architecture in production, where 40,000+ managers used it over six months, with adoption patterns indicating selective use for difficult conversations.

Chinese Translation

有效的管理者与员工沟通对于留住高绩效员工和发展低绩效员工至关重要，但培训管理者这些技能仍然成本高昂。基于文本的聊天机器人提供了一种可扩展的方法，但无法提供真实的排练：管理者需要大声练习以增强信心，特别是在高风险对话之前。本文提出了对话教练（Conversation Coach），一个以语音为主的人工智能系统，使管理者能够以真实的口语格式排练困难的职场对话。该系统解决了三个挑战：实现低延迟的交互并具备强大的语言理解能力，通过可配置的机器人个性模拟不同员工类型以实现自适应对话，以及生成关于内容和政策合规性的个性化反馈。我们比较了一种端到端的语音到语音模型与一种级联方法，后者结合了自动语音识别、大型语言模型和文本到语音合成。端到端的方法在原生插入能力下实现了3倍更低的中位数（P50）延迟，估计成本降低了8倍，而级联方法则提供了对教练质量至关重要的更优推理。我们在生产中部署了级联架构，超过40,000名管理者在六个月内使用了该系统，采用模式表明其在困难对话中的选择性使用。

View on arXiv Download PDF AI Translation

cs.AI / 40 / 2609.00453

mimeo: Compiling Public Expert Corpora into Agent Skills and Testing What Transfers

mimeo：将公共专家语料库编译为代理技能并测试转移效果

Kassis, Timothy

Abstract

Giving an agent a file about a named expert can supply hard-to-find material, produce a recognizable persona, or change what the agent decides. These are different claims. We test each one. mimeo is an open-source tool that finds a person's public work, checks each extracted quotation against the cached source text, and writes a file an agent can load. Eight logged builds averaged 38 model calls; the check rejects 13.2% of extracted quotations. We tested four expert files with one coding-agent harness. Knowledge access was clearest: mimeo answered all 20 obscure, quotation-heavy questions; no closed-book condition answered more than 10. Keyword search (BM25) over the same pages answered 15-17, a gap this sample cannot resolve. Grounding showed one clear benefit: personas written from model memory misstated a documented position on 1-4 of 20 answers under every grader; the plain agent and mimeo never did. Every persona was easy to spot on short open prompts, and adding task material lowered identification by 18-23 points. mimeo was no more identifiable than a from-memory profile. Judgment transfer remained unresolved because both tests hit their ceiling: every condition found 94-97% of the problems planted in engineering tasks and scored 94-100% on 16 new application scenarios. An AI-judged "sounds like the expert" score changed with the judge: two of four preferred answers based on a model's stereotype, while two found no difference on the same text. That is a caution against relying on a single AI judge. The evidence supports mimeo as a compact, inspectable reference on a person, not as a demonstrated transfer of their judgment. Toolkit and expert profiles: https://github.com/K-Dense-AI/mimeo

Chinese Translation

为代理提供有关某位专家的文件可以提供难以获取的材料，塑造可识别的人物形象，或改变代理的决策。这些是不同的主张。我们对每一个主张进行了测试。mimeo 是一个开源工具，它查找某人的公开工作，检查每个提取的引用与缓存源文本的匹配情况，并生成一个代理可以加载的文件。八个记录的构建平均进行了 38 次模型调用；检查拒绝了 13.2% 的提取引用。我们使用一个编码代理工具测试了四个专家文件。知识获取的效果最为明显：mimeo 回答了所有 20 个晦涩、引用密集的问题；没有任何闭卷条件的回答超过 10 个。对相同页面的关键词搜索（BM25）回答了 15-17 个问题，这一差距本样本无法解决。基础验证显示出一个明显的好处：从模型记忆中编写的人物形象在 20 个答案中有 1-4 个错误陈述了已记录的立场；而普通代理和 mimeo 从未出现这种情况。在简短的开放提示中，每个人物形象都很容易被识别，添加任务材料降低了识别率 18-23 分。mimeo 的可识别性与基于记忆的个人档案相当。判断转移仍未解决，因为两个测试都达到了上限：每个条件发现了 94-97% 在工程任务中植入的问题，并在 16 个新的应用场景中得分 94-100%。由 AI 评判的“听起来像专家”的评分因评判者而异：四个评判者中有两个基于模型的刻板印象偏好某些答案，而两个在相同文本上没有发现差异。这提醒我们不要依赖单一的 AI 评判者。证据支持 mimeo 作为一个紧凑、可检查的个人参考，而不是作为其判断转移的证明。工具包和专家档案：https://github.com/K-Dense-AI/mimeo

View on arXiv Download PDF AI Translation

cs.AI / 41 / 2609.00455

Towards a Belief-Based World Model for LLM Agents

面向基于信念的世界模型的LLM代理

Kumar, Shubham, Kumar, Harshit, Ahuja, Narendra, Jha, Saurabh

Abstract

Large language models (LLMs) are being used as policies for autonomous decision-making and planning in many domains. Despite their strong reasoning capabilities, LLMs struggle with long-horizon tasks, especially under partial observability. World models are a promising way to enhance policy performance, both during training and inference. During inference, agents currently use world models to simulate the consequences of candidate actions before committing to an action, which can improve decision-making. However, we argue that simulation alone is an incomplete interface for decision-making under partial observability: simulation doesn't adequately capture uncertainty about the current state, which agents may need for accurate decision-making. We address this limitation with Belief-Based World Models (BB-WMs), which model and maintain a belief that LLMs can query to access information on what is known and uncertain about the current state. Before developing methods to learn accurate BB-WMs, we first ask a more fundamental question: does exposing a world model's belief directly to an LLM policy improve decision-making? Our results show that giving LLM agents access to world model beliefs improves task performance under partial observability, while remaining complementary to existing simulation-based world models. Code is released at https://github.com/skumar-ml/belief-world-models.

Chinese Translation

大型语言模型（LLMs）正被用于许多领域的自主决策和规划。尽管它们具有强大的推理能力，但LLMs在长时间跨度的任务中表现不佳，尤其是在部分可观察性条件下。世界模型是一种有前景的方法，可以增强策略的性能，无论是在训练还是推理期间。在推理过程中，代理当前使用世界模型来模拟候选行动的后果，然后再决定采取行动，这可以改善决策。然而，我们认为，仅仅依靠模拟并不足以为部分可观察性下的决策提供完整的接口：模拟并未充分捕捉当前状态的不确定性，而代理可能需要这些信息以进行准确的决策。我们通过基于信念的世界模型（BB-WMs）来解决这一局限性，该模型对LLMs进行建模并维护一个信念，LLMs可以查询该信念以获取关于当前状态已知和不确定的信息。在开发学习准确的BB-WMs的方法之前，我们首先提出一个更为根本的问题：直接向LLM策略暴露世界模型的信念是否能改善决策？我们的结果表明，允许LLM代理访问世界模型信念可以在部分可观察性下提高任务性能，同时与现有的基于模拟的世界模型互为补充。代码已发布在 https://github.com/skumar-ml/belief-world-models。

View on arXiv Download PDF AI Translation

cs.AI / 42 / 2609.00479

EGT-KG: Evidence-Grounded Typed KG Retrieval for Practical Scientific QA with Small Language Models

EGT-KG：基于证据的类型化知识图谱检索用于小型语言模型的实用科学问答

Yu, Muran, Gao, Jiechao, Pan, Yuandong, Miao, Barney H., Lesh, Andrew C., Law, Kincho H., Wang, Jie, Lepech, Michael D.

Abstract

For emerging scientific research domains, local Small Language Models (SLMs) are becoming more attractive, as they offer stronger privacy control and more stable deployment pipelines than Large Language Models. However, in practice, scientific question-answering on SLMs often operates under inevitable constraints: small literature collections, fragmented evidence, limited context window and reasoning abilities. We propose the Evidence-Grounded Typed Knowledge Graph (EGT-KG), a retrieval framework to improve information retrieval with local SLMs. We assessed three question-answering settings: a vanilla Retrieval-Augmented Generation (RAG) workflow and two EGT-KG workflows: an automatically generated relation schema (AS) and an expert-defined relation schema (ES). Our experiments were evaluated with a six-dimensional evaluation framework (S3CRF: Soundness, Correctness, Completeness, Conciseness, Relevance, Fluency) on a Biopolymer-bound Soil Composite literature benchmark, showing that EGT-KG outperforms the vanilla RAG method in most settings, with the best improvement from llama3:8b: a Final Score of 70.37 (+14.67%) and 68.82 (+12.14%) by AS/ES EGT-KG variants.

Chinese Translation

对于新兴的科学研究领域，本地小型语言模型（SLMs）变得越来越具有吸引力，因为它们提供了比大型语言模型更强的隐私控制和更稳定的部署流程。然而，在实际应用中，SLMs进行科学问答时往往面临不可避免的限制：文献集合小、证据碎片化、上下文窗口有限以及推理能力不足。我们提出了基于证据的类型化知识图谱（EGT-KG），这是一个检索框架，旨在改善本地SLMs的信息检索。我们评估了三种问答设置：一种普通的检索增强生成（RAG）工作流程和两种EGT-KG工作流程：自动生成的关系模式（AS）和专家定义的关系模式（ES）。我们的实验采用六维评估框架（S3CRF：合理性、正确性、完整性、简洁性、相关性、流畅性）在生物聚合物绑定土壤复合材料文献基准上进行评估，结果表明，EGT-KG在大多数设置中优于普通的RAG方法，其中来自llama3:8b的最佳改进为：AS/ES EGT-KG变体的最终得分为70.37（+14.67%）和68.82（+12.14%）。

View on arXiv Download PDF AI Translation

cs.AI / 43 / 2609.00492

The Privacy-Hallucination Tradeoff in Differentially Private Language Models

差分隐私语言模型中的隐私-幻觉权衡

Ramesh, Krithika, Pillutla, Krishna, Pruthi, Danish, Field, Anjalie

Abstract

Both privacy and factual accuracy are paramount in high-stakes domains like healthcare. Concerningly, we uncover and investigate a privacy-hallucination tradeoff in differentially private (DP) language models. First, we empirically show that models pre-trained or fine-tuned with DP tend to produce more hallucinations than non-DP counterparts, with increased severity as the privacy budget grows stricter. Second, we investigate model properties driving this tradeoff, demonstrating that DP mechanisms flatten output distributions, potentially redistributing probability mass toward factually incorrect alternatives. Third, through experiments where we control fact frequency in training data, we characterize how information frequency can reduce hallucination risks in DP models. Overall, our findings underscore the need for more nuanced privacy-preserving interventions that offer rigorous privacy guarantees without compromising factual accuracy.

Chinese Translation

在医疗等高风险领域，隐私和事实准确性至关重要。令人担忧的是，我们发现并研究了差分隐私（DP）语言模型中的隐私-幻觉权衡。首先，我们通过实证研究表明，经过DP预训练或微调的模型相比于非DP模型更容易产生幻觉，且随着隐私预算的严格程度增加，幻觉的严重性也随之加剧。其次，我们探讨了驱动这一权衡的模型特性，证明DP机制会扁平化输出分布，可能会将概率质量重新分配到事实不正确的替代选项上。第三，通过控制训练数据中的事实频率的实验，我们描述了信息频率如何降低DP模型中的幻觉风险。总体而言，我们的研究结果强调了需要更细致的隐私保护干预措施，以提供严格的隐私保障而不损害事实准确性。

View on arXiv Download PDF AI Translation

cs.AI / 44 / 2609.00498

Validity-Aware Jailbreak Evaluation for Large Language Models

面向有效性的越狱评估方法用于大型语言模型

Wu, Qilong, Wadhwa, Sahil, Mohanty, Pranab, Iyengar, Giri, Chandrasekaran, Varun

Abstract

Jailbreak robustness has become central to large language model (LLM) safety evaluation, yet prevailing methodologies rely primarily on refusal behavior, semantic resemblance, and intent-matching heuristics that emphasize linguistic plausibility rather than correctness. We identify a key limitation in existing evaluations: many jailbreak intents depend on instructional validity rather than epistemic factuality, allowing realistic-looking responses to be labeled successful despite being factually or procedurally incorrect. To address this gap, we propose Sequential Epistemic and Action-Level Validation (SEAV), a verification-centric jailbreak evaluation framework that decomposes responses into ordered steps and evaluates both validity and correctness. SEAV combines LLM-as-a-judge mechanisms for semantic interpretation with retrieval-grounded verification using external knowledge sources, assessing whether generated content is factually correct, structurally consistent, and operationally capable of advancing harmful objectives. Empirically, SEAV cuts the false-positive rate on SD-A (a curated strategic-dishonesty diagnostic) by 14.9\,pp vs. the strongest baseline, and reclassifies 22.1\%--51.0\% of sampled prior-labeled successes as invalid across three of four public benchmarks. Together, these results show that enforcing correctness substantially reshapes measured robustness: many previously labeled jailbreak successes are reclassified as invalid, and results are stable across the tested search backends and evaluator models. Code and data are available at https://github.com/Ardor-Wu/SEAV.

Chinese Translation

越狱鲁棒性已成为大型语言模型（LLM）安全评估的核心，但现有的方法主要依赖于拒绝行为、语义相似性和意图匹配启发式，这些方法强调语言的合理性而非正确性。我们识别出现有评估中的一个关键局限性：许多越狱意图依赖于指令有效性而非认知事实性，这使得看似合理的响应被标记为成功，尽管它们在事实或程序上是错误的。为了填补这一空白，我们提出了顺序认知和行动级验证（Sequential Epistemic and Action-Level Validation，SEAV），这是一个以验证为中心的越狱评估框架，它将响应分解为有序步骤，并评估有效性和正确性。SEAV结合了大型语言模型作为评判者的语义解释机制与基于检索的外部知识源验证，评估生成内容是否在事实上正确、结构上连贯，并在操作上能够推进有害目标。实证结果表明，SEAV在SD-A（一个策划的战略不诚实诊断）上的假阳性率比最强基线降低了14.9个百分点，并在四个公共基准中的三个上将22.1%至51.0%的先前标记为成功的样本重新分类为无效。这些结果表明，强制执行正确性显著重塑了测量的鲁棒性：许多先前标记的越狱成功被重新分类为无效，且结果在测试的搜索后端和评估模型中保持稳定。代码和数据可在 https://github.com/Ardor-Wu/SEAV 获取。

View on arXiv Download PDF AI Translation

cs.AI / 45 / 2609.00503

Wave Function Backpropagation with Explicit Temporal-Interval Dynamics

具有显式时间间隔动态的波函数反向传播

Yu, Byunggu, Kim, Justin

Abstract

Conventional neural networks learn predominantly through affine transformations followed by nonlinear activations, while elapsed time is often treated as an auxiliary feature or assumed to be uniformly sampled. This paper introduces Wave Function Backpropagation (WFB), a wave-parameterized learning formulation in which neural responses are represented by learnable amplitude, wavenumber, angular frequency, and phase. The formulation associates an observed state with its temporal interval Delta t through the phase of a differentiable spatiotemporal wave. We derive standard WFB gradients and a spatial-curvature correction based on the Laplacian of the wave response. WFB is instantiated in a deliberately feed-forward trajectory predictor to provide a controlled proof of concept; sequence learning is outside the scope of the present evaluation. With motion features, STD-WFB using real intervals reduces average displacement error (ADE) by 20.4% relative to the original FFN baseline. In a new position-only evaluation that removes temporal leakage through precomputed velocity and acceleration, real-interval WFB reduces ADE by 10.4% relative to the original FFN and remains competitive with parameter-matched ReLU controls, obtaining 2.1% lower mean ADE than the matched FFN with explicit Delta t. Shuffled-interval WFB attains the lowest mean ADE, indicating that the present evidence supports the effectiveness of the wave representation but does not attribute the gain to interval alignment. These results establish WFB as a viable structured feed-forward learning formulation and define a clear basis for subsequent architectural studies.

Chinese Translation

传统神经网络主要通过仿射变换后跟随非线性激活来学习，而经过的时间通常被视为辅助特征或假定为均匀采样。本文介绍了波函数反向传播（Wave Function Backpropagation, WFB），这是一种波参数化的学习形式，其中神经响应由可学习的幅度、波数、角频率和相位表示。该形式通过可微分的时空波的相位将观察到的状态与其时间间隔Δt关联起来。我们推导了标准WFB梯度和基于波响应拉普拉斯算子的空间曲率修正。WFB在一个故意设计的前馈轨迹预测器中实例化，以提供一个受控的概念验证；序列学习超出了本次评估的范围。使用运动特征的STD-WFB在真实时间间隔下相较于原始前馈网络（FFN）基线减少了20.4%的平均位移误差（ADE）。在一个新的仅位置评估中，通过预计算的速度和加速度消除了时间泄漏，真实时间间隔的WFB相较于原始FFN减少了10.4%的ADE，并且在与参数匹配的ReLU控制下保持竞争力，获得了比具有显式Δt的匹配FFN低2.1%的平均ADE。随机时间间隔的WFB达到了最低的平均ADE，表明目前的证据支持波表示的有效性，但并未将增益归因于时间间隔的对齐。这些结果确立了WFB作为一种可行的结构化前馈学习形式，并为后续的架构研究奠定了明确的基础。

View on arXiv Download PDF AI Translation

cs.AI / 46 / 2609.00508

CoVer: Conflict-Aware Claim Verification

CoVer：冲突感知的主张验证

Zhang, Shuning, Shi, Dai, Chu, Bohao, Wang, Hui, Chuai, Yuwei, Wang, Yifan, Chen, Jingruo, Li, Simin, Yi, Xin, Li, Hewu

Abstract

Social media fact-checking has long been challenged by evidence-level and aggregation-level conflicts, where erroneous evidence mimics authoritative news sources. To capture this challenge and support conflict verification tasks, we present ContraNote, a large-scale real-world dataset curated from X's Community Notes system. It includes 33,686 posts for evaluating evidence-level conflict resolution, and 54,474 instances for evaluating aggregation-level prioritization. Additionally, we propose CoVer, a factual adjudication framework with three-stage pipelines: evidence schema normalization, factual consensus and support verification. This prioritizes evidence over noise to prevent it from compromising the final verdict. Technical evaluations show that CoVer achieves strong performance compared with state-of-the-art baselines across ContraNote (86.0% Acc., 68.0% mac. F1, 64.5 bal. Acc. on Conflict; and 88.5% Acc., 88.5 mac. F1 and 89.2 bal. Acc. on Prioritization), CONFACT-HumC (88.4% Acc.) and CONFACT-ModC (89.4% Acc.).

Chinese Translation

社交媒体的事实核查长期以来受到证据层面和聚合层面冲突的挑战，其中错误的证据模仿权威新闻来源。为了应对这一挑战并支持冲突验证任务，我们提出了ContraNote，这是一个从X的社区笔记系统中整理的大规模真实世界数据集。该数据集包含33,686条帖子用于评估证据层面的冲突解决，以及54,474个实例用于评估聚合层面的优先级。此外，我们提出了CoVer，一个具有三阶段流程的事实裁决框架：证据模式标准化、事实共识和支持验证。该框架优先考虑证据而非噪声，以防止其影响最终裁决。技术评估表明，CoVer在ContraNote（冲突任务的准确率为86.0%，宏观F1为68.0%，平衡准确率为64.5%；优先级任务的准确率为88.5%，宏观F1为88.5%，平衡准确率为89.2%）、CONFACT-HumC（准确率为88.4%）和CONFACT-ModC（准确率为89.4%）等最先进基线中表现出色。

View on arXiv Download PDF AI Translation

cs.AI / 47 / 2609.00510

When the Algorithm Becomes the Brand Crisis: A Sociotechnical Theory of Distributed Responsibility and Accountable Transparency

当算法成为品牌危机：分布式责任与可问责透明度的社会技术理论

Torkestani, Mohammad Saleh, Mansouri, Taha

Abstract

Artificial intelligence systems increasingly enact market-facing promises through chatbots, recommendation systems, automated decisions, and generative interfaces. Their failures, misuse, and misrepresentation raise a question that conventional brand-crisis models do not fully specify: how do stakeholders assign responsibility when technical causation, customer-facing control, and governance duties are distributed across an AI system, developer, deployer, vendor, and user? This conceptual paper develops a sociotechnical process theory from a structured, federated scoping synthesis of verified academic and primary sources. It distinguishes an AI/algorithmic incident from an AI-related organisational crisis and, in turn, from an AI-related organisational scandal. The framework proposes that incident configuration shapes actor-specific attribution; attribution informs capability, integrity, fairness, and relationship appraisals; and public moralisation may, but need not, escalate an incident into scandal. The theory offers a reconciliation of findings that algorithm involvement can buffer negative brand reactions in some settings while robot and chatbot failures can redirect responsibility to an associated firm in others. It introduces accountable transparency as a proposed response configuration that combines timely notice, an intelligible account, role-responsibility acknowledgement, remedy, evidence of correction, and recourse. The evidence supports conditional, proximal inferences about blame, trust, satisfaction, firm evaluation, and communication credibility more strongly than claims about durable reputation, brand equity, or market performance.

Chinese Translation

人工智能系统越来越多地通过聊天机器人、推荐系统、自动决策和生成接口实现面向市场的承诺。它们的失败、误用和误表述提出了一个传统品牌危机模型未能充分说明的问题：当技术因果关系、面向客户的控制和治理责任分布在人工智能系统、开发者、部署者、供应商和用户之间时，利益相关者如何分配责任？本文通过对经过验证的学术和主要来源的结构化、联合范围综合，发展了一种社会技术过程理论。它将人工智能/算法事件与人工智能相关的组织危机区分开来，并进一步与人工智能相关的组织丑闻区分开来。该框架提出事件配置塑造特定参与者的归因；归因影响能力、诚信、公平和关系评估；而公众道德化可能会（但不一定）将事件升级为丑闻。该理论提供了一种调和发现的方式，即算法参与在某些环境中可以缓冲负面品牌反应，而机器人和聊天机器人的失败则可以将责任转移到相关公司。它引入了可问责透明度作为一种建议的响应配置，结合了及时通知、可理解的说明、角色责任承认、补救、纠正证据和追索。证据更强烈地支持关于责任、信任、满意度、公司评估和沟通可信度的条件性、近因推论，而不是关于持久声誉、品牌价值或市场表现的主张。

View on arXiv Download PDF AI Translation

cs.AI / 48 / 2609.00513

ISO-RAG: Isoperimetric Noise Control for Retrieval-Augmented Generation

ISO-RAG：用于检索增强生成的等周噪声控制

Zhang, Siyuan, Wang, Hanchen, Wen, Dong, Zhang, Ying, Zhang, Wenjie

Abstract

Retrieval-Augmented Generation (RAG) mitigates large language models (LLMs) hallucinations, yet conventional dense retrieval struggles with the complex reasoning paths of multi-hop question answering (QA). Graph-based RAG captures multi-step relationships but suffers from severe semantic drift and high online latency due to noisy global graph traversals. Thus, we propose ISO-RAG (ISOperimetric Retrieval-Augmented Generation), a geometry-aware RAG framework. By projecting the underlying knowledge graph into a hyperbolic Poincare ball to precompute node-wise isoperimetric profiles, ISO-RAG prunes spurious edges during retrieval, restricting the search space to a strictly localized subgraph. This topological purification regulates Personalized PageRank (PPR) diffusion driving the retrieval process, ensuring exact and low-latency convergence. Experiments on multi-hop QA benchmarks demonstrate that ISO-RAG outperforms state-of-the-art baselines by average absolute gains of 10.0% in retrieval recall and 4.3% in downstream exact match, achieving a superior accuracy-efficiency trade-off by fundamentally eliminating the latency bottleneck of global traversals. Our source code is available at https://github.com/ZaiizaiZHANG/ISO-RAG.

Chinese Translation

检索增强生成（Retrieval-Augmented Generation, RAG）缓解了大型语言模型（Large Language Models, LLMs）的幻觉问题，但传统的密集检索在多跳问答（multi-hop question answering, QA）的复杂推理路径上表现不佳。基于图的RAG捕捉多步关系，但由于嘈杂的全局图遍历，严重的语义漂移和高在线延迟问题使其受到限制。因此，我们提出了ISO-RAG（等周检索增强生成），一个具有几何意识的RAG框架。通过将基础知识图投影到超球面Poincaré球中，以预计算节点级的等周轮廓，ISO-RAG在检索过程中修剪虚假边缘，将搜索空间限制在严格局部的子图中。这种拓扑净化调节了驱动检索过程的个性化PageRank（Personalized PageRank, PPR）扩散，确保了精确且低延迟的收敛。在多跳QA基准上的实验表明，ISO-RAG在检索召回率上平均提高了10.0%，在下游精确匹配上提高了4.3%，通过从根本上消除全局遍历的延迟瓶颈，实现了优越的准确性与效率的权衡。我们的源代码可在https://github.com/ZaiizaiZHANG/ISO-RAG获取。

View on arXiv Download PDF AI Translation

cs.AI / 49 / 2609.00543

Feedback-Assisted Trust Propagation over Document Relation Graphs for Retrieval-Augmented Generation

基于反馈辅助的文档关系图信任传播用于检索增强生成

Li, Zhuoheng, Chen, Ying

Abstract

Retrieval-augmented generation (RAG) systems rely on external corpora that may contain outdated, contradictory, noisy, or unreliable documents, introducing reliability risks. Prior work has leveraged document relations to improve the answer reliability of RAG. To propagate reliability signals beyond directly compared document pairs, we propose TrustPropRAG, which structures document relations as a graph and estimates document reliability through multi-hop propagation across the graph. TrustPropRAG anchors this propagation with a limited set of human feedback on document reliability, extending these costly-to-collect feedback-based reliability signals across the whole corpus. Specifically, based on the constructed document relation graph, TrustPropRAG estimates a trust score for each document by formulating and solving an optimization problem that jointly captures pairwise document relations and user feedback. These scores are then used to improve the selection of reliable documents and support trust-aware answer generation. Evaluation results show that TrustPropRAG improves both retrieval quality and exact match over baselines, and remains robust under sparse and noisy feedback.

Chinese Translation

检索增强生成（RAG）系统依赖于可能包含过时、矛盾、噪声或不可靠文档的外部语料库，这引入了可靠性风险。之前的研究利用文档关系来提高RAG的答案可靠性。为了在直接比较的文档对之外传播可靠性信号，我们提出了TrustPropRAG，它将文档关系结构化为图，并通过图中的多跳传播来估计文档的可靠性。TrustPropRAG以有限的人类反馈作为文档可靠性的锚点，将这些成本高昂的基于反馈的可靠性信号扩展到整个语料库。具体而言，基于构建的文档关系图，TrustPropRAG通过制定和求解一个优化问题来估计每个文档的信任分数，该问题共同捕捉成对文档关系和用户反馈。这些分数随后用于改善可靠文档的选择，并支持信任感知的答案生成。评估结果表明，TrustPropRAG在检索质量和精确匹配方面均优于基线，并在稀疏和噪声反馈下保持稳健性。

View on arXiv Download PDF AI Translation

cs.AI / 50 / 2609.00570

VoiceLongMemEval: Do Assistants Remember How You Sounded?

VoiceLongMemEval：助手是否记得你的声音？

Pahwa, Ramit, Priye, Parivesh, Beedu, Apoorva

Abstract

With the growing scale of multi-agent architectures and large language models, deployed AI assistants are increasingly tasked with reasoning over long, continuous, multi-session conversation histories. Current benchmarks evaluate this dialogue history as information retrieval over long horizon, temporal reasoning, or knowledge updates, while crucially ignoring the fundamental dynamics of human-agent interaction, i.e. how they said it. To address this gap, we present VoiceLongMemEval (VLME) benchmark, where every answer depends on paralinguistic metadata (emotion labels, prosody descriptors, and voice events) attached to conversational turns, which is otherwise unrecoverable from the words alone. Every item passes a three-stage adversarial gate, ensuring that a strong language model fails when given only the transcript. Evaluating leading frontier and open-weight models reveals a pervasive affect gap; providing text-track paralinguistic metadata yields a 0.09 to 0.38 accuracy boost (0.61 to 0.69 when prompted with evidence hints), while standard ASR pipelines systematically discard this signal. Additionally, audio-native models successfully extract these cues directly from speech (0.354 to 0.412 vs. 0.325 blind). Code and dataset will be made available upon acceptance.

Chinese Translation

随着多智能体架构和大型语言模型规模的不断扩大，部署的人工智能助手越来越多地被要求对长时间、连续的多会话对话历史进行推理。目前的基准测试将这种对话历史评估为长期信息检索、时间推理或知识更新，而关键地忽视了人机交互的基本动态，即他们是如何表达的。为了解决这一问题，我们提出了VoiceLongMemEval (VLME) 基准，其中每个回答都依赖于附加在对话轮次上的副语言元数据（情感标签、韵律描述符和声音事件），这些信息仅从文字中无法恢复。每个项目都经过三阶段的对抗性筛选，确保当仅提供文本记录时，强大的语言模型会失败。对领先的前沿模型和开放权重模型的评估揭示了普遍的情感差距；提供文本跟踪的副语言元数据可提高0.09到0.38的准确率（在提示证据线索时为0.61到0.69），而标准的自动语音识别（ASR）流程系统性地丢弃了这一信号。此外，音频原生模型成功地直接从语音中提取这些线索（0.354到0.412对比0.325盲测）。代码和数据集将在接受后提供。

View on arXiv Download PDF AI Translation

cs.AI / 51 / 2609.00575

Residual Sparsification via Output Importance for Compressing Mixture-of-Experts LLMs

通过输出重要性进行残差稀疏化以压缩混合专家大语言模型

Jung, Seungwoo, Kwon, Dohyeok, Cha, Seungmin, Lee, Junseok, Yoo, Yeonho, Yoo, Chuck, Yang, Gyeongsik

Abstract

Mixture-of-experts (MoE) architectures scale large language models efficiently, but they demand massive GPU memory. To cope with such demand, models are commonly compressed to reduce their memory footprint. Residual sparsification is a representative compression technique that decomposes each projection matrix of an expert into a shared base matrix and per-expert residual matrix, and then compresses the residuals. Existing sparsification methods compress each residual matrix independently by minimizing its compression error, thereby minimizing the error of each projection matrix. However, our analysis shows that this objective is misaligned with preserving model accuracy after compression. In an expert, the final output is produced through computations coupled across multiple projections and hidden representations. Therefore, even small errors in individual matrices can propagate through hidden representations and projection interactions, leading to large expert output errors and accuracy degradation. To address this misalignment, we propose PARSER, a new residual sparsification method that shifts the compression objective from minimizing isolated matrix errors to preserving the expert output error. PARSER achieves this by introducing output importance, which measures the actual contribution to the expert output error. Our experiments show that, compared with existing methods, PARSER narrows the accuracy gap to the uncompressed model by 1.41$\times$ on Qwen and 1.44$\times$ on DeepSeek, while achieving the same peak memory reduction.

Chinese Translation

混合专家（MoE）架构有效地扩展了大型语言模型，但它们需要大量的GPU内存。为了应对这种需求，模型通常会被压缩以减少其内存占用。残差稀疏化是一种典型的压缩技术，它将专家的每个投影矩阵分解为一个共享的基础矩阵和每个专家的残差矩阵，然后对残差进行压缩。现有的稀疏化方法通过最小化每个残差矩阵的压缩误差来独立压缩每个残差矩阵，从而最小化每个投影矩阵的误差。然而，我们的分析表明，这一目标与压缩后保持模型准确性的目标不一致。在一个专家中，最终输出是通过多个投影和隐藏表示之间的计算耦合产生的。因此，即使是单个矩阵中的小错误也可能通过隐藏表示和投影交互传播，导致专家输出错误和准确性下降。为了解决这一不一致，我们提出了PARSER，一种新的残差稀疏化方法，它将压缩目标从最小化孤立矩阵误差转变为保持专家输出误差。PARSER通过引入输出重要性来实现这一目标，输出重要性衡量了对专家输出误差的实际贡献。我们的实验表明，与现有方法相比，PARSER在Qwen上将与未压缩模型的准确性差距缩小了1.41倍，在DeepSeek上缩小了1.44倍，同时实现了相同的峰值内存减少。

View on arXiv Download PDF AI Translation

cs.AI / 52 / 2609.00576

Consistency Without Alignment: Item-Sensitive Language Models Indistinguishable From Random

无对齐的一致性：与随机无异的项目敏感语言模型

Huynh, Cris

Abstract

Item-sensitivity, defined as whether a model's choice depends on the specific input rather than on its own output prior, is widely reported as evidence of task competence. We show this evidence is necessary but not sufficient using a forced-choice signalling task abstracted from the board game Deception: Murder in Hong Kong. In this environment, the reference points against which a coordinate should be judged (a fit-maximising strategy, a posterior-maximising strategy, and uniform random selection) are all computable in closed form. Across seven language models, two model families, a post-training ablation, and three independent scoring rules, every one of 21 model-by-rule cells is reliably item-sensitive. Yet 8 of those 21 cells are not statistically distinguishable from a chooser that ignores the item and selects at random, and 5 score worse than random at describing the target. Item-sensitivity and distance from random correlate at only r = 0.30. We call this consistency without alignment and argue it generalises to any evaluation that relies on item-sensitivity, permutation consistency, or self-consistency without an independent reference for the measured quantity. We further find that a literal-similarity baseline with no pragmatics outperforms most tested language models, that adding a pragmatic layer over two baseline similarity sources moves choosers toward random rather than toward the Bayesian reference, and that a standard labelled multiple-choice format carries no measurable content signal here. All results represent the model side of a pre-registered instrument; a matched human condition is designed and piloted but not yet collected.

Chinese Translation

项目敏感性被定义为模型的选择是否依赖于特定输入而非其自身的输出先验，这被广泛报道为任务能力的证据。我们通过一个抽象自桌游《香港谋杀：欺骗》的强制选择信号任务展示了这一证据是必要但不充分的。在该环境中，用于判断坐标的参考点（一个最优拟合策略、一个后验最大化策略和均匀随机选择）均可用封闭形式计算。在七个语言模型、两个模型家族、一个后训练消融实验和三个独立评分规则中，21个模型与规则的组合单元中每一个都可靠地表现出项目敏感性。然而，这21个单元中有8个在统计上与忽略项目并随机选择的选择者无显著区别，且有5个在描述目标时的得分低于随机选择。项目敏感性与随机选择的距离仅相关性为 r = 0.30。我们称之为无对齐的一致性，并认为它可以推广到任何依赖于项目敏感性、置换一致性或自一致性而没有独立参考的评估。我们进一步发现，缺乏语用学的字面相似性基线优于大多数测试的语言模型，且在两个基线相似性来源上添加语用层使选择者更趋向于随机而非贝叶斯参考，并且标准标记的多项选择格式在此没有可测量的内容信号。所有结果代表了一个预注册工具的模型侧；一个匹配的人类条件已设计并试点，但尚未收集。

View on arXiv Download PDF AI Translation

cs.AI / 53 / 2609.00578

Same Request, Different Boundary: Evaluating Cybersecurity Assistance across Conversational Contexts

相同请求，不同边界：评估对话上下文中的网络安全援助

Yang, Rui, Hong, Yang, Xu, Yichao, Liu, Zhengyu, Li, Ziyang, Cao, Yinzhi

Abstract

Large Language Models (LLMs) can solve complex problems, but their misuse in high-risk domains can lead to severe consequences. Model providers therefore restrict assistance for potentially harmful requests. Refusing all cybersecurity requests would therefore harm legitimate users. Providers need a mechanism to block malicious use without denying legitimate assistance to defenders. Existing cybersecurity-specific datasets evaluate this mechanism, but none considers the conversational context of a request. We introduce 3R-Bench (Refusal, Repetition, and Revision), a benchmark of 150 real-world cybersecurity requests augmented with two adversarial conversational settings, and evaluate eight LLMs on it. Prior assistant behavior strongly changes responses to an unchanged request: among 376 available pairs from a 400-pair panel, compliance rises from 62.0% after refused history to 85.1% after accepted history. The opposite pattern appears under dialogue decomposition. In comparison, compliance falls from 501/800 direct responses to 172/800 after dialogue; among 738 pairs returning model-authored text in both conditions, the decrease is 45.1 points. Failure feedback recovers only a small fraction of this loss.

Chinese Translation

大型语言模型（LLMs）能够解决复杂问题，但在高风险领域的误用可能导致严重后果。因此，模型提供者限制对潜在有害请求的援助。然而，拒绝所有网络安全请求将对合法用户造成伤害。提供者需要一种机制，以阻止恶意使用，同时不拒绝对防御者的合法援助。现有的网络安全特定数据集评估了这一机制，但没有考虑请求的对话上下文。我们引入了3R-Bench（拒绝、重复和修订），这是一个包含150个真实世界网络安全请求的基准，增强了两个对抗性对话设置，并在此基础上评估了八个LLM。先前的助手行为显著改变了对未变请求的响应：在400对样本中，376对可用样本中，拒绝历史后的合规率从62.0%上升到接受历史后的85.1%。在对话分解下则出现相反的模式。相比之下，合规率从501/800的直接响应下降到对话后的172/800；在738对在两种条件下返回模型创作文本的样本中，下降幅度为45.1个百分点。失败反馈仅恢复了这一损失的一小部分。

View on arXiv Download PDF AI Translation

cs.AI / 54 / 2609.00584

Socrates went Nuclear: Comparing Interaction Strategies for AI systems in a Learning Context using Brain Sensing

苏格拉底走向核能：在学习环境中使用脑部感应比较人工智能系统的互动策略

Deffarges, Alexandre Clin, Kosmyna, Nataliya, Maes, Pattie

Abstract

Does unrestricted AI access bypass the cognitive effort required for learning, or does it streamline knowledge acquisition? This paper reports on a study where we compare three designs for user-AI interaction in a learning context: (1) an unrestricted conversational bot like ChatGPT, (2) a pedagogically constrained bot that guides through hints without giving final answers, which we refer to as the Socratic mode; and (3) a non-conversational adaptive tutoring system that adjusts difficulty in real-time based on the user's cognitive engagement derived from the brain signals. Fifty study participants were tasked with learning about nuclear safety protocols, a domain chosen for its zero-prior knowledge baseline. The participants progressed through an instructional video, a pre-test, an AI-driven assessment phase, which varied in the three conditions, and an immediate post-test. The nature of the questions centered primarily on factual knowledge acquisition, but it still required participants to have a global understanding of the concepts in order to answer the questions correctly. A Muse headband was used to derive the cognitive engagement of all users in all conditions. The unrestricted chatbot produced higher learning gains (delta) than both constrained modes (p < .03, d > 0.80), while the adaptive condition generated significantly higher EEG engagement (p = .018). The cluster analysis of chatbot usage and discussion patterns by users showed that most participants in the unrestricted-mode adopted a direct answer-retrieval strategy, while participants in the Socratic-mode initially attempted to reason through the hints before progressively disengaging. Consequently, this also suggests that the success of the unrestricted AI is not an evidence of deeper learning, but rather a result of the immediate post-test evaluation after the training phase.

Chinese Translation

无限制的人工智能访问是否绕过了学习所需的认知努力，还是简化了知识获取？本文报告了一项研究，我们比较了三种用户与人工智能在学习环境中的互动设计：(1) 像 ChatGPT 这样的无限制对话机器人，(2) 一种教育约束的机器人，通过提示引导而不提供最终答案，我们称之为苏格拉底模式；以及 (3) 一种非对话的自适应辅导系统，基于用户的脑信号推导出的认知参与度实时调整难度。五十名参与者被要求学习核安全协议，这是一个选择为零先验知识基线的领域。参与者经历了一段教学视频、一个前测、一个基于人工智能的评估阶段（在三种条件下有所不同）以及一个即时后测。问题的性质主要集中在事实知识的获取上，但仍然要求参与者对概念有整体理解，以便正确回答问题。使用 Muse 头带来推导所有用户在所有条件下的认知参与度。无限制的聊天机器人产生的学习增益（delta）高于两种约束模式（p < .03，d > 0.80），而自适应条件产生的脑电图（EEG）参与度显著更高（p = .018）。对聊天机器人使用和用户讨论模式的聚类分析显示，大多数无限制模式的参与者采用了直接答案检索策略，而苏格拉底模式的参与者最初尝试通过提示进行推理，然后逐渐 disengage。因此，这也表明无限制人工智能的成功并不是更深层学习的证据，而是培训阶段后即时后测评估的结果。

View on arXiv Download PDF AI Translation

cs.AI / 55 / 2609.00621

Control-Data Flow Separation: Stable Prompt Optimization in Multi-Agent LLMs

控制-数据流分离：多智能体大语言模型中的稳定提示优化

Zhang, Wentao, Murtaza, Syed Shariyar, Bhatti, Junaid Ahmad, Soni, Utkarsh, Nie, Yifan, Wen, Eugene, Deng, Yuntian

Abstract

Prompt optimization can improve multi-agent LLM systems, but the prompts being optimized often serve two entangled roles: generating task-relevant content and specifying execution-critical protocols, such as message routing, output formatting, and termination signals, on which the underlying code relies. As a result, a prompt edit intended to improve content generation can inadvertently corrupt the protocol and cause the entire agent pipeline to fail. Our key observation is that these two roles have different representations: execution protocols are typically structured, while task-relevant content is usually expressed in unstructured language. Based on this, we propose control-data flow separation, where execution-critical control is represented as typed, validated program objects, while task-relevant language remains the optimizable data flow for agent communication. This design allows optimizers to improve multi-agent behavior without exposing the routing or formatting interface to prompt drift. Across synthetic reasoning, collaborative review generation, and insurance rating workflows, our framework empirically achieves 100% eventual protocol validity while consistently improving task performance.

Chinese Translation

提示优化可以提升多智能体大语言模型系统的性能，但被优化的提示通常承担着两个交织的角色：生成与任务相关的内容和指定执行关键的协议，例如消息路由、输出格式和终止信号，这些都是底层代码所依赖的。因此，旨在改善内容生成的提示编辑可能会无意中破坏协议，导致整个智能体管道的失败。我们关键的观察是，这两个角色具有不同的表现形式：执行协议通常是结构化的，而与任务相关的内容通常以非结构化语言表达。基于此，我们提出了控制-数据流分离的概念，其中执行关键的控制被表示为类型化、经过验证的程序对象，而与任务相关的语言则保持为可优化的数据流，以便于智能体之间的通信。这一设计使得优化器能够在不暴露路由或格式接口给提示漂移的情况下改善多智能体行为。在合成推理、协作审查生成和保险评级工作流中，我们的框架在经验上实现了100%的最终协议有效性，同时持续提升任务性能。

View on arXiv Download PDF AI Translation

cs.AI / 56 / 2609.00643

REVISE: Validity-Guided Recovery for Online Revisions in Agent Workflows

REVISE：基于有效性的在线修订恢复在代理工作流中的应用

Qi, Ruoling, Wu, Xuaner, Liu, Penghang, Chen, Jian, Liu, Yirui

Abstract

Agent revisions expose a fundamental correctness--efficiency trade-off during concurrent execution. Discarding ongoing work preserves latest-version correctness but wastes progress that may remain valid, whereas reusing prior work preserves efficiency but risks propagating stale state into outputs and tool effects. Existing recovery strategies resolve this trade-off in an imbalanced way with coarse-grained policies: they either favor efficiency by allowing potentially stale work to continue, or favor correctness by restarting the workflow or recomputing a linear suffix from the earliest conflict, thereby discarding unaffected progress. We present \textsc{Revise}, a validity-guided runtime for fine-grained recovery in structured agent workflows. When a revision arrives, \textsc{Revise} first intersects its delta with recorded data and control dependencies and propagates the resulting impact through the partially executed DAG to identify affected work. It then stops invalid work, preserves validity-established progress beyond the earliest conflict, and recomputes only the affected region. Incomplete provenance conservatively expands recovery, while reused results are revalidated before commit. Analysis of real coding-agent traces show online recovery opportunities: 118 sessions retain observable work before a queued later message is delivered; across 167 overlapping assistant responses, enqueue-to-completion overlap reaches 56.55~s at p95. Across 300 challenging revision/commit executions, \textsc{Revise} matches a latest-version oracle with no stale outputs or effects. On unmodified LangGraph and LLMCompiler applications using Qwen3-14B, it reduces model calls by 40.6--56.0\% relative to full restart and by 31.3--43.6\% relative to suffix recomputation. Under serving pressure, it further reduces revision-to-correct-completion tokens by 13.26\% and improves SLO goodput by 3.07--5.43\%.

Chinese Translation

代理修订在并发执行过程中暴露了一个基本的正确性与效率的权衡。丢弃正在进行的工作可以保持最新版本的正确性，但会浪费可能仍然有效的进展；而重用先前的工作则保持了效率，但有可能将过时的状态传播到输出和工具效果中。现有的恢复策略以不平衡的方式解决了这一权衡，采用粗粒度的策略：它们要么通过允许潜在的过时工作继续进行来偏向效率，要么通过重启工作流或从最早的冲突重新计算线性后缀来偏向正确性，从而丢弃未受影响的进展。我们提出了 extsc{Revise}，一种用于结构化代理工作流的细粒度恢复的有效性引导运行时。当修订到达时， extsc{Revise} 首先将其增量与记录的数据和控制依赖进行交集，并通过部分执行的有向无环图（DAG）传播结果影响，以识别受影响的工作。然后，它停止无效工作，保留在最早冲突之后建立的有效进展，并仅重新计算受影响的区域。不完整的来源保守地扩展恢复，而重用的结果在提交之前会重新验证。对真实编码代理轨迹的分析显示了在线恢复的机会：118个会话在排队的后续消息送达之前保留了可观察的工作；在167个重叠的助手响应中，排队到完成的重叠达到了56.55秒（p95）。在300次具有挑战性的修订/提交执行中， extsc{Revise} 匹配了最新版本的预言机，没有过时的输出或效果。在使用 Qwen3-14B 的未修改的 LangGraph 和 LLMCompiler 应用中，相较于完全重启，它减少了40.6%至56.0%的模型调用，相较于后缀重新计算则减少了31.3%至43.6%。在服务压力下，它进一步减少了修订到正确完成的令牌数量13.26%，并提高了服务水平目标（SLO）的良好吞吐量3.07%至5.43%。

View on arXiv Download PDF AI Translation

cs.AI / 57 / 2609.00646

DramaChain Bench: An End-to-End Benchmark for Short-Drama Generation

DramaChain Bench：一个端到端的短剧生成基准

Shi, Haoyuan, Chen, Mingtao, Jiang, Shuo, Chen, Ziyan, Sheng, Xuyi, Liu, Yiming, Zhang, Ying, Wang, Miao, Lu, Jianxiang, Lu, Fanyang, Lu, Songyuanyi, Wu, Xiele, Hu, Zhichao, Liu, Yuhong, Xuan, Richeng

Abstract

Commercial short-drama production follows a multi-stage chain: script, storyboard, keyframe imagery, shot-level video, and the finished short drama. Most existing benchmarks evaluate solely the video-generation stage using pre-authored inputs instead of real upstream pipeline outputs. This leaves two critical questions unanswerable: whether each stage adheres to the original script intent (rather than only its immediate input prompt), and whether disparate shots remain coherent after assembly into multi-episode releases. We present DramaChain Bench, the first short-drama benchmark that evaluates every stage of the complete production chain. It is built upon three in-house systems sharing one dimension system, DramaChain Dimensions: five evaluation axes instantiated at every stage, resolving into 63 leaf dimensions. DramaChain Agent is calibrated against commercial short-drama platforms in both workflow and finished short-drama quality, enabling stage-wise fair comparison across models. DramaChain Labeling System has each of the 5,785 items scored independently by three professional annotators, with all defects spatio-temporally localised and selected from a predefined defect list. This process produces 17,488 valid scores and 255,925 traceable attribution records. The human annotations confirm that upstream defects cascade across the pipeline, demonstrating that final episode quality is not governed by video generation alone. DramaChain Agentic Judge then scores every leaf dimension automatically, gathering evidence over multiple agentic rounds before judging against a per-item checklist; it reproduces the model ranking at a mean PLCC of 0.918, enough to admit new models at no annotation cost.

Chinese Translation

商业短剧制作遵循一个多阶段链条：剧本、分镜头脚本、关键帧图像、镜头级视频以及最终的短剧。现有的大多数基准仅评估视频生成阶段，使用预先编写的输入，而不是实际的上游管道输出。这留下了两个关键问题无法回答：每个阶段是否遵循原始剧本的意图（而不仅仅是其直接输入提示），以及在组装成多集发布后，不同镜头是否仍然保持一致性。我们提出了DramaChain Bench，这是第一个评估完整制作链每个阶段的短剧基准。它建立在三个内部系统之上，共享一个维度系统，DramaChain Dimensions：在每个阶段实例化的五个评估轴，最终分解为63个叶维度。DramaChain Agent在工作流程和最终短剧质量上与商业短剧平台进行了校准，使得不同模型之间能够进行阶段性的公平比较。DramaChain Labeling System对5,785个项目进行了独立评分，由三位专业注释员完成，所有缺陷都在时空上进行定位，并从预定义的缺陷列表中选择。这个过程产生了17,488个有效评分和255,925个可追溯的归因记录。人工注释确认上游缺陷在管道中级联，表明最终集的质量不仅仅由视频生成决定。DramaChain Agentic Judge随后自动对每个叶维度进行评分，在多个代理回合中收集证据，然后根据每个项目的检查表进行评判；它以平均PLCC 0.918再现模型排名，足以在没有注释成本的情况下接纳新模型。

View on arXiv Download PDF AI Translation

cs.AI / 58 / 2609.00652

Self-Reports Are Not Verification: Environment-Grounded Auditing of LLM Operators in Evolutionary Search

自我报告并非验证：基于环境的进化搜索中的大型语言模型操作员审计

Pan, Enrong, Zhou, Ryan, Hu, Ting

Abstract

Language model agents increasingly propose actions, observe external feedback, and explain their own behavior. Their confidence and rationales are convenient monitoring signals, but convenience is not verification. We introduce an environment-grounded audit in which every intermediate proposal receives an exact outcome. A language model operates an evolutionary Contexto search whose feedback function assigns every valid guess an exact rank without human annotation. Across 200 runs spanning five configurations and three model families, four reporting configurations produce 12,249 self-reports. We test three assumptions: stated confidence is calibrated, inherited rationales affect later proposals, and fitness-based selection improves report quality. All three fail. Operators overstate top-100 success by factors of 4.8 to 9.3, while calibration and discrimination dissociate across model families. Controlled interventions on 754 inherited rationales bound any measured benefit of the genuine rationale to roughly 250 ranks. Neither fitness-based nor random selection produces a detectable selection differential or parent-to-offspring transmission in report accuracy, despite sharply different search behavior. Agent self-reports should therefore be treated as claims to verify against the environment, not as evidence of their own reliability.

Chinese Translation

语言模型代理越来越多地提出行动、观察外部反馈并解释自身行为。它们的信心和理由是便利的监测信号，但便利并不等同于验证。我们引入了一种基于环境的审计，其中每个中间提案都获得确切的结果。一个语言模型操作一个进化的 Contexto 搜索，其反馈函数为每个有效猜测分配一个确切的排名，而无需人工注释。在跨越五种配置和三个模型家族的200次运行中，四种报告配置产生了12,249个自我报告。我们测试了三个假设：声明的信心是经过校准的，继承的理由影响后续提案，基于适应度的选择提高报告质量。所有三个假设均未成立。操作员高估了前100名的成功率，倍数在4.8到9.3之间，而校准和区分在不同模型家族之间表现出解耦。在754个继承理由上进行的控制干预将任何真实理由的测量效益限制在大约250个排名内。尽管搜索行为截然不同，基于适应度的选择和随机选择都未产生可检测的选择差异或报告准确性的亲代到子代传递。因此，代理的自我报告应被视为需要与环境进行验证的主张，而不是其自身可靠性的证据。

View on arXiv Download PDF AI Translation

cs.AI / 59 / 2609.00654

SciTrue: Reliable Scientific Claim Validation with Frontier and Open Language Models at the NTCIR SciClaimEval Task

SciTrue：在NTCIR SciClaimEval任务中利用前沿和开放语言模型进行可靠的科学声明验证

Bao, Qiming, Tan, Neşet Özkan, Wang, Siyuan, Gahegan, Mark

Abstract

We describe the SciTrue team's participation in both subtasks of the NTCIR-19 SciClaimEval task~\cite{sciclaimeval}, which asks systems to verify scientific claims against the tables and figures of a paper. Rather than tuning a single model, we benchmark eleven frontier and open multimodal models under one honest, per-sample protocol and combine them with light, transparent post-processing. On the official, blind test leaderboard (Section~\ref{sec:results}), SciTrue placed first by a clear margin in three of the four evidence-category/subtask combinations, and tied for first on the primary metric in the fourth. Three findings explain the result. First, strong instruction-tuned models are already competitive: Claude Opus~4.8 and Gemma-4-31B each exceed the strongest public baseline (o4-mini), and GPT-5.5 and Claude Fable~5 lead both subtasks (97.7 on Subtask~2). Second, the task's pairing structure is the largest lever: a \emph{leak-free pair prior} that recovers the Supported/Refuted pairing from the claim text alone (a visible field) and assigns Supported to the higher-confidence evidence raises Subtask-1 pair-accuracy from 72.2 to 93.5, far more than any model swap or ensemble weighting. Third, a case-by-case audit finds that most residual errors are visually-undetectable label-mapping swaps or dataset label noise, so measured accuracy understates the true ability and the fixable-by-modeling headroom is small. Controlled fine-tuning, distillation, and agentic consistency-checking support the same conclusions, and we document throughout a measurement leak---label information reaching a system through the packaging of the data rather than its content---in which the released file ordering encodes the label, including one instance that briefly misled our own pipeline.

Chinese Translation

我们描述了SciTrue团队在NTCIR-19 SciClaimEval任务中的两个子任务的参与，该任务要求系统验证科学声明与论文中的表格和图形的关系。我们没有调优单一模型，而是在一个诚实的逐样本协议下对十一种前沿和开放的多模态模型进行了基准测试，并结合轻量、透明的后处理。在官方的盲测排行榜（第 ef{sec:results}节）上，SciTrue在四个证据类别/子任务组合中的三个中以明显优势名列第一，并在第四个主要指标上并列第一。三个发现解释了这一结果。首先，强大的指令调优模型已经具备竞争力：Claude Opus~4.8和Gemma-4-31B各自超过了最强的公共基线（o4-mini），而GPT-5.5和Claude Fable~5在两个子任务中均领先（在子任务2中达到97.7）。其次，任务的配对结构是最大的杠杆：一个 extit{无泄漏配对先验}，仅通过声明文本（一个可见的领域）恢复支持/反驳配对，并将支持分配给更高置信度的证据，使得子任务1的配对准确率从72.2提升至93.5，远超任何模型替换或集成加权的效果。第三，逐案审计发现，大多数剩余错误是视觉上不可检测的标签映射交换或数据集标签噪声，因此测量的准确率低估了真实能力，而通过建模可修复的提升空间很小。受控的微调、蒸馏和代理一致性检查支持相同的结论，我们在文中记录了一个测量泄漏——标签信息通过数据的包装而非内容到达系统——其中发布的文件顺序编码了标签，包括一个短暂误导我们自己管道的实例。

View on arXiv Download PDF AI Translation

cs.AI / 60 / 2609.00662

Drift-Aware LLM Routing with Sparse Contexts and Shared Budgets

漂移感知的稀疏上下文与共享预算的LLM路由

Lee, Cheung Hao, Wong, Patrick

Abstract

A multi-model language service must route each request while preserving workload-level budgets for compute, latency, memory, or monetary cost. Two features make this problem materially harder than static model selection. Prompt representations are high dimensional, so only a small subset of embedding directions may predict the incremental value of a model, and both the request mix and the model frontier drift after launches, fine-tunes, quantization changes, and system updates. We formulate nonstationary sparse contextual routing with multiple knapsack constraints and an optional shadow-audit stream that evaluates a small fraction of prompts on several models. We propose Drift-Aware Sparse Routing (DRS). The policy estimates reward and resource use from a rolling audit window, routes using pessimistic reward and optimistic cost estimates, updates resource shadow prices online, and applies a hard meter before commitment. The analysis separates control from statistics. On any event with uniform prediction radii $\{\beta_t\}$, regret against a paced dynamic fluid benchmark is bounded by the sum of the radii, a capacity-buffer term, and an $O(\sqrt{T})$ pacing term. Under a sparse linear model and bounded drift $V_T$, rolling estimation gives \[ \widetilde O\left( T\sqrt{\frac{s}{\rho W}}+WV_T+\sqrt{T} \right), \] where $s$ is sparsity, $\rho$ is the audit rate, and $W$ is the window length. Optimizing $W$ yields the usual stationary $O(\sqrt{sT/\rho})$ rate when $V_T=0$ and a $O(T^{2/3}(s/\rho)^{1/3}V_T^{1/3})$ adaptation term under drift.

Chinese Translation

多模型语言服务必须在保持计算、延迟、内存或货币成本的工作负载级预算的同时路由每个请求。两个特征使得这个问题比静态模型选择更为复杂。提示表示是高维的，因此只有一小部分嵌入方向可能预测模型的增量价值，并且请求组合和模型边界在发布、微调、量化变化和系统更新后都会发生漂移。我们提出了具有多个背包约束的非平稳稀疏上下文路由，并引入可选的影子审计流，以评估少量提示在多个模型上的表现。我们提出了漂移感知稀疏路由（Drift-Aware Sparse Routing，DRS）。该策略从滚动审计窗口中估计奖励和资源使用，使用悲观的奖励和乐观的成本估计进行路由，在承诺之前在线更新资源影子价格，并应用硬计量。分析将控制与统计分开。在任何具有均匀预测半径 $eta_t$ 的事件中，相对于一个有节奏的动态流体基准的遗憾被半径的总和、一个容量缓冲项和一个 $O( ext{sqrt}{T})$ 的节奏项所界定。在稀疏线性模型和有界漂移 $V_T$ 的情况下，滚动估计给出 \\[ ilde{O}igg( T ext{sqrt}igg(rac{s}{ ho W}igg)+WV_T+ ext{sqrt}{T} igg), \\] 其中 $s$ 是稀疏性，$ ho$ 是审计率，$W$ 是窗口长度。优化 $W$ 在 $V_T=0$ 时产生通常的平稳 $O( ext{sqrt}{sT/ ho})$ 速率，而在漂移情况下产生 $O(T^{2/3}(s/ ho)^{1/3}V_T^{1/3})$ 的适应项。

View on arXiv Download PDF AI Translation

cs.AI / 61 / 2609.00665

Triple-Bottom-Line Sustainability of Language Models for Edge AI: A Comparison Between SLMs and Quantized LLMs

边缘人工智能语言模型的三重底线可持续性：小语言模型与量化大语言模型的比较

Shah, Jainil Dharmil

Abstract

Edge-AI model selection is commonly driven by one isolated metric - accuracy, latency, memory, energy, or safety, even though a deployable language model must balance all five. Our work focuses on answering the question whether na- tively trained small language models (SLMs) or large language models (LLMs) compressed through post-training quantization offer the more sustainable edge- deployment trade-off. We introduce a reproducible Holistic Sustainability Score (HSS) organized around the triple bottom line: an economic pillar for capability and systems efficiency, an environmental pillar for operational GPU energy and a social pillar for harmful-prompt robustness. Five BF16 SLMs and five LLMs under different quantization approaches - BF16, INT8, NF4 4-bit, GPTQ 4-bit, and GGUF Q4 produce 30 measured configurations. Capability is assessed on five zero-shot benchmarks; efficiency uses latency, throughput, peak VRAM and energy; and safety is approximated by attack success rate on five harmful prompts. Qwen3-30B-A3B/GGUF Q4 ranks first in the combined pool (93.38), followed by Mistral-Small-24B/GGUF Q4 (92.40), while Phi-4-mini/BF16 is the highest- ranked SLM in that pool (89.49). Thus, the hypothesis that native SLMs must be the most sustainable edge choice is not supported universally; optimized quantized LLMs can win overall, while SLMs remain competitive through lower resource demand. Quantization is a systems-level choice rather than a monotonic precision- efficiency trade-off and HSS remains relative to its comparison pool and proxy definitions.

Chinese Translation

边缘人工智能模型的选择通常仅由一个孤立的指标驱动——准确性、延迟、内存、能耗或安全性，尽管可部署的语言模型必须在这五个方面之间取得平衡。我们的研究重点在于回答一个问题：原生训练的小语言模型（SLMs）与通过后训练量化压缩的大语言模型（LLMs）哪个在边缘部署中提供更可持续的权衡。我们引入了一个可重复的整体可持续性评分（Holistic Sustainability Score, HSS），该评分围绕三重底线组织：经济支柱关注能力和系统效率，环境支柱关注操作GPU能耗，社会支柱关注对有害提示的鲁棒性。我们评估了五个BF16小语言模型和五个采用不同量化方法的大语言模型——BF16、INT8、NF4 4位、GPTQ 4位和GGUF Q4，产生了30个测量配置。能力在五个零样本基准上进行评估；效率使用延迟、吞吐量、峰值显存和能量进行衡量；安全性通过对五个有害提示的攻击成功率进行近似。Qwen3-30B-A3B/GGUF Q4在综合池中排名第一（93.38），其次是Mistral-Small-24B/GGUF Q4（92.40），而Phi-4-mini/BF16是该池中排名最高的小语言模型（89.49）。因此，原生小语言模型必须是最可持续的边缘选择这一假设并不普遍成立；经过优化的量化大语言模型在整体上可以胜出，而小语言模型通过较低的资源需求保持竞争力。量化是一种系统级选择，而不是单调的精度-效率权衡，HSS仍然相对于其比较池和代理定义保持相对性。

View on arXiv Download PDF AI Translation

cs.AI / 62 / 2609.00700

Value Over Language Model: Detecting Original Contribution in Writing

超越语言模型的价值：检测写作中的原创贡献

Sharma, Vibhhu, Joachims, Thorsten, Dean, Sarah

Abstract

LLMs have been rapidly adopted across writing tasks, prompting the development of tools for detecting LLM-generated text. Yet, these tools largely measure how much of a document's surface text was written by an LLM and aren't fundamentally designed to measure how much of the information content or ideas originated from the LLM itself rather than being supplied by the user in the prompt. In this work, we design a framework that measures how much value a person adds on top of what a language model could have easily produced by itself. The method requires no training or labeled data and never scores the document's surface text, insulating it from stylistic confounders. Instead, it extracts the document's content at increasing levels of granularity, uses an LLM to reconstruct the document from each partial representation, and compares these reconstructions with those produced from the task description alone. We call this framework Value Over Language Model (VOLM), which measures a document's contribution relative to a replacement-level document that an LLM could produce from the task description alone. We evaluate VOLM with a specific instantiation of this framework across three domains: news articles, ICLR peer reviews, and argumentative essays. VOLM separates human-authored documents from matched LLM-generated documents produced from generic task descriptions, while remaining substantially invariant to content-preserving transformations, including LLM-based reconstruction and round-trip translation. We further find that increasingly constrained content extractors reduce residual differences between LLM-generated and humanized text, demonstrating the importance of disentangling informational content from stylistic variation. We hope these results encourage further work on specialized instantiations of the framework and on assessing human contributions in LLM-assisted writing more generally.

Chinese Translation

大型语言模型（LLMs）在写作任务中被迅速采用，这促使了用于检测LLM生成文本的工具的发展。然而，这些工具主要测量文档表面文本中有多少是由LLM撰写的，并未从根本上设计用于衡量信息内容或思想中有多少源自LLM本身，而不是由用户在提示中提供。在本研究中，我们设计了一个框架，用于衡量一个人相对于语言模型本身能够轻松生成的内容所增加的价值。该方法不需要训练或标记数据，并且从不对文档的表面文本进行评分，从而使其免受风格干扰。相反，它以逐步细化的方式提取文档内容，利用LLM从每个部分表示重构文档，并将这些重构与仅从任务描述生成的内容进行比较。我们将该框架称为超越语言模型的价值（Value Over Language Model，VOLM），它衡量文档相对于LLM仅能从任务描述生成的替代文档的贡献。我们在三个领域评估了VOLM的特定实例：新闻文章、ICLR同行评审和论证性论文。VOLM能够区分人类撰写的文档与从通用任务描述生成的匹配LLM生成的文档，同时对内容保留变换（包括基于LLM的重构和往返翻译）保持实质性不变。我们进一步发现，越来越严格的内容提取器减少了LLM生成文本与人性化文本之间的残余差异，证明了将信息内容与风格变异解耦的重要性。我们希望这些结果能够鼓励对该框架的专门实例化以及对LLM辅助写作中人类贡献的更广泛评估的进一步研究。

View on arXiv Download PDF AI Translation

cs.AI / 63 / 2609.00714

ChatDev 2.0: A No-Code Multi-Agent Platform for Developing Everything

ChatDev 2.0：一个无代码的多智能体平台，用于开发一切

Dang, Yufan, Yao, Shu, Lai, Bowen, Xu, Chenting, Shi, Ruijie, Leung, Wai-Shing, Li, Huatao, Qian, Chen, Liu, Zhiyuan

Abstract

Large language model (LLM)-based multi-agent systems (MAS) have shown strong potential for solving complex tasks, yet their development forces a tradeoff: code frameworks are expressive but engineering-intensive, while no-code builders simplify authoring but constrain agent interactions to author-defined workflows. We present ChatDev 2.0: DevAll (hereafter DevAll), a no-code platform for building, executing, and inspecting heterogeneous MAS that delivers both high expressiveness and ease of use. In terms of expressiveness, DevAll pairs a declarative executable graph abstraction with a cycle-aware execution engine, so that heterogeneous agents and dynamic and cyclic interactions can be represented and executed within a single framework. For ease of use, an integrated visual interface lets users author, run, monitor, and inspect MAS, including human-in-the-loop steps, entirely without writing code. Experiments demonstrate that DevAll reproduces state-of-the-art MAS across three representative tasks at competitive performance and without task-specific orchestration code, highlighting its effectiveness as a general-purpose platform for LLM-based MAS. DevAll is available at https://github.com/OpenBMB/ChatDev.

Chinese Translation

基于大型语言模型（LLM）的多智能体系统（MAS）在解决复杂任务方面展现出强大的潜力，但其开发面临权衡：代码框架表达能力强但工程投入大，而无代码构建工具简化了创作过程，但限制了智能体交互在作者定义的工作流程内。我们提出了ChatDev 2.0：DevAll（以下简称DevAll），这是一个无代码平台，用于构建、执行和检查异构的MAS，提供高表达能力和易用性。在表达能力方面，DevAll将声明式可执行图抽象与循环感知执行引擎相结合，使得异构智能体及动态和循环交互能够在单一框架内表示和执行。在易用性方面，集成的可视化界面使用户能够完全不编写代码地创作、运行、监控和检查MAS，包括人机协作步骤。实验表明，DevAll在三个代表性任务上再现了最先进的MAS，且性能具有竞争力，无需特定任务的编排代码，突显了其作为基于LLM的MAS通用平台的有效性。DevAll可在https://github.com/OpenBMB/ChatDev获取。

View on arXiv Download PDF AI Translation

cs.AI / 64 / 2609.00718

A Closed-Loop Evaluation of Capability Loss and Recovery in Compressed Driving Policies

压缩驾驶策略中能力损失与恢复的闭环评估

Irfan, Ahmad Alfan Alfian, Khatim, Nur Ahmad, Arief, Mansur

Abstract

Many automobile and mobility companies deploy learned driving policies on embedded computers with limited memory and power. Pruning, knowledge distillation, and quantization are the standard methods to reduce the size and the inference cost of these policies. However, these methods are commonly assessed by aggregate numerical scores, and such scores may not reflect the ability of the policy to drive safely when interacting with other road users. In this study, we propose a stage-wise closed-loop evaluation approach to follow a driving policy through a compression pipeline. We formulate the driving task as a partially observable Markov decision process (POMDP) and train a belief-state policy with proximal policy optimization (PPO) in Gym-Duckietown. We then extract the actor, compress it one stage at a time, and evaluate it on five driving curricula. We show that structured pruning is the stage at which the driving capability is first lost. Meanwhile, distillation improves the pruned actor, but the improvement is limited by its rehearsal data. Integer quantization of the improved actor loses some of the curricula that require the vehicle to stop and then resume. Interestingly, the same procedure on the unpruned actor preserves all five curricula. Our study thus provides an empirical analysis aiming to answer the currently active discussions on how to accept a compressed driving policy, so as to achieve a safe and statistically reliable deployment of automated driving functions.

Chinese Translation

许多汽车和出行公司在内存和功耗有限的嵌入式计算机上部署学习到的驾驶策略。剪枝、知识蒸馏和量化是减少这些策略大小和推理成本的标准方法。然而，这些方法通常通过汇总的数值评分进行评估，而这些评分可能无法反映策略在与其他道路使用者互动时安全驾驶的能力。在本研究中，我们提出了一种阶段性闭环评估方法，以跟踪压缩管道中的驾驶策略。我们将驾驶任务形式化为部分可观察的马尔可夫决策过程（POMDP），并在Gym-Duckietown中使用近端策略优化（PPO）训练信念状态策略。然后，我们逐阶段提取演员，进行压缩，并在五个驾驶课程上进行评估。我们展示了结构化剪枝是驾驶能力首次丧失的阶段。同时，蒸馏改善了剪枝后的演员，但这种改善受到其重放数据的限制。改进后的演员的整数量化丢失了一些要求车辆停止然后重新启动的课程。有趣的是，在未剪枝的演员上进行相同的程序可以保留所有五个课程。因此，我们的研究提供了一项实证分析，旨在回答当前关于如何接受压缩驾驶策略的活跃讨论，以实现自动驾驶功能的安全和统计可靠部署。

View on arXiv Download PDF AI Translation

cs.AI / 65 / 2609.00728

SOVER: Formal Certification of Optimization Reformulations via LLM-Assisted SMT Verification

SOVER：通过LLM辅助的SMT验证进行优化重构的形式化认证

Bhattacharyya, Swapnil, Baranwal, Mayank

Abstract

Large Language Models (LLMs) have shown remarkable promise in translating and reformulating complex mathematical optimization problems across modeling languages. However, validating such transformations through empirical solver executions alone is unreliable, as solver outcomes may be affected by local minima, structural timeouts, numerical artifacts, and subtle semantic divergence between formulations. We introduce SOVER, an LLM-assisted SMT framework that separates semantic mapping from formal certification: Z3 checks domain cross-feasibility and global objective-order preservation for mixed-integer linear formulations, while dReal provides tolerance-aware feasibility/range and $\epsilon$-argmin checks for continuous nonlinear formulations. We also introduce NLEquiv-150, a public benchmark of 100 equivalent and 50 deliberately hard non-equivalent nonlinear reformulation pairs. With LLM-extracted mappings, SOVER classifies 149/150 pairs (99.33%) correctly, including all 50 hard negatives; the sole error is an incomplete mapping extraction.

Chinese Translation

大型语言模型（LLMs）在跨建模语言翻译和重构复杂数学优化问题方面展现出了显著的潜力。然而，仅通过经验求解器执行来验证此类变换是不可靠的，因为求解器的结果可能受到局部最小值、结构超时、数值伪影以及不同表述之间微妙的语义差异的影响。我们提出了SOVER，一个LLM辅助的SMT框架，它将语义映射与形式认证分离：Z3检查混合整数线性表述的领域交叉可行性和全局目标顺序保持，而dReal则为连续非线性表述提供容忍度感知的可行性/范围和$ ext{ε}$-argmin检查。我们还引入了NLEquiv-150，这是一个包含100对等效和50对故意困难的非等效非线性重构对的公共基准。在LLM提取的映射下，SOVER正确分类了149/150对（99.33%），包括所有50个困难负例；唯一的错误是映射提取不完整。

View on arXiv Download PDF AI Translation

cs.AI / 66 / 2609.00731

Agentic Empirical Asset Pricing: Methodological Foundations

代理经验资产定价：方法论基础

Pan, Yingjian, Ding, Xiaowei, Giesecke, Kay

Abstract

Recent advances in LLM agents enable a new paradigm for asset pricing, which we call Agentic Empirical Asset Pricing (AEAP): systems that autonomously conduct the scientific discovery process itself. We define AEAP and identify its core building blocks. Existing evaluation practices backtest only the outputs (factors or trades), not the autonomous discovery system that produced them. We focus on factor discovery, contributing a reference architecture, a rigorous evaluation standard for discovered factors, and a method for out-of-sample backtesting the discovery system. As a concrete instance of that architecture, we evaluate SEADS against five re-implemented baselines on two US equity panels using this standard: no single metric ranks the systems consistently, motivating evaluation on multiple axes at once. A separate rolling re-execution then asks the complementary question of whether the discovery process itself, not one static output, is reliable. We also report negative findings and limitations that surface further evaluation pitfalls for future AEAP systems.

Chinese Translation

近期在大型语言模型（LLM）代理方面的进展使得资产定价出现了一种新的范式，我们称之为代理经验资产定价（Agentic Empirical Asset Pricing, AEAP）：这些系统能够自主进行科学发现过程。我们定义了AEAP并识别其核心构建模块。现有的评估实践仅对输出（因子或交易）进行回测，而不是对产生这些输出的自主发现系统进行评估。我们专注于因子发现，贡献了一个参考架构、对发现因子的严格评估标准，以及一种对发现系统进行样本外回测的方法。作为该架构的具体实例，我们使用这一标准对SEADS在两个美国股票面板上与五个重新实现的基准进行评估：没有单一指标能够一致地对系统进行排名，这促使我们在多个维度上同时进行评估。随后进行的独立滚动重新执行则提出了一个互补性的问题，即发现过程本身（而非单一静态输出）是否可靠。我们还报告了负面发现和局限性，这些问题进一步揭示了未来AEAP系统评估的陷阱。

View on arXiv Download PDF AI Translation

cs.AI / 67 / 2609.00738

Escaping Redundant Reasoning: Structure-Aware Search for Inference-Time LLMs

逃避冗余推理：面向结构的推理时间大型语言模型搜索

Cheng, Lu

Abstract

Inference-time search with large language models (LLMs) often concentrates on a small set of structurally or semantically similar trajectories, leaving alternatives underexplored---a failure mode we call \textit{reasoning basin collapse}. We introduce BASIN, a training-free, structure-aware selection method that groups reasoning states into basins and penalizes repeated visits to the same strategy, thereby reallocating search across genuinely distinct reasoning paths under a fixed compute budget. Under matched inference budgets, BASIN improves over Tree of Thoughts (ToT) by up to $+22$pp on Game of 24 and $+6.7$pp on MuSR. A quality-aware variant, QA-BASIN, further improves robustness by preserving high-quality basins when unconditional diversification over-explores. To explain when basin-aware selection helps, we introduce the redundancy gap $\Delta$, which measures how differently search concentrates for correct versus incorrect predictions: standard ToT often operates near $\Delta \approx 0$, while BASIN consistently shifts $\Delta$ positive. More broadly, BASIN suggests structure-aware selection as a simple and general approach to improving inference-time reasoning. Code can be found at https://github.com/GitHubLuCheng/basin.

Chinese Translation

大型语言模型（LLMs）的推理时间搜索通常集中于一小部分结构上或语义上相似的轨迹，导致其他选择未被充分探索——我们称之为 extit{推理盆地崩溃}的失败模式。我们提出了BASIN，这是一种无训练的、面向结构的选择方法，能够将推理状态分组为盆地，并对重复访问相同策略进行惩罚，从而在固定的计算预算下重新分配搜索到真正不同的推理路径。在匹配的推理预算下，BASIN在24点游戏上比思想树（Tree of Thoughts, ToT）提高了最多$+22$个百分点，在MuSR上提高了$+6.7$个百分点。一个质量感知的变体QA-BASIN在无条件多样化过度探索时，通过保留高质量的盆地进一步提高了鲁棒性。为了说明盆地感知选择何时有帮助，我们引入了冗余差距$ riangle$，该指标衡量搜索在正确预测与错误预测之间的集中程度差异：标准的ToT通常在$ riangle ext{接近} 0$附近运行，而BASIN则始终将$ riangle$向正方向移动。更广泛地说，BASIN建议将面向结构的选择作为一种简单而通用的方法来改善推理时间的推理。代码可以在https://github.com/GitHubLuCheng/basin找到。

View on arXiv Download PDF AI Translation

cs.AI / 68 / 2609.00749

ContextPipe: Database-Inspired Context Assembly for Long-Horizon Agents

ContextPipe：基于数据库的长时间跨度智能体上下文组装

Xu, Peng, Zhang, Zuyu, Sun, Yuze, Tian, Feng, Wang, Long, Zhang, Chen

Abstract

Long-horizon large language model (LLM) agents require context assembly: the runtime must decide what to include in each prompt, in what order, and when to compact history under a hard context-window budget and a byte-sensitive prompt cache. In production agentic systems, this logic is scattered across prompt builders, ad hoc compaction routines, cache-break workarounds, and per-provider shims. We argue that context assembly is structurally isomorphic to query execution in a relational database: both execute under a hard budget, exploit a tiered cache, and leverage statistics. We adopt this discipline in ContextPipe: a five-phase pipeline (Plan Bind Optimize Execute Feedback) backed by a structured data-source catalog, a deterministic cache-aware optimizer, and an EXPLAIN ANALYZE trace. We show that context in ContextPipe is auditable, replayable, and failure-isolated. A preliminary evaluation using the SWE-bench Pro Qutebrowser subset shows that, compared with the append-only context construction policy, ContextPipe reduces total token volume by 31%, LLM calls by 23%, and response time by 9%, at the cost of a lower KV cache-hit ratio.

Chinese Translation

长时间跨度的大型语言模型（LLM）智能体需要上下文组装：在运行时必须决定在每个提示中包含什么内容、以何种顺序以及何时在严格的上下文窗口预算和对字节敏感的提示缓存下压缩历史。在生产智能系统中，这种逻辑分散在提示构建器、临时压缩例程、缓存破坏的变通方法和每个提供者的适配器中。我们认为上下文组装在结构上与关系数据库中的查询执行同构：两者都在严格预算下执行，利用分层缓存，并利用统计信息。我们在ContextPipe中采用这种方法：一个五阶段的管道（计划、绑定、优化、执行、反馈），由结构化数据源目录、确定性的缓存感知优化器和EXPLAIN ANALYZE跟踪支持。我们展示了ContextPipe中的上下文是可审计的、可重放的，并且具有故障隔离性。使用SWE-bench Pro Qutebrowser子集的初步评估表明，与仅追加上下文构建策略相比，ContextPipe将总令牌量减少了31%，LLM调用减少了23%，响应时间减少了9%，但代价是较低的KV缓存命中率。

View on arXiv Download PDF AI Translation

cs.AI / 69 / 2609.00755

S^3martCirc: Self-supervised Smart Circuit Discovery

S^3martCirc：自监督智能电路发现

Zheng, Wendy, He, Yinhan, Wu, Liang, Li, Jundong

Abstract

Large Language Models (LLMs) have demonstrated remarkable performance across diverse tasks, from text summarization to question answering. Despite these capabilities, their black-box nature obscures internal decision-making processes. Mechanistic interpretability (MI) aims to address this by reverse-engineering neural networks into human-understandable algorithms. Current MI approaches for LLMs typically follow a two-stage paradigm: first identifying important components (circuit discovery), where components are typically individual nodes such as an attention head or feedforward neuron, and second determining the role they play in a certain task (functional interpretation). However, this sequential approach overlooks a fundamental insight: a component's importance and its functional role are inherently codependent. Unifying these stages presents two key challenges: (1) functional roles are often tied to specific nodes or components, limiting generalization, and (2) their identification relies on subjective interpretation rather than quantifiable metrics. To address these challenges, we propose S^3martCirc (Self-supervised Smart Circuit Discovery), a unified framework that simultaneously discovers circuits and interprets functionality. S^3martCirc abstracts node behavior into two general computational roles that generalize across tasks and defines a quantitative metric for assigning them, enabling importance and functional role to be discovered jointly rather than in sequence. Extensive experiments show that our framework outperforms existing methods in circuit discovery.

Chinese Translation

大型语言模型（LLMs）在文本摘要到问答等多种任务中表现出色。尽管具备这些能力，它们的黑箱特性却使内部决策过程变得不透明。机制可解释性（MI）旨在通过将神经网络逆向工程为人类可理解的算法来解决这一问题。目前针对LLMs的MI方法通常遵循两阶段范式：第一阶段识别重要组件（电路发现），这些组件通常是单个节点，如注意力头或前馈神经元；第二阶段确定它们在特定任务中的作用（功能解释）。然而，这种顺序方法忽视了一个基本的见解：组件的重要性和其功能角色是相互依赖的。统一这两个阶段面临两个关键挑战：（1）功能角色通常与特定节点或组件相关，限制了泛化能力；（2）其识别依赖于主观解释而非可量化指标。为了解决这些挑战，我们提出了S^3martCirc（自监督智能电路发现），一个统一框架，能够同时发现电路并解释其功能。S^3martCirc将节点行为抽象为两个在任务间泛化的通用计算角色，并定义了一个定量指标来分配这些角色，从而使重要性和功能角色能够共同发现，而非顺序发现。大量实验表明，我们的框架在电路发现方面优于现有方法。

View on arXiv Download PDF AI Translation

cs.AI / 70 / 2609.00763

Automated Tree Knowledge Graph Construction using Ontology Expansion and Retrieval from Vietnamese History Textbooks

基于本体扩展和越南历史教科书检索的自动化树知识图谱构建

Nguyen, Ket Doan, Nguyen, Minh N. H.

Abstract

Hierarchical Knowledge graph (KG)-based retrieval augmented generation (RAG) has emerged as a powerful approach for supporting large language models with structured knowledge. However, there are primary challenges: (i) the lack of methods for automatic KG construction using ontology expansion for low-resource languages such as Vietnamese, (ii) the absence of systematic evaluation for knowledge retrieval strategies leveraging the hierarchical structures. In this paper, we propose an end-to-end pipeline for KG construction and retrieval strategies evaluation. In the KG construction, we employ a three-phase hybrid relation extraction pipeline: intra-batch deduplication via Union-Find, approximate cross-batch search, and LLM extraction with a centroid filter that reduces prompts combined with a five-step dual-LLM validator to prevent bloated ontology. A two-tier architecture consists of unmergeable structural nodes to preserve the document structure and mergeable content nodes. The retrieval evaluation consists of three graph traversal strategies: Top-Down, Horizontal, and Bottom-Up, which are evaluated on a synthetically generated benchmark of 1,210 Vietnamese queries from 109 subgraphs, categorized by five query directions. In this paper, we construct the tree knowledge graph from Vietnamese high school History textbooks (nearly 400 pages) to produce 750 nodes and 4,341 semantic edges with controlled ontology growth from 40 to 41 types. Among experimental graph traversal strategies, the Top-Down strategy with structure surpasses the vector baseline by 4.7 percentage points in NDCG@10. As a result, tree-structural information provides valuable information beyond flat cosine similarity but degrades performance when the query does not require structural context.

Chinese Translation

基于层次知识图谱（KG）的检索增强生成（RAG）已成为支持大型语言模型与结构化知识的强大方法。然而，主要面临以下挑战：（i）缺乏针对越南等低资源语言的自动KG构建方法，尤其是利用本体扩展的方式；（ii）缺乏对利用层次结构的知识检索策略的系统评估。本文提出了一种端到端的KG构建和检索策略评估管道。在KG构建中，我们采用了三阶段混合关系提取管道：通过并查集进行批内去重、近似批间搜索，以及结合质心过滤器的LLM提取，该过滤器减少了提示，并结合五步双LLM验证器以防止本体膨胀。该架构由不可合并的结构节点和可合并的内容节点组成，以保持文档结构。检索评估包括三种图遍历策略：自上而下、水平和自下而上，这些策略在从109个子图生成的1,210个越南查询的合成基准上进行评估，并按五个查询方向分类。本文从越南高中历史教科书（近400页）构建树知识图谱，生成750个节点和4,341个语义边，控制本体从40种类型增长到41种。在实验图遍历策略中，自上而下策略的结构在NDCG@10中超越了向量基线4.7个百分点。因此，树结构信息提供了超越平面余弦相似度的有价值信息，但在查询不需要结构上下文时会降低性能。

View on arXiv Download PDF AI Translation

cs.AI / 71 / 2609.00768

DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory

DiagEvo：通过层次错误记忆的诊断引导自我进化

Wei, Xincheng, Ding, Yifan, Li, Yoshua, Ma, Dongsheng, Weng, Rongxiang, Cai, Xunliang, Ding, Wenjian, Zhang, Yao

Abstract

Self-play is an effective paradigm for language-model self-evolution, but without guidance, solver performance can plateau or decline across rounds. Unguided methods steer question generation with signals such as difficulty, learnability, or diversity. These signals keep questions challenging and varied but do not specify which unresolved reasoning weaknesses later rounds should target. Guided methods obtain direction from external task resources, including human examples, document corpora, or specified difficulty targets, and therefore rely on task information supplied outside the self-play loop. We show that the needed direction can instead be derived from the solver's own failure history. We introduce DiagEvo, whose diagnostician extracts recurring error causes from this history and stores them in a hierarchical error-cause memory. The memory groups related causes under skill nodes and tracks each as Active or Mastered according to self-consistency on targeted questions. The challenger uses these states and recurrence counts to balance cause-targeted generation with free exploration. Double-confidence filtering retains intermediate-difficulty questions only when the most common solver answer has a clear vote lead. DiagEvo derives its curriculum from information produced during self-play, without external task resources. With the default 4B diagnostician, DiagEvo outperforms every baseline in mean accuracy across all nine benchmarks for each of the three solvers: Qwen3-4B, Qwen3-8B, and OctoThinker-8B. On Qwen3-8B, it reaches 72.3% mean accuracy across five mathematical reasoning benchmarks, 4.5 percentage points above R-Zero. Its mean accuracy across all nine benchmarks is 57.4%, 1.1 percentage points above DARC. Ablations show that the hierarchical error-cause memory and double-confidence filtering both contribute to these gains.

Chinese Translation

自我对弈是一种有效的语言模型自我进化范式，但在没有指导的情况下，求解器的性能可能在多个回合中停滞不前或下降。无指导的方法通过难度、可学习性或多样性等信号来引导问题生成。这些信号保持问题的挑战性和多样性，但并未明确指出后续回合应针对哪些未解决的推理弱点。受指导的方法从外部任务资源中获取方向，包括人类示例、文档语料库或指定的难度目标，因此依赖于自我对弈循环外提供的任务信息。我们展示了所需的方向可以从求解器自身的失败历史中推导出来。我们引入了DiagEvo，其诊断器从这一历史中提取反复出现的错误原因，并将其存储在层次错误原因记忆中。该记忆将相关原因归类于技能节点，并根据在目标问题上的自我一致性跟踪每个原因的状态为“活跃”或“已掌握”。挑战者利用这些状态和重复计数来平衡针对原因的生成与自由探索。双重置信过滤仅在最常见的求解器答案具有明显投票优势时保留中等难度问题。DiagEvo从自我对弈过程中产生的信息中推导其课程，而无需外部任务资源。在默认的4B诊断器下，DiagEvo在所有九个基准测试中对每个求解器（Qwen3-4B、Qwen3-8B和OctoThinker-8B）的平均准确率均优于所有基线。在Qwen3-8B上，它在五个数学推理基准测试中达到了72.3%的平均准确率，比R-Zero高出4.5个百分点。其在所有九个基准测试中的平均准确率为57.4%，比DARC高出1.1个百分点。消融实验表明，层次错误原因记忆和双重置信过滤均对这些提升有所贡献。

View on arXiv Download PDF AI Translation

cs.AI / 72 / 2609.00782

When Features Become Instances: Inverted Contrastive Learning for Unsupervised Feature Selection

当特征变为实例：用于无监督特征选择的反向对比学习

Ghosh, Utsab, Chakraborty, Roshni

Abstract

Unsupervised feature selection seeks a compact subset of informative features without access to class labels, making feature utility difficult to define. Existing UFS methods therefore rely on indirect structural criteria, such as similarity preservation, locality, sparsity, cluster geometry, or reconstruction quality. In this paper, we instead study UFS through representation consistency and propose Inverted Contrastive Learning for Unsupervised Feature Selection (ICLFS), a feature-wise contrastive framework that reformulates UFS as a representation learning problem over features rather than samples. ICLFS first inverts the data matrix so that each feature is represented by its sample-profile vector, then constructs multiple masked positive views together with a shuffled negative view, and learns projector-space representations that remain consistent across these structured perturbations under an InfoNCE-based objective. Motivated by recent findings that cosine-based and InfoNCE-based training affect embedding norms, we use projector-space embedding magnitude as the saliency signal for ranking features. The resulting norm-based ranking is subsequently refined through Laplacian-Gated Ranking Correction, which suppresses locally redundant candidates while preserving salient ones. Extensive experiments on 12 benchmark datasets show that ICLFS achieves the best clustering accuracy on 10 datasets against both classical and neural baselines under the standard clustering-based UFS evaluation protocol, while remaining competitive on the other two. These results show that feature-wise contrastive representation consistency provides a strong and effective alternative to neighborhood, cluster, and reconstruction-based UFS formulations.

Chinese Translation

无监督特征选择旨在在没有类别标签的情况下寻找一个紧凑的有信息量的特征子集，这使得特征的效用难以定义。因此，现有的无监督特征选择（UFS）方法依赖于间接的结构标准，如相似性保持、局部性、稀疏性、聚类几何形状或重建质量。在本文中，我们通过表示一致性研究无监督特征选择，并提出反向对比学习用于无监督特征选择（Inverted Contrastive Learning for Unsupervised Feature Selection, ICLFS），这是一种特征级对比框架，将UFS重新表述为一个针对特征而非样本的表示学习问题。ICLFS首先反转数据矩阵，使每个特征由其样本特征向量表示，然后构建多个掩蔽的正视图以及一个打乱的负视图，并在基于InfoNCE的目标下学习在这些结构扰动中保持一致的投影空间表示。受到近期研究发现的启发，即基于余弦和基于InfoNCE的训练会影响嵌入的范数，我们使用投影空间嵌入的大小作为特征排名的显著性信号。最终的基于范数的排名通过拉普拉斯门控排名修正（Laplacian-Gated Ranking Correction）进行精细化，该方法抑制了局部冗余候选项，同时保留了显著的特征。在12个基准数据集上的大量实验表明，ICLFS在标准的基于聚类的UFS评估协议下，在10个数据集上实现了最佳的聚类准确性，相较于经典和神经基线表现出色，而在其他两个数据集上也保持了竞争力。这些结果表明，特征级对比表示一致性为基于邻域、聚类和重建的UFS公式提供了一种强大而有效的替代方案。

View on arXiv Download PDF AI Translation

cs.AI / 73 / 2609.00787

StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability?

StudyBench：自我进化能否挤压教科书以提升奥林匹克能力？

Chen, Yinghao, Chen, Zixi, He, Bingxiang, Qiao, Ziqing, Gao, Huan-ang, Xu, Yinuo, Zuo, Yuxin, Liu, Zeyuan, Zhan, Yuhao, Xiao, Chaojun

Abstract

Humans need to study only a handful of well-written textbooks to master a discipline and attempt its hardest problems. We argue that an ideal self-evolution method should share the same property, that is autonomously learning from raw training material for transferable problem-solving capability. However, we still lack a direct measurement for it. We introduce StudyBench, a controlled physics benchmark that directly measures how efficiently a self-evolution method converts training material into capability. We organise the test set into an Application Set, consisting of difficult textbook problems and evaluating absorption ability, and a Transfer Set, consisting of olympiad-level problems and evaluating transfer ability. Benchmarking representative self-evolution methods across three base models, we find that improvements on the Application Set rarely translate to the harder Transfer Set. A guidance ablation exposes a Guidance Gap: even the strongest method closes only a small fraction of what the same material unlocks when supplied as in-context guidance. Besides, every method hits a Compute Plateau, saturating well before exhausting its compute budget. The remaining gap is therefore a method problem rather than a data or compute problem. By offering a clean and controlled benchmark, StudyBench turns self-evolution progress from an open-ended pursuit into a measurable target for future research. Our code is released at https://github.com/thunlp/StudyBench.

Chinese Translation

人类只需学习少数几本优秀的教科书便能掌握一门学科并尝试其最难的问题。我们认为，理想的自我进化方法应具备相同的特性，即能够从原始训练材料中自主学习，以获得可转移的问题解决能力。然而，我们仍然缺乏对其的直接测量。我们引入了StudyBench，一个受控的物理基准，直接测量自我进化方法将训练材料转化为能力的效率。我们将测试集组织为应用集（Application Set），由困难的教科书问题组成，用于评估吸收能力，以及转移集（Transfer Set），由奥林匹克水平的问题组成，用于评估转移能力。在对三种基础模型进行代表性自我进化方法的基准测试时，我们发现应用集的改进很少转化为更难的转移集。指导消融实验揭示了一个指导差距（Guidance Gap）：即使是最强的方法也只能实现同一材料在作为上下文指导时所解锁的能力的一小部分。此外，每种方法都达到了计算平台（Compute Plateau），在耗尽其计算预算之前就已经饱和。因此，剩余的差距是方法问题，而不是数据或计算问题。通过提供一个干净且受控的基准，StudyBench将自我进化的进展从一个开放式追求转变为未来研究的可测量目标。我们的代码已发布在 https://github.com/thunlp/StudyBench。

View on arXiv Download PDF AI Translation

cs.AI / 74 / 2609.00805

Towards a Reliable and Practical Eval Pipeline

迈向可靠且实用的评估管道

Nguyen, Emma Thuong, Ghose, Abhishek

Abstract

LLM-based software systems increasingly require effective "evals" as quality gates in the development lifecycle. However, existing work typically addresses individual aspects of eval reliability rather than the full set of practical requirements. We present an end-to-end eval pipeline that combines eval checklist creation, with learned aggregation for checklist responses, to improve agreement across LLM judges and accuracy against human judgments. The framework additionally pro- vides self-consistency, explanations, and prediction uncertainty, and we empirically demonstrate its effectiveness.

Chinese Translation

基于大型语言模型（LLM）的软件系统越来越需要有效的“评估”作为开发生命周期中的质量门槛。然而，现有的研究通常只关注评估可靠性的个别方面，而不是完整的实际需求集合。我们提出了一种端到端的评估管道，该管道结合了评估检查表的创建与对检查表响应的学习聚合，以提高LLM评审者之间的一致性和与人类判断的准确性。该框架还提供自我一致性、解释和预测不确定性，并且我们通过实证研究证明了其有效性。

View on arXiv Download PDF AI Translation

cs.AI / 75 / 2609.00813

One Policy, Any Budget: Internalizing Budget-Aware Search via Reinforcement Learning

一项政策，任何预算：通过强化学习内化预算感知搜索

Sun, Xiaowei, Li, Jin, Hong, Yili, Fu, Yikun, Xiao, Yanghua

Abstract

While reinforcement learning has enabled LLM-based search agents to invoke external tools, existing methods train under fixed budgets and cannot adapt when constraints vary at deployment. We propose AnySearch, a framework that enables a single policy to perform budget-aware search under any budget constraint through a training scaffold and curriculum reinforcement learning. In the first phase, we train the agent with explicit budget state injection and structured reasoning prompts that guide efficient allocation under linearly decaying budgets. In the second phase, the scaffold is removed and the agent learns to operate autonomously under adaptively sampled budget constraints, matching inference conditions. Both phases are optimized with a composite reward that couples answer accuracy with budget efficiency through absolute and relative signals, where an adaptive weight amplifies the efficiency signal for high-accuracy queries and attenuates it for low-accuracy ones. Extensive experiments on seven general and multi-hop QA benchmarks show that our method outperforms baselines across all budget scales, generalizes to unseen constraints beyond the training range, and achieves superior tool productivity without excessive token overhead. Our code is available at https://github.com/xwsun01/AnySearch.

Chinese Translation

尽管强化学习使基于大型语言模型（LLM）的搜索代理能够调用外部工具，但现有方法在固定预算下进行训练，无法在部署时适应变化的约束。我们提出了AnySearch，一个框架，使单一策略能够在任何预算约束下执行预算感知搜索，采用训练支架和课程强化学习。在第一阶段，我们通过显式的预算状态注入和结构化推理提示来训练代理，这些提示指导在线性衰减预算下的有效分配。在第二阶段，支架被移除，代理学习在自适应采样的预算约束下自主操作，以匹配推理条件。两个阶段都通过复合奖励进行优化，该奖励将答案准确性与预算效率结合，通过绝对和相对信号进行反馈，其中自适应权重增强高准确性查询的效率信号，而减弱低准确性查询的效率信号。在七个通用和多跳问答基准上的大量实验表明，我们的方法在所有预算规模上均优于基线，能够推广到超出训练范围的未见约束，并在不产生过多标记开销的情况下实现更高的工具生产力。我们的代码可在 https://github.com/xwsun01/AnySearch 获取。

View on arXiv Download PDF AI Translation

cs.AI / 76 / 2609.00818

AnalysisBank: An Expert Analysis Pattern Library for Financial Report Generation

AnalysisBank：用于财务报告生成的专家分析模式库

Yang, Yajing, Ma, Yunshan, Koa, Kelvin J. L., Kan, Min-Yen

Abstract

We argue that financial report generation should operate at the analytical rather than structural level, composing content from data-derived insights rather than high-level topics or sections. To this end, we propose AnalysisBank, which distills expert reports into a reusable library of Analyses, each pairing a data signal, an analytical move, and the expert span it was derived from. At inference time, AnalysisBank matches input signals to library entries and applies the retrieved moves to compose the report. A study of Analyses distilled from 550 expert reports reveals a heavy-tailed distribution of 47-52 signal types spanning 13 move types. On two financial benchmarks across four LLM backbones, AnalysisBank increases the proportion of novel, data-grounded insights by 1.7-3.7x over structural-level baselines. Transfer to scientific writing suggests that the distinction generalizes beyond finance. Code and the distilled Analysis library are available at https://github.com/yajingyang/AnalysisBank.

Chinese Translation

我们认为财务报告生成应在分析层面而非结构层面进行，从数据衍生的洞察中构建内容，而不是基于高层主题或部分。为此，我们提出了AnalysisBank，它将专家报告提炼为可重用的分析库，每个分析都配对一个数据信号、一个分析动作以及其来源的专家范围。在推理时，AnalysisBank将输入信号与库中的条目匹配，并应用检索到的动作来编写报告。对550份专家报告提炼出的分析的研究显示，47-52种信号类型呈现重尾分布，涵盖13种动作类型。在四个大型语言模型（LLM）基础上进行的两个财务基准测试中，AnalysisBank将新颖、基于数据的洞察的比例提高了1.7-3.7倍，超越了结构层面的基准。转向科学写作表明，这一区分超越了金融领域。代码和提炼的分析库可在https://github.com/yajingyang/AnalysisBank获取。

View on arXiv Download PDF AI Translation

cs.AI / 77 / 2609.00823

Polished but Unresolved: Identifying Late-Stage Pressure States in Long-Horizon Tool-Use Agents

抛光但未解决：识别长时间工具使用代理的晚期压力状态

Chen, Haoyang, Liu, Yi, Shao, Jianzhi, Xu, Xiaozhou, Sun, Zhe, Hu, Wei

Abstract

Long-horizon tool-use agents need not only to search and plan, but also to decide when to finalize. We study late-stage pressure states, in which an agent is biased toward submitting a final answer that appears complete and polished while key constraints remain unresolved. We first train a linear probe to show that this pressure state is identifiable from the agent's hidden states. Then, we use activation interventions along this pressure direction and find that shifting the hidden states changes both the pressure score and whether the agent continues tool use or submits early. Through controlled context manipulations, we further see that the pressure is mitigated by constraint clarity and action mapping. Based on these findings, we propose Probe-Sensed Pressure Relief (PSPR), a plugin that applies lightweight pressure relief direction under moderate pressure and moves to structured organization under high pressure risk. Experiments on multiple long-horizon benchmarks show that our method consistently strengthens existing agent methods.

Chinese Translation

长时间工具使用代理不仅需要搜索和规划，还需要决定何时最终确定答案。我们研究晚期压力状态，在这种状态下，代理倾向于提交一个看似完整且抛光的最终答案，而关键约束仍未解决。我们首先训练一个线性探针，以表明这一压力状态可以从代理的隐藏状态中识别出来。然后，我们沿着这一压力方向进行激活干预，发现改变隐藏状态会同时改变压力评分以及代理是否继续使用工具或提前提交答案。通过控制上下文的操控，我们进一步观察到，约束的清晰度和行动映射可以减轻压力。基于这些发现，我们提出了探针感知压力缓解（Probe-Sensed Pressure Relief, PSPR），这是一种在中等压力下应用轻量级压力缓解方向，并在高压力风险下转向结构化组织的插件。在多个长时间基准测试中的实验表明，我们的方法始终增强了现有代理方法的效果。

View on arXiv Download PDF AI Translation

cs.AI / 78 / 2609.00831

FLaG: Frequency-Domain Latent-attention Gated Pooling for Token Aggregation

FLaG：用于标记聚合的频域潜在注意力门控池化

Li, Kewei, Zhang, Rongying, Wang, Xueli, Gong, Xiwen, Wang, Zhongjian, Zhao, Qiuchen, Huang, Lan, Zhang, Ruochi, Zhou, Fengfeng

Abstract

Token aggregation converts token-level representations into fixed-dimensional sample representations, but most pooling methods operate only in the original token space. We introduce Frequency-Domain Latent-attention Gated Pooling (FLaG), a plug-in aggregation module that re-expresses encoder outputs in the Fourier domain before final pooling. FLaG represents the nonredundant rFFT spectrum through concatenated real and imaginary components, summarizes spectral tokens with learnable latent queries, derives a sample-conditioned channel gate, and reconstructs modulated token representations for downstream aggregation. We evaluate the same architecture across ESM2-based antimicrobial peptide (AMP) activity prediction, ResNet18 image classification on CIFAR-10 and CIFAR-100, and three RoBERTa-based language tasks. FLaG achieves the best macro-averaged Spearman correlation coefficient, RMSE, and Recall@50 across four AMP backbone-species settings and the highest top-1 accuracy on CIFAR 10. It also achieves the best mean results on five of seven language metrics, although mean pooling remains strongest on STSBenchmark. AMP-side mechanistic analyses reveal low-frequency prediction sensitivity across most encoder layers, with increased relative high-frequency sensitivity in the final layer, and pronounced peptide-specific positional responses. The residual gate broadly amplifies spectral channels while preserving the low-frequency-dominated energy profile, whereas latent cross-attention exhibits sample- and species-specific spectral allocation. Overall, FLaG provides a transferable frequency-domain aggregation bias across protein, visual, and textual representations, with benefits that depend on the backbone and downstream task. Supplementary materials, source code, and data are available at https://www.healthinformaticslab.org/supp/ and https://github.com/Kewei2023/AMPCliff/tree/FLaG.

Chinese Translation

标记聚合将标记级表示转换为固定维度的样本表示，但大多数池化方法仅在原始标记空间中操作。我们提出了频域潜在注意力门控池化（FLaG），这是一种插件聚合模块，在最终池化之前在傅里叶域中重新表达编码器输出。FLaG通过连接的实部和虚部表示非冗余的 rFFT 频谱，使用可学习的潜在查询总结频谱标记，推导样本条件的通道门，并重建调制的标记表示以进行下游聚合。我们在基于 ESM2 的抗菌肽（AMP）活性预测、CIFAR-10 和 CIFAR-100 上的 ResNet18 图像分类以及三个基于 RoBERTa 的语言任务中评估了相同的架构。FLaG 在四个 AMP 主干物种设置中实现了最佳的宏平均斯皮尔曼相关系数、均方根误差（RMSE）和 Recall@50，并在 CIFAR 10 上获得了最高的 top-1 准确率。它在七个语言指标中的五个上也取得了最佳平均结果，尽管在 STSBenchmark 上均值池化仍然最强。AMP 侧的机制分析揭示了大多数编码器层的低频预测敏感性，在最后一层中相对高频敏感性增加，并且表现出明显的肽特异性位置响应。残差门广泛放大频谱通道，同时保持低频主导的能量特征，而潜在的交叉注意力则表现出样本和物种特异的频谱分配。总体而言，FLaG 提供了一种可转移的频域聚合偏差，适用于蛋白质、视觉和文本表示，其益处依赖于主干和下游任务。补充材料、源代码和数据可在 https://www.healthinformaticslab.org/supp/ 和 https://github.com/Kewei2023/AMPCliff/tree/FLaG 获取。

View on arXiv Download PDF AI Translation

cs.AI / 79 / 2609.00845

Towards Generalizable Visually Grounded Exploration of Household Devices

面向可推广的家庭设备视觉基础探索

Zheng, Linhao, Liu, Zeming, Chen, Wangke, Zeng, Li, Che, Wanxiang, Huang, Heyan, Guo, Yuhang

Abstract

Recent advancements in Vision-Language Models (VLMs) have demonstrated impressive capabilities in static visual recognition and high-level semantic reasoning. However, current embodied exploration paradigms still heavily rely on imitation learning from human-annotated trajectories, which severely limits agents' generalization ability. The key bottleneck of realizing general autonomous embodied agents lies in Generalizable Visually Grounded Exploration: the ability to operate novel devices without manuals or specific training by actively grounding abstract world knowledge into fine-grained visual affordances. Yet, existing benchmarks fail to evaluate this capability: they generally rely on explicit documents and annotated trajectories, neglecting the dynamic Hypothesis-Interaction-Refinement process essential for functional device operation. To bridge this gap, we introduce VGEBench, a comprehensive benchmark designed to evaluate the generalizable visually grounded exploration capabilities of VLMs. Unlike static datasets, we construct a Logic-Driven State Machine framework. This framework simulates multi-turn interaction loops, compelling agents to achieve goals by active visual perception and feedback-driven correction. Experimental results demonstrate that existing VLMs face significant challenges in translating semantic knowledge into physical execution and maintaining long-horizon state tracking.

Chinese Translation

近期视觉-语言模型（Vision-Language Models, VLMs）的进展展示了在静态视觉识别和高层次语义推理方面的卓越能力。然而，当前的具身探索范式仍然严重依赖于从人类标注轨迹中进行模仿学习，这大大限制了智能体的泛化能力。实现通用自主具身智能体的关键瓶颈在于可推广的视觉基础探索：即在没有手册或特定训练的情况下，能够操作新设备的能力，通过主动将抽象的世界知识与细粒度的视觉可用性相结合。然而，现有基准测试未能评估这一能力：它们通常依赖于显式文档和标注轨迹，忽视了功能设备操作所必需的动态假设-交互-精炼过程。为了解决这一问题，我们引入了VGEBench，这是一个全面的基准，旨在评估VLMs的可推广视觉基础探索能力。与静态数据集不同，我们构建了一个逻辑驱动状态机框架。该框架模拟多轮交互循环，迫使智能体通过主动视觉感知和反馈驱动的修正来实现目标。实验结果表明，现有的VLMs在将语义知识转化为物理执行和维持长期状态跟踪方面面临重大挑战。

View on arXiv Download PDF AI Translation

cs.AI / 80 / 2609.00858

Verifiable Disaster Storylines and Causal Knowledge Graphs: A Citation-Grounded Pipeline from Heterogeneous Humanitarian Sources

可验证的灾难故事线和因果知识图谱：来自异构人道主义来源的引用基础管道

Decostanzi, Ivan, Ronco, Michele, Consoli, Sergio, Corbane, Christina, Bertolini, Lorenzo, Biazzo, Indaco, Mihaila, Daria, Garcia-Herranz, Manuel, Schwebel, Felix, Mejova, Yelena, Kalimeri, Kyriaki

Abstract

Effective humanitarian response depends on the rapid synthesis of heterogeneous, high-volume information sources - a task that routinely exceeds human analytical capacity in the critical early hours of a crisis. We present a pipeline that combines structured disaster records from EM-DAT with unstructured documents from ReliefWeb and the European Media Monitor (EMM) to produce source-grounded disaster storylines and causal knowledge graphs supporting situational awareness for responders and analysts. Using Retrieval-Augmented Generation, the pipeline extracts structured storylines - tabular event profiles covering 17 fields, from severity and key drivers to child-sensitive impact indicators - and constructs causal knowledge graphs where each node and edge is enriched with citation-grounded explanatory narratives, enabling full traceability back to primary sources. We evaluate the system on three diverse crisis use cases through a human evaluation involving 9 domain expert and 9 non-expert evaluators. Results confirm high retrieval precision, strong faithfulness of extracted causal relations, and a clear expert preference for citation-grounded components over ungrounded alternatives. The pipeline is designed to scale to the full EM-DAT catalogue, with the goal of publicly releasing a narrative-enriched version of the database.

Chinese Translation

有效的人道主义响应依赖于对异构、高容量信息源的快速综合——这一任务在危机的关键早期阶段通常超出了人类的分析能力。我们提出了一种管道，将来自 EM-DAT 的结构化灾难记录与来自 ReliefWeb 和欧洲媒体监测（EMM）的非结构化文档相结合，以生成基于来源的灾难故事线和因果知识图谱，支持响应者和分析师的情境意识。通过检索增强生成（Retrieval-Augmented Generation），该管道提取结构化故事线——涵盖严重程度、关键驱动因素到儿童敏感影响指标等 17 个领域的表格事件概况——并构建因果知识图谱，其中每个节点和边缘都附加了基于引用的解释性叙述，实现了对原始来源的完全可追溯性。我们通过涉及 9 位领域专家和 9 位非专家评估者的人类评估，对该系统在三个不同的危机使用案例中进行了评估。结果确认了高检索精度、提取的因果关系的强忠实度，以及专家对基于引用的组件相较于无基础替代方案的明显偏好。该管道旨在扩展到完整的 EM-DAT 目录，目标是公开发布一个叙述丰富的数据库版本。

View on arXiv Download PDF AI Translation

cs.AI / 81 / 2609.00859

Reinforcement Learning Enhanced LLM Agents for Complex Vehicle Routing Problems

强化学习增强的LLM代理用于复杂车辆路径规划问题

Chen, Yi, Yu, Zikang, Wang, Jiahai, Chen, Jinbiao, Zhou, Jianpeng, Zhang, Zizhen

Abstract

Vehicle Routing Problems (VRPs) are fundamental combinatorial optimization problems with widespread applications in various scenarios. The advanced optimization solvers can effectively solve such problems. However, modeling complex VRP variants for solvers often requires substantial domain expertise, which limits the accessibility of advanced optimization technologies. In this paper, we propose Reinforcement Learning Enhanced LLMAgents(RLEA), a multi-agent framework designed to automate the modeling of complex VRPs. RLEA introduces a lightweight neural Planner trained with Soft Q-learning to efficiently orchestrate the actions of LLM-based agents. In addition, we equip the system with an evolutionary memory module and retrieval-augmented generation, enabling the agent to leverage both accumulated experience and external solver knowledge during program generation and refinement for solving VRPs. We evaluated 48 distinct VRP variants across various solvers. The experimental results demonstrate that RLEA outperforms the previous state-of-the-ar method, achieving a 16.67% higher success rate while significantly reducing runtime errors. These results validate that integrating reinforcement learning with LLM-based reasoning is highly effective for automated optimization modeling. The appendix is available at: https://doi.org/10.5281/zenodo.19134435.

Chinese Translation

车辆路径规划问题（VRPs）是基本的组合优化问题，在各种场景中具有广泛的应用。先进的优化求解器能够有效地解决此类问题。然而，为求解器建模复杂的VRP变体通常需要大量的领域专业知识，这限制了先进优化技术的可及性。本文提出了强化学习增强的LLM代理（Reinforcement Learning Enhanced LLM Agents, RLEA），这是一个旨在自动化复杂VRP建模的多代理框架。RLEA引入了一种轻量级神经规划器，该规划器通过软Q学习（Soft Q-learning）进行训练，以高效协调基于LLM的代理的行动。此外，我们为系统配备了一个进化记忆模块和检索增强生成（retrieval-augmented generation），使代理能够在程序生成和优化过程中利用累积经验和外部求解器知识。我们在多种求解器上评估了48种不同的VRP变体。实验结果表明，RLEA的成功率比之前的最先进方法提高了16.67%，同时显著减少了运行时错误。这些结果验证了将强化学习与基于LLM的推理相结合在自动化优化建模中的高效性。附录可在以下链接获取：https://doi.org/10.5281/zenodo.19134435。

View on arXiv Download PDF AI Translation

cs.AI / 82 / 2609.00874

Beyond the Clock: Measuring the Value of Adaptive Revision

超越时钟：测量自适应修订的价值

Chadha, Ayushi

Abstract

As agentic systems become compound systems, increasingly important decisions move above task execution itself: when should a higher-level controller preserve the strategy guiding another process, and when should it revise it? We study this meta-level control problem in a hierarchical latent reasoner whose manager can retain or replace a commitment governing lower-level computation. Across three precommitted training seeds, learned revision timing produces qualitatively different policies, ranging from an almost deterministic early clock to substantially more state conditioned schedule distributions, yet none outperforms the best forced timing policy evaluated on the same frozen checkpoint. This separates state dependence from decision value: a controller can vary its actions with internal state without turning that variation into a reproducible task-performance benefit. A deeper intervention study on the original checkpoint shows that timing itself is consequential and order-sensitive, while exhaustive enumeration reveals that a strong fixed schedule captures most of the measurable value available from timing at this decision budget. Counterfactual PERSIST/REPLAN diagnostics further show why score-level evidence can be misleading when predictability is dominated by decision position rather than within-position discrimination. Together, these results argue that learned meta-level control should be evaluated along three separate axes: whether its score depends on state, whether that dependence changes realized behavior, and whether those changes capture outcome value beyond a strong non-adaptive policy.

Chinese Translation

随着自主系统变得更加复杂，越来越重要的决策超越了任务执行本身：何时应由高层控制器保留指导其他过程的策略，何时应对其进行修订？我们研究了这一元级控制问题，聚焦于一个层次化的潜在推理器，其管理者可以保留或替换支配低层计算的承诺。在三个预设的训练种子中，学习到的修订时机产生了质上不同的策略，从几乎确定性的早期时钟到显著更具状态条件的调度分布，但没有一种策略在同一冻结检查点上超越最佳的强制时机策略。这将状态依赖性与决策价值区分开来：控制器可以根据内部状态变化其行为，而不将这种变化转化为可重复的任务表现收益。对原始检查点的更深入干预研究表明，时机本身是重要的且对顺序敏感，而全面枚举则揭示出强固定调度捕获了在这一决策预算下可从时机获得的大部分可测量价值。反事实的 PERSIST/REPLAN 诊断进一步表明，当可预测性受决策位置主导而非位置内区分时，得分水平证据可能具有误导性。综合来看，这些结果表明，学习到的元级控制应沿着三个独立的维度进行评估：其得分是否依赖于状态，这种依赖是否改变了实现的行为，以及这些变化是否捕获了超越强非适应性政策的结果价值。

View on arXiv Download PDF AI Translation

cs.AI / 83 / 2609.00875

FractalNet-Based Heterogeneous Federated Learning for Orbital Edge Intelligence in Satellite Mega-Constellations: A Wildfire Case Study

基于分形网络的异构联邦学习在卫星巨型星座中的轨道边缘智能应用：以野火为案例研究

Puppala, Sai, Sinha, Koushik

Abstract

Satellite mega-constellations are emerging as large-scale sensing, communication, and computation fabrics, yet their learning architectures remain largely inherited from terrestrial federated learning and ground-centric mission operations--- ill-suited to satellites that differ by orders of magnitude in Size, Weight, Power, and Cost (SWAP-C), radiation tolerance, link availability, and propagation delay. We propose a heterogeneous federated learning method based on the FractalNet architecture for orbital edge intelligence. We formalize contact-window-constrained, depth-heterogeneous federated optimization and introduce a distributed path scheduler that assigns model depth as a function of SWAP-C constraints, predicted inter-satellite contacts, and training statistics. To reduce message overhead and energy consumption, each tier pools updates periodically rather than at every contact opportunity, and a three-tier agentic control plane governs in-space scheduling, anomaly escalation, and policy-governed autonomy. As a case study, we apply the framework to wildfire detection, where each orbital shell naturally learns a different semantic level of situational awareness: pixel-scale thermal anomalies at low Earth orbit (LEO), regional fire-front dynamics at medium Earth orbit (MEO), and larger-scale risk propagation at geostationary or high Earth orbit (GEO/HEO). Experiments on simulated mega-constellations validate the approach across convergence, communication efficiency, energy adaptation, scheduled-pooling savings, robustness, and latency.

Chinese Translation

卫星巨型星座作为大规模传感、通信和计算的网络正在兴起，但其学习架构在很大程度上仍然继承自地面联邦学习和以地面为中心的任务操作，这与在尺寸、重量、功耗和成本（SWAP-C）、辐射容忍度、链路可用性和传播延迟等方面相差几个数量级的卫星并不匹配。我们提出了一种基于分形网络（FractalNet）架构的异构联邦学习方法，以实现轨道边缘智能。我们形式化了受接触窗口约束的深度异构联邦优化，并引入了一种分布式路径调度器，根据SWAP-C约束、预测的卫星间接触和训练统计信息来分配模型深度。为了减少消息开销和能耗，每个层级定期汇总更新，而不是在每次接触机会时都进行更新，并且一个三层代理控制平面负责在轨调度、异常升级和政策驱动的自主控制。作为案例研究，我们将该框架应用于野火检测，其中每个轨道层自然学习不同语义层次的态势感知：低地球轨道（LEO）下的像素级热异常、中地球轨道（MEO）下的区域火前动态，以及静止轨道或高地球轨道（GEO/HEO）下的大规模风险传播。对模拟巨型星座的实验验证了该方法在收敛性、通信效率、能量适应性、调度汇聚节省、鲁棒性和延迟等方面的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 84 / 2609.00879

Towards reliable multimodal disaster severity assessment through preference optimization and explainable vision-language reasoning

通过偏好优化和可解释的视觉-语言推理实现可靠的多模态灾害严重性评估

Zhang, Yuanjun, Shaik, Fuzel Ahamed, Acharjee, Suvojit, Khalid, Fahad, Oussalah, Mourad

Abstract

Reliable disaster damage assessment requires models that provide both accurate predictions and transparent explanations. However, existing multimodal approaches are limited by scarce annotated data and insufficient evaluation of reasoning quality. This study proposes a two-stage training framework that integrates Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) within a unified data construction pipeline. From a single Human-in-the-Loop (HITL) annotation workflow, two complementary datasets are derived, namely ReasoningSet, which contains validated rationales for SFT, and PreferenceSet, which comprises paired rationales for DPO-based alignment. The framework evaluates both classification performance and explanation quality using automatic metrics, model-based scoring, and human ranking. Experimental results show that SFT improves accuracy from 73.64% to 78.29% and increases Macro-F1 by 29% compared to the baseline, while explanation quality improves by approximately 25%. Subsequent DPO alignment further enhances interpretability on the PreferenceSet. Cross-model validation on InternVL-3-8B and LLaVA-1.5-7B demonstrates the robustness and generalizability of the approach. The proposed framework improves detection of underrepresented mild damage cases, reduces high-risk misclassifications, and strengthens alignment between model reasoning and human judgment. Overall, it provides a reproducible pathway to develop reliable multimodal systems that deliver auditable, actionable disaster insights for emergency management.

Chinese Translation

可靠的灾害损害评估需要提供准确预测和透明解释的模型。然而，现有的多模态方法受到标注数据稀缺和推理质量评估不足的限制。本研究提出了一种两阶段训练框架，将监督微调（Supervised Fine-Tuning, SFT）和直接偏好优化（Direct Preference Optimization, DPO）整合在一个统一的数据构建流程中。从单一的人机交互（Human-in-the-Loop, HITL）标注工作流中，衍生出两个互补的数据集，即ReasoningSet，包含用于SFT的验证推理，以及PreferenceSet，包含用于基于DPO的对齐的成对推理。该框架使用自动指标、基于模型的评分和人工排名评估分类性能和解释质量。实验结果表明，与基线相比，SFT将准确率从73.64%提高到78.29%，宏观F1值提高了29%，而解释质量提高了约25%。后续的DPO对齐进一步增强了PreferenceSet上的可解释性。在InternVL-3-8B和LLaVA-1.5-7B上的跨模型验证证明了该方法的鲁棒性和普适性。所提出的框架提高了对代表性不足的轻微损害案例的检测，减少了高风险的错误分类，并加强了模型推理与人类判断之间的对齐。总体而言，它提供了一条可重复的路径，以开发可靠的多模态系统，为应急管理提供可审计、可操作的灾害洞察。

View on arXiv Download PDF AI Translation

cs.AI / 85 / 2609.00885

Denoising Diffusion Generative Models Secretly Calculate Attentions

去噪扩散生成模型秘密计算注意力

Haddadi, Farzan, Monfared, Leila, Rezaii, Ebrahim, Malek-Mohammadi, Mohammadreza, Zakalvand, Pejman, Mokhtari, Narges

Abstract

Denoising diffusion models are the dominant architecture for image generation, whereas most natural language generation and modeling are primarily handled by well-known transformer architectures employing attention mechanism. Here, we show that diffusion models also inherently use an attention mechanism very similar to that of transformers. Therefore, attention emerges as a universal machine learning principle, based on a general training objective. We also show similarities in basic functional principle of auto-encoders and attention-based models. These equivalences allows us to interchange these designs based on practical requirements. As an example, we can reformulate the diffusion framework to reduce the lengthy training process and computation-intensive image generation. Using this approach, a simplified algorithm is proposed for image generation which is based on attention mechanism. Results show that the attention-based implementation achieves comparable performance with significantly less effort and computational resources.

Chinese Translation

去噪扩散模型是图像生成的主流架构，而大多数自然语言生成和建模主要由采用注意力机制的知名变换器架构处理。在这里，我们展示了扩散模型也固有地使用与变换器非常相似的注意力机制。因此，注意力作为一种通用的机器学习原则出现，基于一般的训练目标。我们还展示了自编码器和基于注意力的模型在基本功能原理上的相似性。这些等价性使我们能够根据实际需求互换这些设计。作为一个例子，我们可以重新构建扩散框架，以减少冗长的训练过程和计算密集型的图像生成。采用这种方法，提出了一种基于注意力机制的简化算法用于图像生成。结果表明，基于注意力的实现以显著较少的努力和计算资源达到了可比的性能。

View on arXiv Download PDF AI Translation

cs.AI / 86 / 2609.00891

CacheBridge: Efficient Cross-Model KV Cache Transfer

CacheBridge：高效的跨模型 KV 缓存转移

Qu, Xingyu, Lu, Siyuan, Chen, Zhiyu, Wang, Sheng, Lin, Tao

Abstract

Sharing context between LLMs in a multi-model system requires the receiving model to prefill the shared prefix because KV caches are model-specific. Recent closed-form cross-model KV transfer, hereafter Full-Head Mapping, avoids this replay by fitting a training-free affine mapper from source to target caches. However, its full-head design maps each target KV head from every source KV head in the selected layers, making transfer quality sensitive to architectural differences and causing mapper storage and application cost to grow with layer support. To this end, we introduce CacheBridge, which co-designs architecture-indexed mapper support, attention-aligned calibration, and bounded mapper construction while retaining a closed-form affine interface for online deployment. CacheBridge restricts each target head to a matched source head, weights reconstruction errors by causal attention sensitivity, and uses a fused GPU kernel to construct weighted sufficient statistics without materializing full observation tensors. Across three transfer directions, CacheBridge recovers the two Ministral 3 transfer directions where Full-Head Mapping loses substantial accuracy while preserving 99.83\% mean target retention on Qwen3. On Qwen3 $14\mathrm{B}\to32\mathrm{B}$, it reduces mapper storage by $8\times$, accelerates application by up to $3.0\times$, matches \fullhead with one tenth of the calibration data, and reduces 500-sequence construction from 92.63 to 8.63 seconds ($10.7\times$).

Chinese Translation

在多模型系统中，LLMs 之间共享上下文需要接收模型预填充共享前缀，因为 KV 缓存是模型特定的。最近的闭式跨模型 KV 转移（以下简称为 Full-Head Mapping）通过拟合一个无训练的仿射映射器从源缓存到目标缓存，避免了这一重放。然而，其全头设计将每个目标 KV 头映射到所选层中的每个源 KV 头，使得转移质量对架构差异敏感，并导致映射器的存储和应用成本随着层支持的增加而增长。为此，我们引入了 CacheBridge，它共同设计了架构索引映射器支持、注意力对齐校准和有界映射器构建，同时保留了用于在线部署的闭式仿射接口。CacheBridge 将每个目标头限制为匹配的源头，通过因果注意力敏感性加权重构误差，并使用融合的 GPU 内核构建加权充分统计量，而无需实现完整的观察张量。在三个转移方向上，CacheBridge 恢复了两个 Ministral 3 转移方向，其中 Full-Head Mapping 丧失了大量准确性，同时在 Qwen3 上保持了 99.83\% 的平均目标保留。在 Qwen3 $14 ext{B} o32 ext{B}$ 的情况下，它将映射器存储减少了 $8 imes$，将应用加速提高至 $3.0 imes$，用十分之一的校准数据匹配了 ullhead，并将 500 序列构建时间从 92.63 秒减少到 8.63 秒（$10.7 imes$）。

View on arXiv Download PDF AI Translation

cs.AI / 87 / 2609.00892

CARE: Contrastive Anchor-based Rubric Evolution for Large Language Model Post-Training

CARE：基于对比锚点的评分标准演化用于大型语言模型后训练

Li, Siyuan, Song, Xinxin, Ruinian, Chen, Fan, Jingjing, Xiao, Tingxiong, Hu, Yangen, Zeng, Ke, Suo, Jinli

Abstract

Rubric-based reinforcement learning decomposes open-ended instructions into prompt-specific, flexible rubrics, making it better suited than reinforcement learning with verifiable rewards for post-training LLMs on open-ended tasks. However, static rubrics are inevitably hacked as the policy evolves, and existing dynamic approaches introduce new problems: undirected rubric extraction, unreliable hack detection, and unbounded rubric proliferation. We propose $\textbf{CARE}$ ($\textbf{C}$ontrastive $\textbf{A}$nchor-based $\textbf{R}$ubric $\textbf{E}$volution), which grounds every rubric evolution step in a high-quality anchor response generated by a frontier model conditioned on the prompt and its rubrics. At each training step, CARE contrasts the highest-scoring rollout against the anchor, enabling two complementary mechanisms: an Adaptive branch that reactively repairs reward misspecification; and a Chase branch that proactively converts frontier-level quality gaps into sharper rubrics. Together, the two branches $\textbf{maintain discriminative accuracy in the high-reward region}$---the precise region where reward over-optimization mostly originates. Experiments on WildChecklist-9K with Qwen2.5-7B-Base and Qwen2.5-7B-Instruct show that CARE achieves state-of-the-art performance on Arena-Hard-2.0, InfoBench, and FollowBench, and is the $\textbf{only}$ method whose win rate against GPT-4.1 anchor responses shows sustained improvement throughout 300 training steps; additional results on Llama-3.1-8B-Instruct and Qwen3-8B further indicate that CARE generalizes across model families.

Chinese Translation

基于评分标准的强化学习将开放式指令分解为特定于提示的灵活评分标准，使其在开放式任务中比具有可验证奖励的强化学习更适合后训练的大型语言模型（LLMs）。然而，随着策略的演变，静态评分标准不可避免地会被破解，而现有的动态方法则引入了新的问题：无方向的评分标准提取、不可靠的破解检测和无界的评分标准扩散。我们提出了$ extbf{CARE}$（$ extbf{C}$ontrastive $ extbf{A}$nchor-based $ extbf{R}$ubric $ extbf{E}$volution），该方法将每一步评分标准的演化基于由前沿模型生成的高质量锚点响应，这些响应是根据提示及其评分标准生成的。在每个训练步骤中，CARE将得分最高的回滚与锚点进行对比，从而实现两种互补机制：一个自适应分支，反应性地修复奖励错误指定；以及一个追逐分支，主动将前沿水平的质量差距转化为更精确的评分标准。两个分支共同$ extbf{维持高奖励区域的区分准确性}$——这是奖励过度优化主要来源的精确区域。在使用Qwen2.5-7B-Base和Qwen2.5-7B-Instruct进行的WildChecklist-9K实验中，结果表明CARE在Arena-Hard-2.0、InfoBench和FollowBench上达到了最先进的性能，并且是$ extbf{唯一}$一种在300个训练步骤中与GPT-4.1锚点响应的胜率持续改善的方法；在Llama-3.1-8B-Instruct和Qwen3-8B上的额外结果进一步表明CARE在模型家族之间具有良好的泛化能力。

View on arXiv Download PDF AI Translation

cs.AI / 88 / 2609.00904

In-Context Neurofeedback: Can LLMs Control Their Internal Representations through Privileged Access?

上下文神经反馈：大型语言模型能否通过特权访问控制其内部表征？

Aoki, Koshiro, Takatsuki, Ryota, Minegishi, Gouki, Haruki, Yusuke, Kawahara, Daisuke

Abstract

Whether large language models (LLMs) can control their own internal representations matters for both machine metacognition and AI safety. A recent study applied neurofeedback to LLMs and claimed that they can control their internal representations. However, the reported control may rely on superficial mechanisms rather than genuine internal access because the control targets in that study are not privileged, meaning that a third party can infer them from the prompt. We redesign the neurofeedback paradigm for LLMs so that the control target satisfies the privileged access requirement, which is closer to neurofeedback experiments in human cognitive neuroscience. Under this stricter setting, the models do not demonstrate reliable control over privileged internal representations, suggesting that previously reported control cannot exclude the possibility that it relies on superficial mechanisms. Our results indicate that rigorous assessments of metacognition in LLMs require evaluation methods that demand privileged access.

Chinese Translation

大型语言模型（LLMs）是否能够控制自身的内部表征对于机器元认知和人工智能安全具有重要意义。最近的一项研究将神经反馈应用于LLMs，并声称它们能够控制其内部表征。然而，该研究报告的控制可能依赖于表面机制而非真正的内部访问，因为该研究中的控制目标并不具备特权，这意味着第三方可以从提示中推断出这些目标。我们重新设计了LLMs的神经反馈范式，使控制目标满足特权访问的要求，这更接近于人类认知神经科学中的神经反馈实验。在这一更严格的设置下，模型未能可靠地控制特权内部表征，这表明之前报告的控制不能排除其依赖于表面机制的可能性。我们的结果表明，对LLMs的元认知进行严格评估需要采用要求特权访问的评估方法。

View on arXiv Download PDF AI Translation

cs.AI / 89 / 2609.00918

RPCBench: A Benchmark for Proactive Premise Critique in LLM-based Recommendation

RPCBench：用于基于大语言模型的推荐系统中的主动前提批评基准测试

Chen, Zhongru, Wu, Yuan, Chang, Yi

Abstract

Large language models are increasingly used as interactive recommender assistants. Their evaluation should therefore go beyond plausible item recommendation and test whether they can recognize flawed recommendation requests. Existing recommender benchmarks mainly assess ranking, generation, or preference satisfaction, while existing error-detection benchmarks are usually not grounded in recommendation-specific user and candidate evidence. To address this gap, we introduce RPCBench, a benchmark for evaluating Recommender-Premise Critique: the ability to detect, diagnose, and properly handle faulty premises in natural-language recommendation requests. RPCBench contains evidence-grounded test instances from five recommendation domains and covers ten types of premise failures. Each instance provides a visible recommendation context and a corrupted user query. We further design a fine-grained evaluation framework that measures proactive detection, error localization, post-detection handling strategy, and evidence faithfulness. Through a systematic evaluation of 11 LLMs, we find that proactive detection is the main bottleneck in Recommender-Premise Critique, and models perform worst on underspecified-premise errors. We also observe that target-critical information density matters more than redundant evidence, and that longer reasoning does not monotonically improve critique quality: performance peaks at intermediate reasoning length, while overly long reasoning is accompanied by an overthinking penalty. The code is available at https://github.com/ZhongruChen/RPCBench.

Chinese Translation

大型语言模型越来越多地被用作交互式推荐助手。因此，它们的评估应该超越合理的项目推荐，测试它们是否能够识别存在缺陷的推荐请求。现有的推荐基准主要评估排名、生成或偏好满足，而现有的错误检测基准通常并未基于推荐特定的用户和候选证据。为了解决这一空白，我们引入了RPCBench，这是一个用于评估推荐前提批评（Recommender-Premise Critique）的基准：即检测、诊断和妥善处理自然语言推荐请求中错误前提的能力。RPCBench包含来自五个推荐领域的基于证据的测试实例，并涵盖十种前提失败类型。每个实例提供了可见的推荐上下文和一个损坏的用户查询。我们进一步设计了一个细粒度的评估框架，衡量主动检测、错误定位、后检测处理策略和证据的可信度。通过对11个大型语言模型的系统评估，我们发现主动检测是推荐前提批评的主要瓶颈，模型在不明确前提错误上的表现最差。我们还观察到，目标关键性信息的密度比冗余证据更为重要，并且更长的推理并不单调地提高批评质量：性能在中等推理长度时达到峰值，而过长的推理则伴随着过度思考的惩罚。代码可在https://github.com/ZhongruChen/RPCBench获取。

View on arXiv Download PDF AI Translation

cs.AI / 90 / 2609.00921

VIBE-Bench: Evaluating Personalized Large Language Models When Profiles Don't Mean Preferences

VIBE-Bench：在个人资料不等于偏好的情况下评估个性化大型语言模型

Jiang, Yiwen, Deng, Yang, Fong, Stephanie, Wang, Zimu, Shen, Yaling, Feng, Wei, Yang, Hongxi, Zhao, Xiangyu, Xu, Zhongxing, Mehta, Deval, Cheng, Xuelian, Ge, Zongyuan

Abstract

Personalized Large Language Models (PLLMs) aim to tailor responses to individual users, where a central challenge is preference reasoning: inferring query-relevant preferences from user-related history. Existing benchmarks, however, largely assume that such preference can be retrieved from semantically related history. We study an underexplored but practically important regime, profile-preference conceptual misalignment (PRCM), where observable profile cues and query-specific preferences lie in different concept spaces, making semantic retrieval inconsistent for personalization. We introduce VIBE-Bench, a benchmark with two psychology-grounded tasks, 3,504 personas and 12,239 dialogues, including a manually verified gold test set, and requires cross-concept preference reasoning beyond surface semantic overlap. Experiments with several personalization methods show that current PLLMs largely rely on shallow semantic correlations and fail to acquire robust cross-concept mappings. These findings establish PRCM as a distinct failure regime in PLLMs and position VIBE-Bench as a focused testbed for advancing preference reasoning beyond semantic matching.

Chinese Translation

个性化大型语言模型（PLLMs）旨在根据个体用户量身定制响应，其中一个核心挑战是偏好推理：从用户相关历史中推断与查询相关的偏好。然而，现有基准大多假设可以从语义相关的历史中检索到这样的偏好。我们研究了一个未被充分探索但在实践中重要的领域，即个人资料-偏好概念不一致（PRCM），在该领域中，可观察的个人资料线索和查询特定的偏好位于不同的概念空间，使得语义检索在个性化方面不一致。我们引入了VIBE-Bench，这是一个包含两个基于心理学的任务、3504个角色和12239个对话的基准，其中包括一个经过人工验证的黄金测试集，并要求超越表面语义重叠进行跨概念偏好推理。对几种个性化方法的实验表明，目前的PLLMs在很大程度上依赖于浅层语义关联，未能获得稳健的跨概念映射。这些发现确立了PRCM作为PLLMs中的一个独特失败领域，并将VIBE-Bench定位为一个专注于推动偏好推理超越语义匹配的测试平台。

View on arXiv Download PDF AI Translation

cs.AI / 91 / 2609.00961

Few-Shot Out of Domain Intent Detection with Covariance Corrected Mahalanobis Distance

基于协方差修正的马哈拉诺比斯距离的少样本域外意图检测

Talur, Jayasimha, Smirnov, Oleg, Missault, Paul

Abstract

Conversational agents like chatbots and voice assistants are trained to understand and respond to user intents. On encountering an utterance with an intent different from the ones they have been trained on, these agents are expected to classify the intent as `unknown' or `out of domain'. This problem is known as out of domain (OOD) intent detection. Podolskiy et al. (2021), showed that Mahalanobis distance can be used effectively for identifying OOD intents, outperforming competing approaches. However, their method fails to outperform the baselines in the practically important few-shot setting. In this paper we analyze the reason for low performance and propose a covariance corrected Mahalanobis distance for detecting out-of-domain intents.

Chinese Translation

对话代理如聊天机器人和语音助手被训练以理解和响应用户意图。当遇到与其训练过的意图不同的发言时，这些代理被期望将该意图分类为‘未知’或‘域外’。这个问题被称为域外（OOD）意图检测。Podolskiy 等（2021）表明，马哈拉诺比斯距离可以有效地用于识别域外意图，且优于其他竞争方法。然而，他们的方法在实际重要的少样本设置中未能超越基线。本文分析了低性能的原因，并提出了一种用于检测域外意图的协方差修正马哈拉诺比斯距离。

View on arXiv Download PDF AI Translation

cs.AI / 92 / 2609.00967

CoBRA: Learning Tool-Use Boundaries via Counterfactual Margins

CoBRA：通过反事实边界学习工具使用边界

Zou, Wenhao, Liu, Xianglong, Bi, Wendong, Wang, Hanjie, Zhao, Simin, Zhi, Gong

Abstract

As large language models increasingly act through external tools, deciding when to call a tool has become a central problem alongside deciding how to use it. Unnecessary tool calls introduce latency, cost, retrieval noise, and error propagation, while missed calls hurt knowledge-intensive queries or questions requiring up-to-date evidence. Existing methods typically trigger tools from absolute query or generation signals, such as difficulty, confidence, or final task reward, and therefore lack an explicit estimate of the instance-level marginal benefit of tool use. We propose CoBRA, a counterfactual boundary-learning framework for tool-augmented language models. CoBRA first constructs internal and external experts from the same base model, collects paired trajectories, and estimates the reward margin between answering with and without tools. This margin partitions data into internal-favored, external-favored, and ambiguous cases. CoBRA then uses clear-margin samples for Boundary-Aware Cold-Start SFT, followed by MARS-RL with reference-split rollouts and counterfactual marginal advantages to optimize boundary decisions. Experiments with retrieval as the main tool on Qwen3-4B show that CoBRA improves tool-use efficiency and boundary-sensitive answer accuracy while maintaining strong performance on tool-dependent out-of-distribution questions.

Chinese Translation

随着大型语言模型越来越多地通过外部工具进行操作，决定何时调用工具已成为一个核心问题，与决定如何使用工具并列。不必要的工具调用会引入延迟、成本、检索噪声和错误传播，而漏掉的调用则会影响知识密集型查询或需要最新证据的问题。现有方法通常基于绝对查询或生成信号（如难度、信心或最终任务奖励）来触发工具，因此缺乏对实例级工具使用边际收益的明确估计。我们提出了CoBRA，一个用于工具增强语言模型的反事实边界学习框架。CoBRA首先从同一基础模型构建内部和外部专家，收集配对轨迹，并估计使用工具与不使用工具之间的奖励边际。该边际将数据划分为内部偏好、外部偏好和模糊案例。然后，CoBRA利用清晰边际样本进行边界感知冷启动的监督微调（Boundary-Aware Cold-Start SFT），随后通过参考分割回放和反事实边际优势进行MARS-RL，以优化边界决策。在以检索作为主要工具的Qwen3-4B实验中，CoBRA提高了工具使用效率和边界敏感的答案准确性，同时在工具依赖的分布外问题上保持了强劲的表现。

View on arXiv Download PDF AI Translation

cs.AI / 93 / 2609.01006

Figures as Programs: Recursive Generation of Editable Scientific Figures

图形作为程序：可编辑科学图形的递归生成

Liu, Yepeng, Dai, Dasen, Liu, Chengzhi, Song, Yiren, Ci, Hai, Zhang, Yu, Zhang, Qi, Shou, Mike Zheng, Wang, Xin Eric, Bu, Yuheng

Abstract

Scientific methodology figures are essential for communicating complex methods clearly, yet creating them remains labor-intensive and typically requires multiple rounds of refinement. Recent image-generation models can synthesize visually appealing raster figures, but producing a human-satisfactory result in a single generation step remains difficult. Moreover, precise edits to raster figures are challenging for both humans and models. We formulate scientific figure generation as recursive SVG program construction and propose \textsc{FigTree}, a \textit{multi-agent} system that automatically transforms a scientific paper into a structured vector figure. \textsc{FigTree} grounds figure content in the source paper, decomposes a figure into a hierarchy of local regions, generates each region as a short SVG program, and assembles the resulting fragments. A render-critic refinement loop jointly inspects the rendered figure and its underlying program, enabling visual defects to be traced to specific statements and accurately repaired. We conduct extensive evaluations of \textsc{FigTree} on figure quality and editability, showing that \textsc{FigTree} produces high-quality figures, while also enabling more effective editing than existing raster-based methods.

Chinese Translation

科学方法论图形对于清晰地传达复杂方法至关重要，但创建这些图形仍然劳动密集，通常需要多轮的精细化。近期的图像生成模型能够合成视觉上吸引人的光栅图形，但在单次生成步骤中产生令人满意的结果仍然困难。此外，对光栅图形进行精确编辑对人类和模型来说都是具有挑战性的。我们将科学图形生成形式化为递归SVG程序构建，并提出了 extsc{FigTree}，一个自动将科学论文转化为结构化矢量图形的 extit{多代理}系统。 extsc{FigTree}将图形内容与源论文相结合，将图形分解为局部区域的层次结构，生成每个区域作为简短的SVG程序，并组装生成的片段。一个渲染-评估的精细化循环共同检查渲染的图形及其底层程序，使得视觉缺陷能够追溯到特定的语句并得到准确修复。我们对 extsc{FigTree}在图形质量和可编辑性方面进行了广泛评估，结果表明 extsc{FigTree}生成高质量图形，同时也比现有的基于光栅的方法更有效地支持编辑。

View on arXiv Download PDF AI Translation

cs.AI / 94 / 2609.01035

Spawn Freely, Act Sparingly: Progressive Risk Vesting for Recursive LLM-Agent Trees

自由生成，谨慎行动：递归 LLM-代理树的渐进风险归属

Wang, Molly

Abstract

Recursive LLM agents can broaden their search by spawning specialists. Some branches later request tools that send data or deploy code. When should a branch receive authority to act? We distinguish sandbox spawning, in which external controls prevent the specified harm, from capability activation, in which a selected branch crosses an irreversible-action boundary. Progressive Risk Vesting (PRV) holds a trajectory-level risk budget in escrow and debits it as branches are activated. We prove an anytime harm bound for adaptively generated trees. Branch outcomes may be dependent, but each local certificate needs to remain valid conditional on the full pre-activation history, including the information used to select the request. When activation gates, branch charges, and compute constraints are held fixed, delayed vesting preserves every policy available under irrevocable spawn charging. Marginal risk estimates can still fail after branch selection. In a stylized branching model, trajectory harm changes as the authority reproduction number $\mathcal{R}_A$ crosses one. As local risk $p$ approaches zero, trajectory harm is proportional to $p$ below criticality, proportional to $\sqrt{p}$ at criticality, and retains a positive floor above it. A finite-type occupancy model yields risk and compute shadow prices. For nested fanout modes with decreasing marginal value per unit risk, these prices produce a threshold rule. Branching calculations and a split-sample experiment illustrate the results. These synthetic studies do not estimate safety in deployed agents. The analysis suggests a design rule: search broadly in the sandbox and grant recursive authority sparingly, with an explicit risk charge.

Chinese Translation

递归 LLM 代理可以通过生成专家来扩大搜索范围。一些分支随后请求工具以发送数据或部署代码。何时应当给予分支行动的权力？我们区分沙箱生成（sandbox spawning），在这种情况下，外部控制防止了指定的危害，以及能力激活（capability activation），在这种情况下，所选分支跨越了不可逆行动的边界。渐进风险归属（Progressive Risk Vesting, PRV）在托管中保持轨迹级别的风险预算，并在激活分支时进行扣除。我们证明了自适应生成树的任意危害界限。分支结果可能是相互依赖的，但每个局部证书需要在全预激活历史的条件下保持有效，包括用于选择请求的信息。当激活门、分支费用和计算约束保持固定时，延迟归属保留了在不可撤销生成收费下可用的每一政策。边际风险估计在分支选择后仍可能失败。在一个简化的分支模型中，轨迹危害随着权力再生产数 $ ext{R}_A$ 的变化而变化。当局部风险 $p$ 接近零时，轨迹危害在临界点以下与 $p$ 成正比，在临界点时与 $ ext{sqrt}(p)$ 成正比，并在其上保持一个正的下限。有限类型占用模型产生风险和计算影子价格。对于边际价值随风险单位减少的嵌套分支模式，这些价格产生了一个阈值规则。分支计算和分样实验说明了结果。这些合成研究并未估计已部署代理的安全性。分析建议了一条设计规则：在沙箱中广泛搜索，并谨慎授予递归权力，同时明确风险收费。

View on arXiv Download PDF AI Translation

cs.AI / 95 / 2609.01038

Data-Driven Persona-Conditioned Agents for A/B Test Simulation

基于数据驱动的人物条件代理的A/B测试模拟

Benomar, Ziyad, Łajewska, Weronika, Perelli, Leonardo, Mansour, Saab

Abstract

A/B testing is the gold standard for evaluating product changes, but each experiment requires real user traffic, engineering effort, and weeks of measurement. We propose a simulation framework that predicts A/B test outcomes using LLM-powered agents conditioned on data-driven personas grounded in real user behavioral signals. Unlike prior work that relies on synthetic or rule-based personas, our agents are constructed from anonymized behavioral data-activity patterns, engagement signals, and inferred demographics-enabling more faithful population modeling. We frame A/B test simulation as a structured question task and systematically study (i) question design formats, (ii) the impact of persona data source and domain alignment, (iii) the trade-off between per-persona behavioral depth and population diversity, and (iv) efficient population subsampling. On a benchmark of 40 A/B tests spanning two metric types, our best configuration achieves 0.75-0.90 directional accuracy depending on the test metric, demonstrating that data-driven personas are a viable path toward fast, low-cost experiment pre-screening.

Chinese Translation

A/B测试是评估产品变更的黄金标准，但每个实验都需要真实用户流量、工程投入和数周的测量时间。我们提出了一种模拟框架，通过基于真实用户行为信号的数据驱动人物构建的LLM（大语言模型）驱动代理来预测A/B测试结果。与依赖于合成或基于规则的人物的先前工作不同，我们的代理是由匿名化的行为数据、活动模式、参与信号和推断的人口统计信息构建的，从而实现了更真实的人口建模。我们将A/B测试模拟框架视为一个结构化问题任务，并系统性地研究了（i）问题设计格式，（ii）人物数据源和领域对齐的影响，（iii）每个人物行为深度与人口多样性之间的权衡，以及（iv）有效的人口子抽样。在涵盖两种指标类型的40个A/B测试基准上，我们的最佳配置在测试指标上实现了0.75-0.90的方向准确率，证明了数据驱动的人物是快速、低成本实验预筛选的可行路径。

View on arXiv Download PDF AI Translation

cs.AI / 96 / 2609.01045

AgentFactory: Towards Automated Agentic System Design and Optimization

AgentFactory：面向自动化代理系统设计与优化

Zhang, Enci, Wang, Haofeng, Zhu, Yuesheng, Cui, Xiaole, Luo, Guibo

Abstract

Large Language Models (LLMs) have demonstrated remarkable capabilities as powerful components in agentic systems, enabling sophisticated reasoning and complex task execution. However, current approaches to manually designing and optimizing agentic systems heavily rely on manual effort, limiting their adaptability and scalability. Recent work has explored the automated optimization of workflow designs. However, these approaches often overlook the crucial role of model capabilities and focus on single performance metrics, failing to address real-world deployment constraints. In this paper, we present AgentFactory, a framework that jointly optimizes both foundation models and workflow structures in agentic systems while considering multiple objectives including performance, cost, and efficiency. AgentFactory leverages advanced LLMs as optimizers to navigate the vast search space of possible configurations, employing a three-stage optimization pipeline to automatically discover effective combinations of fine-tuned models and optimized workflows. Through an iterative optimization process, our framework systematically explores and evaluates different agentic system designs, adapting to task-specific requirements while maintaining operational efficiency. We evaluate AgentFactory across eight benchmarks spanning five domains, including general reasoning, coding, mathematics, medicine, and finance. Our experiments demonstrate that AgentFactory consistently outperforms both manually designed methods and existing automated approaches, achieving an average improvement of 9.1% across all benchmarks, with particularly significant gains in domain-specific tasks (19.6% on MedQA and 18.7% on FinEval). These results establish AgentFactory as a promising approach for developing more capable and efficient agentic systems through automated optimization.

Chinese Translation

大型语言模型（LLMs）作为代理系统中的强大组件，展现了卓越的能力，能够实现复杂的推理和任务执行。然而，目前手动设计和优化代理系统的方法在很大程度上依赖于人工努力，这限制了其适应性和可扩展性。近期的研究探讨了工作流设计的自动化优化，但这些方法往往忽视了模型能力的重要作用，且仅关注单一性能指标，未能解决实际部署的约束。在本文中，我们提出了AgentFactory，一个框架，它在考虑多个目标（包括性能、成本和效率）的同时，联合优化代理系统中的基础模型和工作流结构。AgentFactory利用先进的LLMs作为优化器，导航可能配置的广阔搜索空间，采用三阶段优化流程自动发现有效的微调模型和优化工作流的组合。通过迭代优化过程，我们的框架系统地探索和评估不同的代理系统设计，适应特定任务的要求，同时保持操作效率。我们在涵盖一般推理、编码、数学、医学和金融等五个领域的八个基准测试中评估AgentFactory。实验结果表明，AgentFactory在所有基准测试中均优于手动设计的方法和现有的自动化方法，平均提升达到9.1%，在特定领域任务中（MedQA提升19.6%，FinEval提升18.7%）尤其显著。这些结果确立了AgentFactory作为通过自动化优化开发更强大和高效的代理系统的有前景的方法。

View on arXiv Download PDF AI Translation

cs.AI / 97 / 2609.01049

QILP-0: Constructing Observational Declarative Twins of Quantum Circuits

QILP-0：构建量子电路的观察性声明双胞胎

Echeandía, Marina de la Cruz, Alonso, César Luis, Ribeiro, Tony, de la Puente, Alfonso Ortega

Abstract

This paper introduces QXymb, a general framework for constructing observational declarative twins of quantum circuits, and develops QILP-0, its first complete order-0 specialization. QILP-0 constructs a finite multi-valued propositional logic program from observed circuit behaviour within a declared observational scope. The pipeline traverses a declared family of quantum observables incrementally according to a reproducible structural grading and a declared observational reference horizon. Progress is quantified through reference-relative coverage against a fixed target-independent reference. Observable responses are organized through target-independent geometry, while retained latent structure is mapped deterministically back to original observable columns before symbolic processing, preserving observational semantics and provenance. Selected observable profiles are converted into a finite relation through admissible target-independent discretization. The target is used only afterwards to audit twin-admissibility and induce the declarative theory. A theory is certified as an exact observational declarative twin when it completely and correctly reconstructs the resulting finite task-conditioned discrete relation. Logical exactness is therefore separated from numerical, backend, provider, and discretization uncertainty, which is retained as audit metadata. Validation uses two complementary QML settings. Exhaustive Bars & Stripes experiments compare product and grid-CZ embeddings from 16 to 100 qubits and exercise the native-discrete branch. Low-Depth MNIST analyses all 14,708 digit-0/1 instances before and after a trained variational quantum transformation and exercises continuous discretization. In every reported relation, the induced QILP-0 theory achieves complete, conflict-free reconstruction with strict accuracy equal to one.

Chinese Translation

本文介绍了QXymb，一个用于构建量子电路观察性声明双胞胎的通用框架，并开发了其第一个完整的0阶特化版本QILP-0。QILP-0从观察到的电路行为中构建一个有限的多值命题逻辑程序，范围在声明的观察范围内。该流程根据可重复的结构分级和声明的观察参考视野，逐步遍历声明的量子可观察量家族。进展通过相对于固定目标独立参考的覆盖量来量化。可观察响应通过目标独立几何进行组织，而保留的潜在结构在符号处理之前确定性地映射回原始可观察列，从而保持观察语义和来源。在经过可接受的目标独立离散化后，选定的可观察特征被转换为有限关系。目标仅在此之后用于审核双胞胎可接受性并引入声明理论。当理论完全且正确地重建所得到的有限任务条件离散关系时，它被认证为一个精确的观察性声明双胞胎。因此，逻辑精确性与数值、后端、提供者和离散化的不确定性分开，这些不确定性作为审核元数据保留。验证使用了两个互补的量子机器学习设置。全面的条形与条纹实验比较了从16到100个量子比特的产品和网格-CZ 嵌入，并测试了原生离散分支。低深度MNIST分析了14,708个数字0/1实例在经过训练的变分量子变换之前和之后的表现，并进行了连续离散化。在每个报告的关系中，诱导的QILP-0理论实现了完全、无冲突的重建，严格的准确性等于1。

View on arXiv Download PDF AI Translation

cs.AI / 98 / 2609.01056

WorldBench: Culturally Grounded Benchmark for Multilingual Agents

WorldBench：面向多语言智能体的文化基础基准

Ranaldi, Leonardo, Shen, Sherrie, Kai, Jushi, Birch, Alexandra

Abstract

Despite the growing use of LLM-powered agents to solve multi-step tasks in complex environments, existing benchmarks rarely test state preservation, performance across languages, and application to realistic, grounded scenarios. To address these concerns, we present WorldBench: a comprehensive, multilingual benchmark of genuine, persona-grounded everyday workflows, where agents can act in a sandbox via structured actions. WorldBench comprises 1,600 tasks across seven languages and eight cultures, filtered and refined through feedback from human annotators with language- and culture-specific expertise. For evaluation, we extend metrics from previous works and introduce Constrained Task Success (CTS), which combines natural language instructions and testbeds to score task completion, minimal modification, and other complementary metrics through deterministic and LLM-as-a-Judge evaluations. Our experiments show that frontier models reach only 49.2% CTS, with all models demonstrating large gaps between correctness and environment preservation. We thereby show that current agents remain brittle in multilingual, agentic scenarios, especially for long-horizon tasks and under state-preservation constraints

Chinese Translation

尽管基于大型语言模型（LLM）的智能体在复杂环境中解决多步骤任务的应用日益增长，但现有基准很少测试状态保持、跨语言性能以及在真实、基于情境的场景中的应用。为了解决这些问题，我们提出了WorldBench：一个全面的多语言基准，涵盖真实的、基于角色的日常工作流程，智能体可以通过结构化的行动在沙盒环境中进行操作。WorldBench包含1600个任务，覆盖七种语言和八种文化，经过具备语言和文化特定专业知识的人类注释者的反馈进行筛选和完善。为了评估，我们扩展了先前工作的指标，并引入了受限任务成功率（Constrained Task Success, CTS），该指标结合自然语言指令和测试平台，通过确定性评估和LLM作为评判者的评估来评分任务完成度、最小修改和其他互补指标。我们的实验表明，前沿模型的CTS仅为49.2%，所有模型在正确性和环境保持之间存在较大差距。因此，我们表明当前的智能体在多语言、智能体驱动的场景中仍然脆弱，尤其是在长时间跨度任务和状态保持约束下。

View on arXiv Download PDF AI Translation

cs.AI / 99 / 2609.01057

User Representation via Cross Multi-source Behavior Pre-training for Mobile Games

通过跨多源行为预训练进行用户表示学习的移动游戏研究

Yang, Chengqi, Qiao, Yiran, Liu, Feng, Lou, Xingyu, Zhou, Zijun, Mo, Xiaoyun, Zhang, Changwang, Xu, Jiayuan, Wang, Jun, Ao, Xiang

Abstract

User representation pre-training has become a fundamental paradigm for alleviating data sparsity in downstream personalization tasks. However, existing studies predominantly focus on single-app or app-level behaviors, overlooking the inherently cross-source and multi-granular nature of user activities on mobile devices. At the device level, user intent emerges from complex interactions among heterogeneous behavior sources and hierarchical action structures, posing challenges that cannot be addressed by conventional app-centric modeling. To tackle this issue, we propose CM-PTM, a novel Cross Multi-source Behavior Pre-Training Model tailored for mobile game user representation learning on device-level behavioral logs. CM-PTM employs hierarchical cascaded mask-then-predict proxy tasks that first infer the source of the next behavior and then progressively refine predictions at the app-action level. This design enables unified modeling of cross-source dependencies and fine-grained behavioral dynamics within a single pre-training paradigm. Extensive experiments on large-scale real-world mobile datasets demonstrate that CM-PTM effectively captures users' endogenous interests and consistently delivers significant performance gains on downstream mobile game recommendation tasks.

Chinese Translation

用户表示预训练已成为缓解下游个性化任务中数据稀疏问题的基本范式。然而，现有研究主要集中于单应用或应用级别的行为，忽视了移动设备上用户活动固有的跨源和多粒度特性。在设备层面，用户意图源于异构行为源和层次化行动结构之间的复杂交互，这给传统的应用中心建模带来了挑战。为了解决这一问题，我们提出了CM-PTM，一种新颖的跨多源行为预训练模型，旨在基于设备级行为日志进行移动游戏用户表示学习。CM-PTM采用层次级联的掩蔽-预测代理任务，首先推断下一个行为的来源，然后逐步在应用-动作级别上细化预测。这一设计使得在单一预训练范式中能够统一建模跨源依赖关系和细粒度行为动态。在大规模真实世界移动数据集上的广泛实验表明，CM-PTM有效捕捉用户的内生兴趣，并在下游移动游戏推荐任务中持续提供显著的性能提升。

View on arXiv Download PDF AI Translation

cs.AI / 100 / 2609.01058

ARISE-RL: Agentic Rubric-Grounded Iterative Self-Evolution with Reinforcement Learning

ARISE-RL：基于代理的评估标准驱动的强化学习迭代自我进化

Zhang, Fanrui, Ding, Ruixue, Zhang, Qiang, Chen, Xi, Chen, Boli, Wang, Shihang, Wang, Qiuchen, Zhan, Hongmin, Bian, Jinxin, xingchao, Li, Zheng, Peijin, cheng, Hao, Xie, Pengjun, Zhang, Kaipeng, Liu, Jiawei, Zha, Zheng-Jun

Abstract

Training open-ended agents via reinforcement learning (RL) is hindered by the lack of verifiable gold answers and scalable rubrics. Moreover, even near the model's capability boundary, long-horizon open-ended agentic tasks often yield brittle and unstable rewards, resulting in weak or noisy rollout contrast that obscures fine-grained optimization signals for group-based policy learning. To address these challenges, we propose ARISE-RL, a novel full-cycle self-evolution framework that couples a task/rubric Generator and a reasoning Solver through rubric-mediated co-evolution. The Generator grounds tool-related rubric criteria in real tool observations and is rewarded for producing valid, intermediate-difficulty tasks aligned with the Solver's evolving capability boundary. The Solver, in turn, learns from fine-grained rubric satisfaction signals through multi-step reasoning and tool use. We further introduce Reward-Gated Self-Evolution Distillation (RG-SED), which selectively distills a memory-augmented variant of the same policy back into itself only when the memory yields empirical reward improvement, thereby reducing distribution mismatch and avoiding blind imitation of noisy guidance. Finally, to support rigorous evaluation, we present ECR-Bench, an expert-calibrated rubric benchmark suite covering single-tool deep research and multi-tool travel planning. Extensive experiments demonstrate that ARISE-RL consistently achieves robust and stable overall state-of-the-art performance across all evaluated benchmarks.

Chinese Translation

通过强化学习（RL）训练开放式智能体受到可验证的黄金答案和可扩展评估标准缺乏的限制。此外，即使在模型能力边界附近，长时间跨度的开放式智能体任务通常会产生脆弱和不稳定的奖励，导致基于群体的策略学习中的细粒度优化信号模糊不清。为了解决这些挑战，我们提出了ARISE-RL，这是一种新颖的全周期自我进化框架，通过评估标准介导的共同进化将任务/评估标准生成器和推理求解器结合在一起。生成器将与工具相关的评估标准标准化为真实工具观察，并因生成与求解器不断演变的能力边界一致的有效中等难度任务而获得奖励。求解器则通过多步推理和工具使用，从细粒度的评估标准满足信号中学习。我们进一步引入了奖励门控自我进化蒸馏（RG-SED），该方法仅在记忆产生经验奖励改善时，选择性地将同一策略的记忆增强变体蒸馏回自身，从而减少分布不匹配，避免盲目模仿噪声指导。最后，为了支持严格的评估，我们提出了ECR-Bench，这是一个经过专家校准的评估标准基准套件，涵盖单工具深度研究和多工具旅行规划。大量实验表明，ARISE-RL在所有评估基准中始终实现了稳健和稳定的最先进整体性能。

View on arXiv Download PDF AI Translation

cs.AI / 101 / 2609.01062

Space Generative AI with Solar Energy Harvesting

利用太阳能收集的空间生成性人工智能

Zhang, Jierui, Huang, Jianhao, Wang, Zhanwei, Huang, Kaibin

Abstract

Satellites are emerging as promising platforms to extend generative \emph{artificial intelligence} (AI) services to remote areas lacking terrestrial infrastructure. However, deploying space generative AI is fundamentally constrained by the limited, time-varying onboard energy supplied by solar \emph{energy harvesting} (EH). This paper presents a framework for solar-powered space generative AI in which a satellite receives a user prompt, executes a diffusion-based image-generation model, and downlinks the compressed result within a strict time window. We identify the fundamental \emph{computation--communication} (C$^2$) trade-offs governed by the shared harvested-energy budgets. Specifically, increasing the number of generation steps improves intrinsic image quality but depletes energy and time available for downlink transmission, whereas prioritizing communication guarantees reliable delivery but sacrifices semantic quality. To balance these trade-offs and maximize \emph{end-to-end} (E2E) generative performance, we exploit the predictable solar-EH dynamics induced by deterministic orbital motion and develop a joint C$^2$ resource-optimization framework using a tractable two-step approach. First, we characterize the maximum downlink throughput for a fixed generation depth under continuous solar EH. This establishes a separation principle that decouples waiting-time selection from optimal transmit-power control. Next, we formulate a joint C$^2$ utility-maximization problem and derive a closed-form, low-complexity step-selection policy in the dominant constant-power regime. Extensive experiments under realistic orbital dynamics demonstrate that the proposed policy dynamically balances generation quality and transmission reliability. This yields significant E2E performance gains over static computation- and communication-centric baselines across diverse solar-EH states.

Chinese Translation

卫星作为有前景的平台，正在扩展生成性人工智能（AI）服务到缺乏地面基础设施的偏远地区。然而，部署空间生成性AI在根本上受到太阳能收集（EH）所提供的有限且随时间变化的机载能量的限制。本文提出了一种太阳能驱动的空间生成性AI框架，其中卫星接收用户提示，执行基于扩散的图像生成模型，并在严格的时间窗口内下行传输压缩结果。我们识别了由共享收集能量预算所主导的基本计算-通信（C$^2$）权衡。具体而言，增加生成步骤的数量可以提高内在图像质量，但会消耗下行传输可用的能量和时间，而优先考虑通信则保证了可靠的交付，但牺牲了语义质量。为了平衡这些权衡并最大化端到端（E2E）生成性能，我们利用由确定性轨道运动引起的可预测太阳能EH动态，并开发了一种联合C$^2$资源优化框架，采用可处理的两步方法。首先，我们在持续的太阳能EH下表征固定生成深度的最大下行吞吐量。这建立了一个分离原则，将等待时间选择与最佳发射功率控制解耦。接下来，我们制定了一个联合C$^2$效用最大化问题，并在主导的恒定功率范围内推导出一个封闭形式的低复杂度步骤选择策略。在现实轨道动态下进行的大量实验表明，所提出的策略动态平衡了生成质量和传输可靠性。这在各种太阳能EH状态下，相较于静态计算和通信中心基线，显著提高了E2E性能。

View on arXiv Download PDF AI Translation

cs.AI / 102 / 2609.01117

Latent Recurrent Thoughts: Recurrent Refinement of Proposed Latents for Reasoning with Frozen LLMs

潜在递归思维：对冻结的大型语言模型进行推理时提议潜在状态的递归优化

Chen, Zhaoliang, Fu, Jie

Abstract

Chain-of-thought reasoning unfolds in discrete token space: each step is committed as text, errors propagate, and eliciting good traces presupposes traces to imitate. Reasoning instead in a model's continuous representation space - where intermediate states are vectors rather than words - sidesteps these constraints, but leaves open how those latent states should be computed. We approach this along two axes. First, we keep a large language model (LLM) frozen and use it for what it is already good at - modeling and decoding sequences - while a small auxiliary network supplies continuous latent thoughts as input. Second, we produce those latents by recurrence: a tiny recurrent reasoner refines them over many steps, decoupling the depth of computation from the size of the model, so that the latents are a product of iterative processing rather than a single forward pass. We instantiate this as Latent Recurrent Thoughts (LRT): a task-dedicated proposer supplies base latents, a recurrent reasoner refines them through bounded residual corrections, and the frozen LLM decodes the answer. On symbolic reasoning with answer supervision but no reasoning traces (Countdown-4, Sudoku) and on natural-language reasoning (HumanEval, MBPP, StrategyQA), LRT substantially outperforms prior frozen-decoder continuous-space reasoning methods under an identical decoder, prompt, data, and training budget, and outperforms non-thinking-mode chain-of-thought prompting on the same backbone at a small fraction of its inference compute.

Chinese Translation

思维链推理在离散的符号空间中展开：每一步都以文本形式记录，错误会传播，产生良好轨迹的前提是存在可模仿的轨迹。相反，在模型的连续表示空间中进行推理——其中中间状态是向量而非单词——可以规避这些限制，但如何计算这些潜在状态仍然是一个未解的问题。我们从两个方面入手。首先，我们保持大型语言模型（LLM）的冻结状态，利用其擅长的任务——建模和解码序列，同时一个小型辅助网络提供连续的潜在思维作为输入。其次，我们通过递归生成这些潜在状态：一个微小的递归推理器在多个步骤中对其进行优化，将计算深度与模型大小解耦，使得潜在状态成为迭代处理的产物，而非单次前向传播的结果。我们将其实现为潜在递归思维（Latent Recurrent Thoughts, LRT）：一个专门的提议者提供基础潜在状态，一个递归推理器通过有界的残差修正对其进行优化，冻结的LLM则解码出答案。在符号推理（有答案监督但没有推理轨迹的任务，如Countdown-4和数独）以及自然语言推理（如HumanEval、MBPP和StrategyQA）中，LRT在相同的解码器、提示、数据和训练预算下，显著优于之前的冻结解码器连续空间推理方法，并在相同基础架构下以极小的推理计算量超越了非思维模式的思维链提示。

View on arXiv Download PDF AI Translation

cs.AI / 103 / 2609.01168

Jailbreaking Text-to-Image Models Through Cracks: Navigating Heterogeneous Safety Filters via Multi-Agent Debate

通过漏洞破解文本到图像模型：通过多智能体辩论导航异构安全过滤器

Wen, Kaiyan, Zhang, Shijie, Yu, Lu, Bai, Guangdong

Abstract

Text-to-image (T2I) models remain vulnerable to jailbreak attacks that elicit Not-Safe-For-Work (NSFW) content, despite increasingly being guarded by heterogeneous, multi-layer safety stacks combining text filters, image classifiers, and cross-modal detectors. Existing jailbreak studies either optimize against individual filters or query the complete pipeline with aggregate feedback, making it difficult to identify the active constraint and adapt to conflicts across safety layers.In this paper, we introduce the \emph{Detection Surface}, a unified geometric framework that characterizes the decision boundaries induced by heterogeneous T2I safety filters and their joint effect on the jailbreak search space. This formulation reveals that successful evasion is governed by a sparse and non-convex region shaped by cross-layer conflicts, where mutations that bypass one filter may increase exposure to another. Motivated by this analysis, we propose \emph{CRACK}, a multi-agent debate framework for adaptive jailbreak search that decomposes jailbreak search into exploration, diagnosis, and arbitration. CRACK coordinates an Attack Agent, a Defense Agent, and a Judge Agent to iteratively generate prompt mutations, obtain layer-specific diagnostic feedback, and optimize mutation strategies through reward-guided refinement. Through repeated rounds of debate, CRACK adapts its search direction to the evolving cross-layer constraints while preserving the original harmful intent. Extensive experiments across multiple T2I models, datasets, and safety configurations show that CRACK achieves Attack Success Rates (ASR) of up to 99.63\% under composite defenses, while requiring fewer queries than existing methods and maintaining semantic fidelity.

Chinese Translation

尽管文本到图像（T2I）模型越来越受到结合文本过滤器、图像分类器和跨模态检测器的异构多层安全堆栈的保护，但仍然容易受到引发不适合工作的内容（NSFW）的破解攻击。现有的破解研究要么针对单个过滤器进行优化，要么使用汇总反馈查询完整的处理流程，这使得识别活跃约束和适应安全层之间的冲突变得困难。在本文中，我们引入了 extit{检测表面}（Detection Surface），这是一个统一的几何框架，用于表征异构T2I安全过滤器所引起的决策边界及其对破解搜索空间的联合影响。这一表述揭示了成功规避由跨层冲突所塑造的稀疏且非凸区域的规律，其中绕过一个过滤器的变异可能会增加对另一个过滤器的暴露。基于这一分析，我们提出了 extit{CRACK}，一个用于自适应破解搜索的多智能体辩论框架，将破解搜索分解为探索、诊断和仲裁。CRACK协调攻击代理、防御代理和评判代理，迭代生成提示变异，获取层特定的诊断反馈，并通过奖励引导的优化来改进变异策略。通过多轮辩论，CRACK调整其搜索方向以适应不断变化的跨层约束，同时保持原有的有害意图。在多个T2I模型、数据集和安全配置下的广泛实验表明，CRACK在复合防御下实现了高达99.63%的攻击成功率（ASR），同时所需查询次数少于现有方法，并保持语义一致性。

View on arXiv Download PDF AI Translation

cs.AI / 104 / 2609.01198

FinLifeBench: Exhaustive Life-Event History and Financial-State Reconstruction from Longitudinal Banking Dialogue

FinLifeBench：从纵向银行对话中全面重建生活事件历史和财务状态

Lee, Hangyeul, Oh, Juyoung, Ko, Jaeyong, Kim, Sunmin, Park, Jaeik, Kim, Hyunkyu, Son, Jungmin, Kang, Pilsung

Abstract

Repeated banking interactions require assistants to maintain complete, current, and traceable customer records as life changes emerge incidentally in routine requests. Existing benchmarks emphasize question answering, bounded episodes, or targeted recall rather than exhaustive longitudinal reconstruction. We introduce FinLifeBench, which evaluates two tasks over the same cumulative dialogue: reconstructing every life-event instance with its first-establishing session and reconstructing a complete 34-path financial state at consecutive checkpoints. The benchmark contains 6,000 eight-turn Korean banking sessions from 20 independent synthetic trajectories, with deterministic, exhaustive gold for 24 event types and 34 state paths and consensus quality assurance. Across eleven LLMs under a full-context condition, event-anchor recall falls from 0.591 at 15 sessions to 0.445 at 300. Errors are driven primarily by omitted events rather than poor anchor localization, while financial-state reconstruction frequently treats superseded or potentially outdated information as current; the best GCA@15 reaches 0.470. Performance on the two reconstruction tasks is only weakly associated. These results show that models can localize evidence for recovered events while still failing to maintain complete and temporally valid longitudinal records.

Chinese Translation

重复的银行互动要求助手在日常请求中随着生活变化的偶然出现，保持完整、最新和可追溯的客户记录。现有基准强调问答、有限的情节或针对性的回忆，而非全面的纵向重建。我们引入了FinLifeBench，该基准评估在同一累积对话中进行的两个任务：重建每个生活事件实例及其首次建立的会话，以及在连续检查点重建完整的34条财务状态路径。该基准包含来自20条独立合成轨迹的6000个八轮韩语银行会话，提供24种事件类型和34条状态路径的确定性、全面的金标准，并确保共识质量。在全上下文条件下，11个大型语言模型（LLMs）的事件锚点回忆在15个会话时为0.591，300个会话时降至0.445。错误主要由遗漏事件驱动，而非锚点定位不佳，同时财务状态重建常常将被取代或可能过时的信息视为当前；最佳的GCA@15达到0.470。两个重建任务的表现仅弱相关。这些结果表明，模型能够定位恢复事件的证据，但仍未能保持完整且时间有效的纵向记录。

View on arXiv Download PDF AI Translation

cs.AI / 105 / 2609.01216

H2Table: Hierarchical Hypergraph-Enhanced Large Language Models for Complex Table Reasoning

H2Table：用于复杂表格推理的层次化超图增强大型语言模型

Ling, Jia, Wang, Yangfan, Tang, Chen, Tan, Haoming, Yang, Yang, Guan, Yi, Jiang, Jingchi

Abstract

Tables are ubiquitous across diverse domains, yet reasoning over them remains a significant challenge for modern large language models (LLMs). Current approaches typically linearize tables into sequences, inherently overlooking their intrinsic two-dimensional and hierarchical structure. To address this, we propose H2Table (Hierarchical Hypergraph-Enhanced Table Reasoning), a novel framework that represents complex tables as hierarchical nested hypergraphs. To process this representation, we design a tailored hypergraph encoder to facilitate message passing between hyperedges (headers) and nodes (cells), thereby perceiving the semantic entailment relationships between them within complex tables. Furthermore, we introduce a set of learnable query vectors acting as a lightweight bridge to extract representative structural embeddings from the encoder into the LLM. Experimental results demonstrate that our approach effectively handles complex table question answering tasks with hierarchical nested headers. Notably, on the HiTab dataset, H2Table achieves an average improvement of 22.88% over state-of-the-art baselines on highly complex tables with a nesting depth of four. Our code is available at: https://github.com/lila120/h2table.

Chinese Translation

表格在各个领域中无处不在，但对其进行推理仍然是现代大型语言模型（LLMs）面临的一项重大挑战。目前的方法通常将表格线性化为序列，固有地忽视了其内在的二维和层次结构。为了解决这个问题，我们提出了H2Table（层次化超图增强表格推理），这是一个将复杂表格表示为层次嵌套超图的新框架。为了处理这种表示，我们设计了一个定制的超图编码器，以促进超边（表头）和节点（单元格）之间的信息传递，从而感知它们在复杂表格中的语义蕴含关系。此外，我们引入了一组可学习的查询向量，作为轻量级桥梁，将代表性的结构嵌入从编码器提取到LLM中。实验结果表明，我们的方法有效地处理了具有层次嵌套表头的复杂表格问答任务。值得注意的是，在HiTab数据集上，H2Table在嵌套深度为四的高度复杂表格上，相较于最先进的基线平均提高了22.88%。我们的代码可在以下链接获取：https://github.com/lila120/h2table。

View on arXiv Download PDF AI Translation

cs.AI / 106 / 2609.01217

Prompt-Robust Language Models: Which Training Strategies Work?

稳健提示语言模型：哪些训练策略有效？

Sadrieh, Frederic, Štefánik, Michal

Abstract

Despite their strong performance, large language models remain highly sensitive to prompt formulation. Prior work addresses this through refined data construction or through dedicated robustness objectives. We reproduce and compare these strategies under controlled conditions, and measure how effective they are in addressing models' prompt sensitivity. We find the current robustness fine-tuning methods improve over standard fine-tuning and in-context learning, but the best-to-worst prompt gap remains as high as 40-57% of performance. Moreover, the recent robustness-enhancing methods we test - CoIN for contrastive alignment and PPCL for consistency regularization - often fail to outperform the simplest data construction strategy: training on one template per batch. Our diagnostics explain these results. The auxiliary objectives move the quantity they penalize, but do not generalize beyond it. Additionally, data construction strategies differ due to the conflicting signs of per-template gradients on 57-64% of parameters. Thus, batches that mix formulations force the optimizer to reconcile competing updates instead of finding a shared, prompt-agnostic one.

Chinese Translation

尽管大型语言模型表现出色，但仍对提示的表述高度敏感。之前的研究通过精细的数据构建或专门的稳健性目标来解决这一问题。我们在受控条件下重现并比较这些策略，并测量它们在解决模型提示敏感性方面的有效性。我们发现当前的稳健性微调方法在标准微调和上下文学习上有所改进，但最佳与最差提示之间的差距仍高达40-57%的性能。此外，我们测试的最近稳健性增强方法——对比对齐的 CoIN 和一致性正则化的 PPCL——往往未能超越最简单的数据构建策略：每批次训练一个模板。我们的诊断解释了这些结果。辅助目标移动了它们所惩罚的量，但并未超出这一范围进行泛化。此外，由于57-64%参数的每模板梯度存在相互矛盾的符号，数据构建策略也存在差异。因此，混合表述的批次迫使优化器调和竞争更新，而不是找到一个共享的、与提示无关的更新。

View on arXiv Download PDF AI Translation

cs.AI / 107 / 2609.01257

Measuring the Behavioral Fidelity of Long-Horizon Human Activity Simulations

测量长期人类活动模拟的行为忠实度

Cheng, Yi Fei, Yang, Fan, Bas, Iremsu, Niinuma, Koichiro, Abe, Narishige, Lindlbauer, David

Abstract

As LLM-based human simulators are increasingly used for policy, evaluation, and training, they must faithfully reproduce real behavioral patterns. While prior work has examined behavioral fidelity in survey responses and dialogue, longer-horizon real-world activity remains largely unexplored. We introduce a framework for evaluating behavioral fidelity in long-horizon activity simulations across temporal granularities and levels of analysis. As a case study, we collect a 43-hour multi-camera dataset of in-the-wild office activity and compare trace-derived conditioning mechanisms: persona descriptors, few-shot exemplars, and statistical transition and time-of-day priors. We find that behavioral fidelity is not uniform across metrics: statistical priors bring activity and sequence distributions closest to real behavior, yet over-fragment routines and suppress within-person variability. These findings motivate a more holistic evaluation that spans multiple metrics, temporal granularities, and levels of analysis.

Chinese Translation

随着基于大型语言模型（LLM）的人类模拟器在政策、评估和培训中的应用日益增多，它们必须忠实地再现真实的行为模式。虽然之前的研究已经考察了调查响应和对话中的行为忠实度，但长期的现实活动仍然在很大程度上未被探索。我们提出了一个评估长期活动模拟中行为忠实度的框架，涵盖了时间粒度和分析层次。作为案例研究，我们收集了一个43小时的多摄像头野外办公室活动数据集，并比较了基于轨迹的条件机制：个性描述符、少量示例和统计转移及时间优先级。我们发现，行为忠实度在不同指标间并不均匀：统计优先级使活动和序列分布最接近真实行为，但却过度细分了常规并抑制了个体内变异性。这些发现促使我们进行更全面的评估，涵盖多个指标、时间粒度和分析层次。

View on arXiv Download PDF AI Translation

cs.AI / 108 / 2609.01260

Dual Process Motion Planning

双重过程运动规划

Yan, Jiayi, Fabiano, Francesco, Abate, Alessandro

Abstract

Robotic systems are deeply embedded in both industry and everyday life, where they are expected to act with speed, precision, and reliability. Classical control and planning methods have long delivered strong guarantees, but often at the cost of computational efficiency and adaptability. More recently, learning-based approaches have shown promise in overcoming these limitations, enabling agents to leverage experience to accelerate decision-making and address previously intractable problems. In this work, we bridge these two approaches through a neuro-symbolic perspective on nonlinear motion planning. Inspired by the Thinking Fast and Slow paradigm, we introduce a dual-process architecture that combines the strengths of robust reasoning and learning. Our framework integrates state-of-the-art symbolic solvers as a ``System-2'' component with experience-driven ``System-1'' modules. A metacognitive controller dynamically orchestrates their interaction, selecting when to rely on fast intuition versus slower, more precise reasoning. By evaluating the framework across diverse nonlinear benchmark environments, we demonstrate that this architecture yields consistent gains in planning efficiency, accuracy, and generalization, while promoting reuse across tasks. The results suggest that tightly coupling learning with structured reasoning offers a scalable path toward more capable and adaptive robotic systems.

Chinese Translation

机器人系统深深嵌入工业和日常生活中，期望其具备快速、精确和可靠的行动能力。传统的控制和规划方法长期以来提供了强有力的保证，但往往以计算效率和适应性为代价。最近，基于学习的方法在克服这些局限性方面展现了希望，使得智能体能够利用经验加速决策过程并解决以前难以处理的问题。在本研究中，我们通过对非线性运动规划的神经符号视角，桥接了这两种方法。受到《思考，快与慢》范式的启发，我们引入了一种双重过程架构，结合了稳健推理和学习的优势。我们的框架将最先进的符号求解器作为“系统-2”组件与经验驱动的“系统-1”模块相结合。一个元认知控制器动态协调它们的互动，选择何时依赖快速直觉与较慢、更精确的推理。通过在多样的非线性基准环境中评估该框架，我们证明了这一架构在规划效率、准确性和泛化能力上都带来了持续的提升，同时促进了任务间的重用。结果表明，将学习与结构化推理紧密结合，为更强大和适应性的机器人系统提供了一条可扩展的路径。

View on arXiv Download PDF AI Translation

cs.AI / 109 / 2609.01272

Making Prospective Memory SLM-Shaped: Typed Intention Stores for Small-Model Agents

使前瞻性记忆呈现 SLM 形状：小型模型代理的类型化意图存储

Zhao, Jinqing, Wu, Chengcan

Abstract

Prospective memory means carrying out a deferred intention at the right future cue while other work continues. Benchmarks now isolate it as an agent skill, yet frontier LLMs still struggle: the best published PM-Bench scaffold reaches only 65.1% Set-F1. We argue that this loop is schema-constrained state tracking rather than open-ended reasoning, and that small models can execute it when the action space is typed. We propose the Prospective Intention Store (PIS) that puts lifecycle logic in code and scoped language work on the model. The scaffold is agentic and training-free: no selector fine-tuning and no trajectory distillation. On PM-Bench, DeepSeek-Chat with PIS reaches 82.9% Set-F1. On Gemma-E2B, Set-F1 is only 4.2% without a store and at most 6.6% under seven retrospective memories, while PIS reaches 66.2%. PIS further reaches 70.1% Set-F1, where retrospective memory methods stay at most 54.4%. PIS sets a new state of the art on this benchmark and enables small models to surpass the published large-model scaffold.

Chinese Translation

前瞻性记忆指的是在其他工作持续进行的同时，在合适的未来提示下执行延迟意图。目前的基准测试将其孤立为一种代理技能，但前沿的大型语言模型（LLM）仍然面临挑战：已发布的最佳 PM-Bench 框架仅达到 65.1% 的 Set-F1。我们认为这个循环是受模式约束的状态跟踪，而不是开放式推理，并且小型模型在动作空间类型化时可以执行它。我们提出了前瞻性意图存储（Prospective Intention Store, PIS），将生命周期逻辑编码并在模型上进行范围语言工作。该框架是自主的且无需训练：不需要选择器微调，也不需要轨迹蒸馏。在 PM-Bench 上，带有 PIS 的 DeepSeek-Chat 达到了 82.9% 的 Set-F1。在 Gemma-E2B 上，若没有存储，Set-F1 仅为 4.2%，在七个回顾性记忆下最多为 6.6%，而 PIS 达到了 66.2%。PIS 进一步达到了 70.1% 的 Set-F1，而回顾性记忆方法的最高值为 54.4%。PIS 在该基准测试中设定了新的技术标准，并使小型模型能够超越已发布的大型模型框架。

View on arXiv Download PDF AI Translation

cs.AI / 110 / 2609.01286

Analog-DB: An Agent-First Analog Integrated Circuit Database, From Blocks to Systems

Analog-DB：一种以代理为先的模拟集成电路数据库，从模块到系统

Zadeh, Danial Noori, Elamien, Mohamed B.

Abstract

Sharing analog integrated circuit designs remains difficult: foundry non-disclosure agreements restrict the process details a design depends on, and the testbenches behind published results are rarely released. We present analog-db, an open-source, versioned database built on a shareable design representation. A domain-specific language captures each design as a process-neutral topology, reusable testbenches, and a machine-readable datasheet under one schema, so a design is shared in full and re-simulates on the process kits it is bound to. A parameterization scheme exposes functional sub-blocks and device sizes as named parameters that carry their matching constraints, making circuits composable and retargetable; a schema-governed contract and queryable catalog let AI design agents discover and reuse them directly. Across the regulator corpus, all 23 circuit-kit bindings on three open kits meet their own recorded specification bands (typical corner, matched devices, no layout) and 10 of 23 meet a common class band. Seventeen of the 23 imported sizings failed their testbenches and closed under a gm/ID sizing loop driven by the annotated sub-block roles, typically within one to three iterations. In a supervised case study, a coding agent working from the released artifacts sized the op-amp cores of a chopper instrumentation amplifier on an open 130nm kit, locating four hand-entry defects and a missing common-mode feedback loop that the sizing-only baseline did not repair. The database holds 68 circuits across sixteen classes, verifiable at schematic level under a tiered harness and tracked on a power/performance scoreboard, released at https://github.com/MacAnalog/spicexplorer-release.

Chinese Translation

共享模拟集成电路设计仍然面临困难：代工厂的不披露协议限制了设计所依赖的过程细节，而已发布结果背后的测试平台很少被公开。我们提出了analog-db，一个基于可共享设计表示的开源版本数据库。一个领域特定语言将每个设计捕捉为一个与过程无关的拓扑结构、可重用的测试平台和一个机器可读的数据表，所有这些都在一个模式下，因此设计可以完整共享并在其绑定的工艺工具上重新仿真。参数化方案将功能子模块和器件尺寸作为命名参数暴露出来，并携带其匹配的约束，使电路可组合和重新定向；一个模式管理的合同和可查询的目录使得人工智能设计代理能够直接发现和重用它们。在调节器语料库中，所有23个电路工具在三个开放工具上满足其记录的规格带（典型角、匹配器件、无布局），其中10个满足一个共同的类别带。23个导入的尺寸中有17个未通过其测试平台，并在由注释子模块角色驱动的gm/ID尺寸循环中关闭，通常在一到三次迭代内。在一个监督案例研究中，一个基于发布文档工作的编码代理在一个开放的130nm工具上对一个斩波仪表放大器的运算放大器核心进行了尺寸调整，发现了四个手动输入缺陷和一个缺失的共模反馈回路，而仅进行尺寸调整的基线并未修复这些问题。该数据库包含68个电路，跨越十六个类别，在分层测试平台下可在原理图级别进行验证，并在功率/性能记分板上进行跟踪，发布于 https://github.com/MacAnalog/spicexplorer-release。

View on arXiv Download PDF AI Translation

cs.AI / 111 / 2609.01315

A Composable Evaluation System for Reproducible Omni-Modal Foundation Model Evaluation

可组合的全模态基础模型评估系统

Lee, Hodong, Park, Sanghee, Ryu, Dohoon, Kim, Jungwhan, Kim, Junyeob, Kim, Soyoon, Kim, Geewook

Abstract

Building an omni-modal foundation model means evaluating it across text, image, video, and audio. Excellent evaluation toolkits exist for each modality, but their inference engines, prompt conventions, and metric implementations are mutually incompatible, so practitioners end up maintaining separate environments for every toolchain and still struggle to compare results across them. OmniEvaluator grew out of this need in our own model development: rather than reimplementing benchmarks, it connects existing inference engines and curated evaluation libraries at a higher level, exposing four inference backends, four evaluation frameworks, and over a thousand benchmarks through a single interface. Every run is recorded as an artifact capturing the full configuration for exact reproduction, and results flow into a shared dashboard for cross-model comparison. A federated mode shares GPU inference servers across concurrent evaluations, and a built-in verifier, small enough to run on CPU, keeps its score stable across engines and prompts where rule-based scoring fluctuates under configuration mismatch, matching cost-efficient commercial LLM judges without their recurring API cost. The system, demo video, and dashboard are publicly available. (https://github.com/naver-ai/omni-evaluator)

Chinese Translation

构建一个全模态基础模型意味着需要在文本、图像、视频和音频等多个模态上进行评估。虽然每种模态都有优秀的评估工具包，但它们的推理引擎、提示约定和指标实现相互不兼容，因此从业者最终不得不为每个工具链维护独立的环境，并且在它们之间比较结果时仍然面临困难。OmniEvaluator 正是基于我们自身模型开发中的这一需求而产生的：它并不是重新实现基准测试，而是在更高的层次上连接现有的推理引擎和精心策划的评估库，通过单一接口暴露四个推理后端、四个评估框架以及超过一千个基准测试。每次运行都会被记录为一个捕捉完整配置的工件，以便于精确重现，结果流入一个共享仪表板以进行跨模型比较。联邦模式在并发评估中共享 GPU 推理服务器，而一个内置验证器足够小，可以在 CPU 上运行，确保其分数在不同引擎和提示下保持稳定，避免了基于规则的评分在配置不匹配时的波动，匹配了成本效益高的商业 LLM 评判者而无需其重复的 API 成本。该系统、演示视频和仪表板均已公开可用。 (https://github.com/naver-ai/omni-evaluator)

View on arXiv Download PDF AI Translation

cs.AI / 112 / 2609.01320

Automated Event Log Generation from Unstructured Text Using Finetuned LLMs

基于微调大语言模型的非结构化文本自动事件日志生成

Seeth, Maximilian, Tavares, Gabriel Marques, Schuster, Daniel

Abstract

Process mining (PM) provides a powerful framework for discovering and optimizing operational processes from event data. However, the efficacy of PM techniques is strictly predicated on the availability of structured event logs. Thus far, event logs have often been laboriously created by domain and process mining experts. This costly effort causes large portions of organizational knowledge, including incident tickets, manuals, and textual reports, to remain underutilized. We address this bottleneck by investigating the efficacy of Large Language Models (LLMs) as automated data translators. We propose a scalable framework that leverages LLMs as data translators to bridge the gap between unstructured textual resources and structured event data. We finetune LLMs on a newly created text-to-log dataset, demonstrating that the resulting models can extract high-fidelity event logs from unstructured resources. Our results show that this finetuning approach outperforms few-shot or zero-shot prompting by a large amount, highlighting finetuning as a necessary pre-condition for generating reliable event data. We conclude that our method provides a promising pipeline for making previously unused data available to process mining ecosystems, effectively expanding the possibilities of using PM to further investigate organizational workflows.

Chinese Translation

过程挖掘（PM）为从事件数据中发现和优化运营过程提供了强大的框架。然而，PM 技术的有效性严格依赖于结构化事件日志的可用性。迄今为止，事件日志往往由领域和过程挖掘专家费力地创建。这种高成本的工作导致大量组织知识，包括事件工单、手册和文本报告，未得到充分利用。我们通过研究大型语言模型（LLMs）作为自动数据翻译器的有效性来解决这一瓶颈。我们提出了一个可扩展的框架，利用 LLMs 作为数据翻译器，弥合非结构化文本资源与结构化事件数据之间的差距。我们在新创建的文本到日志数据集上微调 LLMs，证明所得到的模型能够从非结构化资源中提取高保真度的事件日志。我们的结果显示，这种微调方法在性能上大幅超越了少量示例或零示例提示，强调了微调作为生成可靠事件数据的必要前提。我们得出结论，方法为将以前未使用的数据提供给过程挖掘生态系统提供了一个有前景的管道，有效扩展了使用 PM 进一步研究组织工作流程的可能性。

View on arXiv Download PDF AI Translation

cs.AI / 113 / 2609.01337

LEAP: Likelihood Elicitation and Aggregation for LLM-based Probabilistic Forecasting

LEAP：基于LLM的概率预测中的似然提取与聚合

Chen, Yufei, Zhao, Yiran, Xu, Xiaogang, Xie, Qipeng, Wu, Jiafei, Liu, Zhe

Abstract

LLM-based forecasting systems have improved on real-world tasks such as financial markets and sports outcomes, largely through stronger search and tool use. Many systems still ask an LLM to read all collected evidence together and produce the final forecast. We call this design Monolithic Prediction. It can obscure how individual evidence items affect the result and collapse uncertainty across competing outcomes. We propose LEAP (Likelihood Elicitation and Aggregation for Probabilistic forecasting), which reorganizes how collected evidence is used in the prediction stage. LEAP examines each evidence item separately and elicits likelihood parameters that describe its implications for the target. An explicit prior and a deterministic probabilistic model then combine these likelihoods into a posterior distribution. This procedure supports continuous, single-choice, and multi-choice forecasts while preserving reproducible evidence contributions. We build a benchmark covering forecasting, information-seeking, and browsing tasks, and evaluate LEAP on our own agent loop and several agent CLI frameworks. Given the same evidence, LEAP improves most prediction and calibration metrics across models and remains stronger under controlled comparisons of prior access, inference budget, and aggregation.

Chinese Translation

基于LLM的预测系统在金融市场和体育结果等现实任务中取得了进展，主要得益于更强的搜索和工具使用。然而，许多系统仍然要求LLM将所有收集到的证据一起阅读并生成最终预测。我们称这种设计为单一预测（Monolithic Prediction）。这种设计可能会掩盖个别证据项如何影响结果，并在竞争结果之间压缩不确定性。我们提出了LEAP（Likelihood Elicitation and Aggregation for Probabilistic forecasting），它重新组织了在预测阶段使用收集证据的方式。LEAP单独检查每个证据项，并提取描述其对目标影响的似然参数。然后，一个明确的先验和一个确定性的概率模型将这些似然结合成后验分布。该程序支持连续的单选和多选预测，同时保留可重复的证据贡献。我们构建了一个涵盖预测、信息搜索和浏览任务的基准，并在我们自己的代理循环和多个代理CLI框架上评估LEAP。在相同证据下，LEAP在大多数模型的预测和校准指标上有所改善，并在先验访问、推理预算和聚合的受控比较中表现更强。

View on arXiv Download PDF AI Translation

cs.AI / 114 / 2609.01345

Cheap Verifiers, Large Blind Spots: Measuring the Reliability Cost of Cost-Saving Cascades

廉价验证者与大盲区：测量节省成本的级联的可靠性成本

Rajput, Dushyant

Abstract

Inference cascades cut cost by answering most queries with a cheap model and escalating a hard tail to a frontier model that acts as verifier. A natural extension closes the loop: fine-tune the cheap student on the verifier's rejections so the escalation rate, and cost, fall each round. We measure this loop on real LLMs and report four findings. First, the verifier's blind spot, the fraction of the student's wrong answers it accepts, is large and moves adversarially: it grows with student capability ($\beta$ from 0.12 to 0.55 as the student scales 0.5B to 32B) and shrinks with verifier capability, so it is worst in the cheap-student, cheap-verifier regime cascades exist to create. Second, buying it away returns the saving: a frontier verifier drives $\beta$ to about 0.05 but then escalates on 46% of hard-MATH queries against a 39% true error rate, paying the frontier price on nearly half of all traffic. Third, naive corrective fine-tuning on the verifier-rejected tail does not improve the small student but degrades and ultimately collapses it, across every teacher we tried (cross-family and same-family), so at this scale the self-improving loop is self-defeating. Fourth, through all of this the cascade's own dashboard, every metric computed through the verifier, reads a flat 3% error while true delivered error swings up to 32%: the system is blind to its own degradation by construction. We then give the theory that explains the blindness, a two-population conservation law, $\epsilon_\infty \lesssim q_0 \beta_0$, under which every in-loop metric improves while true quality does not, and a synthetic study that validates the mechanism. The practical conclusion: the reliability of a self-improving cascade cannot be read from any metric computed through its own verifier.

Chinese Translation

推理级联通过使用廉价模型回答大多数查询并将难以处理的尾部升级到作为验证者的前沿模型来降低成本。一个自然的扩展是闭合循环：对验证者拒绝的情况进行微调廉价学生模型，以便每轮的升级率和成本降低。我们在真实的大型语言模型（LLMs）上测量了这个循环，并报告了四个发现。首先，验证者的盲区，即学生错误答案被接受的比例，较大且呈对抗性变化：随着学生能力的提升（$eta$ 从 0.12 增长到 0.55，当学生规模从 0.5B 增加到 32B），盲区扩大，而随着验证者能力的提升，盲区缩小，因此在廉价学生和廉价验证者的情况下，盲区最严重。第二，购买更好的验证者可以恢复节省：前沿验证者将 $eta$ 降至约 0.05，但在面对 39% 的真实错误率时，对 46% 的困难数学查询仍然升级，几乎在一半的所有流量中支付了前沿价格。第三，对验证者拒绝的尾部进行简单的纠正微调并未改善小型学生模型，反而导致其性能下降并最终崩溃，适用于我们尝试的每个教师模型（跨家族和同家族），因此在这个规模下，自我改进的循环是自我挫败的。第四，在这一切中，级联自身的仪表板，即通过验证者计算的每个指标，显示的错误率始终为 3%，而实际交付的错误率波动高达 32%：系统在结构上对自身的退化视而不见。我们随后给出了解释这种盲区的理论，即一种两人口守恒定律，$ ext{ε}_ ext{∞} extlesssim q_0 eta_0$，在此定律下，每个循环内的指标改善，而真实质量却没有改善，并进行了合成研究以验证该机制。实践结论是：自我改进级联的可靠性无法通过其自身验证者计算的任何指标来判断。

View on arXiv Download PDF AI Translation

cs.AI / 115 / 2609.01353

SymFold: Synergizing Evolutionary and Structural Priors for Accurate Protein Inverse Folding

SymFold：协同进化与结构先验以实现准确的蛋白质逆折叠

Wang, Handong, Qi, Jiaxin, Lai, Baisheng, Huang, Jianqiang

Abstract

Protein inverse folding aims to recover amino acid sequences for a given 3D protein structure, underpinning broad applications such as enzyme engineering and drug discovery.Current methods often follow a serial pipeline, in which a structure encoder predicts a coarse sequence, which is then refined by protein language models (PLMs). However, because PLMs only perform post-hoc sequence edits, the refinement is bounded by the quality of upstream predictions.Thanks to recent multimodal protein language models (MPLMs), we could directly encode structure to generate sequences with pretrained structural knowledge, but we observe that they are not effective for inverse folding. Therefore, we introduce a symmetric dual-path architecture that both leverages PLMs for pretrained sequence evolution knowledge and MPLMs for pretrained structural knowledge to iteratively guide protein sequence generation.Through extensive experiments across standard protein inverse folding benchmarks, our method achieves state-of-the-art performance, surpassing prior approaches, and ablation studies validate the rationale of our symmetric design, revealing a promising direction for the community.

Chinese Translation

蛋白质逆折叠旨在为给定的三维蛋白质结构恢复氨基酸序列，这一过程在酶工程和药物发现等广泛应用中具有重要意义。目前的方法通常遵循串行流程，其中结构编码器预测粗略序列，然后由蛋白质语言模型（PLMs）进行精细化。然而，由于PLMs仅执行事后序列编辑，因此精细化受到上游预测质量的限制。得益于最近的多模态蛋白质语言模型（MPLMs），我们可以直接编码结构以生成具有预训练结构知识的序列，但我们观察到它们在逆折叠方面并不有效。因此，我们提出了一种对称双路径架构，既利用PLMs的预训练序列演化知识，又利用MPLMs的预训练结构知识，以迭代方式指导蛋白质序列生成。通过在标准蛋白质逆折叠基准测试中的广泛实验，我们的方法实现了最先进的性能，超越了之前的方法，消融研究验证了我们对称设计的合理性，为该领域提供了一个有前景的方向。

View on arXiv Download PDF AI Translation

cs.AI / 116 / 2609.01360

EDGE: Error Dependency Graph-Guided Multi-Error Attribution in Multi-Agent LLM Systems

EDGE：基于错误依赖图的多错误归因在多智能体大语言模型系统中的应用

Hou, Jun, Pitre, Priya, Fang, Yi, Wang, Xuan

Abstract

Large language model (LLM) agent failures often contain multiple related errors rather than a single mistake. Existing attribution methods usually identify a responsible agent, step, or root cause, but do not explicitly model dependency between errors. We introduce EDGE, an Error Dependency Graph-guided multi-Error attribution framework. EDGE constructs an error dependency graph from observed error events and validates a reliable causal subset through counterfactual rollout. The inference graph guides a two-stage LLM-as-judge detector for error attribution, and the intervention-validated subgraph provides a more reliable basis for explanation and repair analysis. Experiments on TRAIL and MAST show that EDGE improves category-level multi-error attribution across most evaluated models and settings. Experiments with adapted Who&When-style prompts show that the graph helps across prompting strategies. These results suggest that dependency structure is a useful diagnostic prior for agent failures beyond isolated root-cause prediction.

Chinese Translation

大型语言模型（LLM）代理的失败通常包含多个相关错误，而非单一错误。现有的归因方法通常识别出一个负责的代理、步骤或根本原因，但并未明确建模错误之间的依赖关系。我们提出了EDGE，一个基于错误依赖图的多错误归因框架。EDGE从观察到的错误事件构建错误依赖图，并通过反事实回滚验证一个可靠的因果子集。推理图引导一个两阶段的LLM作为判断者的错误归因检测器，而干预验证的子图为解释和修复分析提供了更可靠的基础。在TRAIL和MAST上的实验表明，EDGE在大多数评估模型和设置中改善了类别级别的多错误归因。使用改编的Who&When风格提示的实验表明，该图在各种提示策略中均有帮助。这些结果表明，依赖结构是诊断代理失败的有用先验，超越了孤立的根本原因预测。

View on arXiv Download PDF AI Translation

cs.AI / 117 / 2609.01408

Neuro-Symbolic Geometric Abstraction (NeuSOGA): From Observations to Symbolic Mathematical Representations

神经符号几何抽象（NeuSOGA）：从观察到符号数学表示

Li, Qingde, Hong, Qingqi, Tian, Jie

Abstract

A fundamental challenge in artificial intelligence is the transformation of observations into explicit symbolic representations suitable for abstraction, interpretation, and reasoning. While modern AI systems achieve remarkable perceptual capabilities through large-scale statistical learning, the resulting knowledge is typically encoded within latent parameters that are difficult to inspect or manipulate analytically. Inspired by Neuro-Symbolic AI and theories of human abstraction, this paper investigates the formation of symbolic mathematical representations from geometric observations. We propose NeuSOGA (Neuro-Symbolic Geometric Abstraction), a framework that progressively transforms observations into topological abstractions, geometric abstractions, and ultimately symbolic mathematical representations. The architecture combines topology-guided structural discovery using Euclidean Distance Transforms, foundation-model perception using Segment Anything, adaptive multi-scale geometric abstraction, and symbolic synthesis through Implicit Area Splines. The resulting representation is an analytical implicit model supporting arbitrary-order smoothness, additive composition, and closed-form evaluation. Unlike neural latent encodings, the generated representation remains interpretable, editable, and mathematically explicit. Experiments on ModelNet40 point clouds, arbitrary-view projections, and segmented optical observations demonstrate that NeuSOGA transforms diverse observations into compact symbolic representations while preserving essential geometric and topological structure across sensing modalities and viewing directions. NeuSOGA provides an interpretable and explainable pathway from observation to symbol and establishes

Chinese Translation

人工智能中的一个基本挑战是将观察转化为适合抽象、解释和推理的明确符号表示。尽管现代人工智能系统通过大规模统计学习实现了显著的感知能力，但所获得的知识通常以潜在参数的形式编码，这些参数难以进行分析性检视或操作。受到神经符号人工智能和人类抽象理论的启发，本文研究了从几何观察中形成符号数学表示的过程。我们提出了NeuSOGA（神经符号几何抽象）框架，该框架逐步将观察转化为拓扑抽象、几何抽象，最终形成符号数学表示。该架构结合了基于拓扑的结构发现（使用欧几里得距离变换）、基础模型感知（使用Segment Anything）、自适应多尺度几何抽象，以及通过隐式区域样条进行的符号合成。生成的表示是一个支持任意阶平滑性、加法组合和封闭形式评估的分析性隐式模型。与神经潜在编码不同，生成的表示保持可解释性、可编辑性和数学明确性。在ModelNet40点云、任意视角投影和分割光学观察的实验中，NeuSOGA能够将多样的观察转化为紧凑的符号表示，同时保留跨感知模态和视角方向的基本几何和拓扑结构。NeuSOGA提供了从观察到符号的可解释和可解释的路径。

View on arXiv Download PDF AI Translation

cs.AI / 118 / 2609.01409

EdiTikZ: Scientific Figure Editing from Revision Trajectories

EdiTikZ：基于修订轨迹的科学图形编辑

Greisinger, Christian, Zhao, Zhixue, Eger, Steffen

Abstract

Vision-language models (VLMs) have shown strong performance in generating scientific figures from text or images. However, producing publication-ready figures requires iterative refinement, making scientific figure editing an important yet largely unexplored task. Existing approaches rely on costly proprietary agentic systems, focus primarily on evaluation, or construct training supervision from synthetically generated edits. Instead, we leverage naturally occurring scientific revision and development trajectories as a scalable source of supervision. To this end, we introduce DaEdiTikZ, the first large-scale dataset of revision-derived scientific figure edits, constructed by mining 391K plausible TikZ edit pairs from arXiv, GitHub, and TeX SE and inferring 781K directed edit instructions with a VLM conditioned on rendered figures and TikZ code. We further introduce DaEdiTikZ-Bench, a human-refined benchmark with 790 instances, and train two compact Qwen3.5-based EdiTikZ models (4B and 9B) by jointly learning reconstruction and editing, followed by reinforcement learning (RL) with complementary rewards for rendered fidelity and edit application. Automatic evaluation places our 9B model above all tested baselines, while human evaluation with 9 annotators and 4,320 ratings places it above GPT-5.6-Sol and on par with Gemini-3.1-Pro. Under severe out-of-distribution shifts, it remains competitive with GPT-5.6-Sol near its 2K training sequence-length regime. Models and datasets will be released.

Chinese Translation

视觉-语言模型（VLMs）在从文本或图像生成科学图形方面表现出色。然而，生成适合出版的图形需要反复的精细调整，使得科学图形编辑成为一个重要但尚未被充分探索的任务。现有的方法依赖于昂贵的专有代理系统，主要集中在评估上，或从合成生成的编辑中构建训练监督。相反，我们利用自然发生的科学修订和发展轨迹作为可扩展的监督来源。为此，我们介绍了DaEdiTikZ，这是第一个基于修订的科学图形编辑的大规模数据集，通过从arXiv、GitHub和TeX SE中挖掘391K个合理的TikZ编辑对，并基于渲染图形和TikZ代码推断出781K个定向编辑指令来构建。我们进一步推出了DaEdiTikZ-Bench，这是一个经过人工精炼的基准数据集，包含790个实例，并通过联合学习重建和编辑训练了两个基于Qwen3.5的紧凑型EdiTikZ模型（4B和9B），随后进行强化学习（RL），以获得渲染保真度和编辑应用的互补奖励。自动评估将我们的9B模型置于所有测试基线之上，而通过9位注释者和4,320条评分的人工评估则将其置于GPT-5.6-Sol之上，并与Gemini-3.1-Pro持平。在严重的分布外变化下，它仍然与GPT-5.6-Sol在其2K训练序列长度范围内保持竞争力。模型和数据集将被发布。

View on arXiv Download PDF AI Translation

cs.AI / 119 / 2609.01466

Parsing the Stream: A Live Trace Model for Long-Horizon Agents and Their Observers

解析流：用于长时间跨度代理及其观察者的实时追踪模型

Pakhomov, Egor, Nijkamp, Erik

Abstract

A long-horizon agent's trace outgrows both of its consumers: the human observer monitoring the run, and the agent itself, whose bounded context the trace must be folded back into. We present a live trace model, an append-only event ledger folded incrementally into typed run state and compiled into per-consumer views, and evaluate it for both consumers against deterministic ground truth. For the observer side, evaluated with an LLM reader as proxy, the compiled view answers monitoring questions using approximately 14x and 15x fewer input tokens (by reader) and at 5-7x lower cost than a budget-capped single-call reading of the raw trace, with higher accuracy (0.85-0.87 versus 0.48). Because the questions were co-designed with the view schema, we treat the token and cost reduction, conditional on schema coverage, as the transferable result. For the agent, on 120-link sequential-dependency tasks, mechanisms that maintain the task's running statistic in per-step state succeed where full-context prompting fails (30/30 versus 8/30 under a clean protocol, n=30, labeled descriptive owing to benchmark-system co-development); a prompt-level scratchpad matches the fold's accuracy at lower cost, and a two-arm decomposition attributes the fold's accuracy to its deterministic aggregate and its cost advantage to its compactness. The fold's remaining value over cheaper alternatives is deterministic auditability and serving the observer from the same state. We derive eleven candidate requirements for trace folding from observed failures and delimit them with an order-sensitive task family on which the fold ceases to help. Code, benchmarks, a regenerable synthetic corpus, and all workbench traces are released.

Chinese Translation

长时间跨度代理的追踪超出了其两个消费者的范围：监控运行的人类观察者和代理自身，其追踪必须被折叠回有限的上下文中。我们提出了一种实时追踪模型，即一个仅追加的事件账本，逐步折叠到类型化的运行状态中，并编译成每个消费者的视图。我们针对这两个消费者对其进行了评估，并与确定性真实值进行了比较。在观察者方面，使用大型语言模型（LLM）阅读器作为代理，编译后的视图在回答监控问题时使用的输入令牌数量比预算限制的原始追踪单次读取少约14倍和15倍，且成本降低了5-7倍，同时准确率更高（0.85-0.87对比0.48）。由于问题与视图模式共同设计，我们将令牌和成本的减少视为可转移的结果，前提是模式覆盖。对于代理，在120链接的顺序依赖任务中，保持任务运行统计信息的机制在逐步状态中成功，而完整上下文提示则失败（在干净协议下30/30对比8/30，n=30，由于基准系统的共同开发而标记为描述性）；一个提示级别的草稿板在较低成本下匹配了折叠的准确性，而一个双臂分解将折叠的准确性归因于其确定性总和，将其成本优势归因于其紧凑性。折叠相较于更便宜的替代方案的剩余价值在于其确定性可审计性以及从相同状态为观察者服务。我们从观察到的失败中推导出十一项追踪折叠的候选要求，并通过一个对顺序敏感的任务家族对其进行界定，在该任务家族上折叠不再有帮助。代码、基准测试、可再生的合成语料库以及所有工作台追踪均已发布。

View on arXiv Download PDF AI Translation

cs.AI / 120 / 2609.01481

Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement

Harness-of-Harness：多天自主软件开发与持续改进

Yan, Haoyang, Su, Min-le, Zhang, Hangfan, Li, Zhanhao, Zhang, Chen, Zhang, Shao, Chen, Yang, Bai, Lei, Hu, Shuyue

Abstract

This paper studies autonomous software development, in which LLM-based coding agents transform high-level requirements into complete, functional, and usable software systems without human intervention. We introduce Harness-of-Harness (HoH), a framework that enables coding agents to continually improve software during autonomous development. HoH operates on existing coding-agent harnesses, and organizes their executions into iterative planning-coding-testing loops. To sustain improvement across loops, HoH balances repair with capability growth, scopes development into small and verifiable increments, separates implementation-time testing from independent evaluation, and constrains verifiable outputs rather than prescribing agent workflows. It progressively exposes deliverables, role-specific tools, and skills, encourages reuse rather than recreation, and maintains versioned project histories. On GameCraft-Bench, FrontierSWE, and ProgramBench, three harness-model pairs (Codex with GPT-5.5, OpenCode with DeepSeek-V4-Pro, and Pi with MiniMax-M3), HoH consistently outperforms the corresponding standalone harnesses, achieving an average relative gain of 52.25 percent and a maximum gain of 82.86 percent after three iterations. In a multi-day deployment with more than 70 iterations, HoH autonomously develops a first-person-shooter game, featuring a coherent storyline, fully implemented core mechanics, human-playable experience, polished visuals and integrated audio. Github: https://github.com/Flesymeb/HarnessOfHarness Project Page: https://flesymeb.github.io/HarnessOfHarness/

Chinese Translation

本文研究了自主软件开发，其中基于大型语言模型（LLM）的编码代理在没有人类干预的情况下将高层次需求转化为完整、功能齐全且可用的软件系统。我们引入了Harness-of-Harness（HoH）框架，该框架使编码代理能够在自主开发过程中持续改进软件。HoH在现有的编码代理工具上运行，并将其执行组织成迭代的规划-编码-测试循环。为了在循环中保持改进，HoH平衡了修复与能力增长，将开发范围缩小到小而可验证的增量，将实施时的测试与独立评估分开，并约束可验证输出而不是规定代理工作流程。它逐步暴露交付成果、角色特定工具和技能，鼓励重用而非重新创建，并维护版本化的项目历史。在GameCraft-Bench、FrontierSWE和ProgramBench上，三个工具-模型对（Codex与GPT-5.5、OpenCode与DeepSeek-V4-Pro、Pi与MiniMax-M3）中，HoH始终优于相应的独立工具，实现了平均相对增益52.25%和最大增益82.86%（经过三次迭代）。在一次超过70次迭代的多天部署中，HoH自主开发了一款第一人称射击游戏，具有连贯的故事情节、完全实现的核心机制、可供人类游玩体验、精美的视觉效果和集成的音频。Github: https://github.com/Flesymeb/HarnessOfHarness 项目页面: https://flesymeb.github.io/HarnessOfHarness/

View on arXiv Download PDF AI Translation

cs.AI / 121 / 2609.01519

When Guardrails Look Effective: Construct Validity Failures in LLM Agent Commerce Evaluation

当护栏看似有效时：大型语言模型代理商业评估中的构念有效性失败

Zhu, Peiying, Chang, Sidi

Abstract

Interactive simulations increasingly evaluate policies in markets populated by language-model agents. Their outputs can look economic---prices, profits, consumer surplus, and welfare---without instantiating the behavior named in the claim. We audit this risk in a multi-turn buyer--seller testbed for configurable hotel transactions. An initial implementation reported welfare gains from two marketplace guardrails of +87.4, +35.0, and +28.8 across a Qwen2.5 1.5B--14B ladder. It also gave guarded and unguarded agents different offer schemas and choice procedures. Holding the schema and buyer chooser fixed changes the paired contrasts to +7.2, -13.9, and +23.8. The four largest 14B single-generation effects averaged +229; after three generations per profile-condition, they averaged +37.6 (95% bootstrap interval [-34.2, 109.3]), while generation residuals account for 49.9% of variation in this post-hoc probe. A seller-incentive check is non-monotone: increasing profit pressure produces less profit than the default seller prompt. Scripted positive controls show why this matters. A profit-maximizing seller already attains first-best welfare, so guardrails mostly redistribute and reduce welfare; they create welfare only when the seller is explicitly programmed to force inefficient bundles. We contribute a construct-validity contract separating incentive validity, protocol isolation, stochastic stability, and welfare accounting, and returning INVALID or INCONCLUSIVE before substantive policy claims. In our case, the original estimate is INVALID under protocol isolation, while the controlled study remains INCONCLUSIVE under incentive validity and stochastic stability. The case does not show that guardrails are ineffective; it shows their apparent value is unidentified until the simulated agents and protocol pass these checks.

Chinese Translation

互动模拟越来越多地用于评估由语言模型代理构成的市场中的政策。它们的输出可能看起来具有经济特征——价格、利润、消费者剩余和福利——但并未体现声明中所提到的行为。我们在一个多轮买卖测试平台上审计了这一风险，该平台用于可配置的酒店交易。初步实施报告了来自两个市场护栏的福利增益，分别为 +87.4、+35.0 和 +28.8，涵盖了 Qwen2.5 1.5B--14B 的梯度。同时，它还为有护栏和无护栏的代理提供了不同的报价模式和选择程序。固定模式和买方选择者的情况下，配对对比的变化为 +7.2、-13.9 和 +23.8。四个最大的 14B 单次生成效应的平均值为 +229；在每个配置条件下进行三次生成后，平均值为 +37.6（95% 自助法区间 [-34.2, 109.3]），而生成残差占该后验探测中变异的 49.9%。卖方激励检查呈非单调性：增加利润压力产生的利润低于默认卖方提示。脚本化的正控制显示了这一点的重要性。一个利润最大化的卖方已经达到了第一最佳福利，因此护栏主要是重新分配并减少福利；只有当卖方被明确编程以强迫低效捆绑时，护栏才会创造福利。我们贡献了一个构念有效性合同，区分激励有效性、协议隔离、随机稳定性和福利核算，并在实质性政策声明之前返回 INVALID 或 INCONCLUSIVE。在我们的案例中，原始估计在协议隔离下为 INVALID，而控制研究在激励有效性和随机稳定性下仍然为 INCONCLUSIVE。该案例并未表明护栏无效；它表明，直到模拟代理和协议通过这些检查，其表面价值才未被识别。

View on arXiv Download PDF AI Translation

cs.AI / 122 / 2609.01526

EvoSCM: Scientific Belief Revision Through Causal Model Evolution and Experimentation

EvoSCM：通过因果模型演化与实验进行科学信念修正

Zhao, Qing, Li, Haowei, Deng, Weijian, Wei, Pengxu, Lin, Liang

Abstract

Scientific agents must learn not only how to reason, but also what to believe. However, existing LLM agents typically express scientific hypotheses in free-form text, leaving their beliefs implicit and difficult to test or revise. We introduce EvoSCM, which equips scientific agents with explicit structural causal models that evolve as new experimental evidence is collected. EvoSCM maintains a population of competing SCM hypotheses, each encoding a candidate causal explanation of the environment, and evolves them through a closed discovery loop. In each round, the agent abduces latent mechanisms from accumulated evidence, designs discriminative interventions, and commits to falsifiable predictions that it tests through experimentation. Discrepancies between prediction and observation are inductively distilled into correction rules that revise the causal structures and mechanisms of each hypothesis, and the agent then deductively validates the revised population against accumulated evidence and structural consistency to guide the next round. We evaluate EvoSCM on DiscoverPhysics, a benchmark requiring agents to uncover the hidden dynamics of noncanonical physical worlds through experimentation. EvoSCM consistently improves scientific discovery over baselines, yielding more accurate explanations and predictions while making more effective use of experimental interactions.

Chinese Translation

科学智能体不仅必须学习如何推理，还必须学习相信什么。然而，现有的大型语言模型（LLM）智能体通常以自由文本的形式表达科学假设，使得它们的信念隐含且难以测试或修正。我们提出了EvoSCM，它为科学智能体提供了明确的结构因果模型，这些模型随着新实验证据的收集而演化。EvoSCM维持一组竞争的结构因果模型（SCM）假设，每个假设编码了环境的候选因果解释，并通过一个封闭的发现循环进行演化。在每一轮中，智能体从累积的证据中推断潜在机制，设计区分性干预，并承诺进行可证伪的预测，通过实验进行测试。预测与观察之间的差异被归纳为修正规则，以修正每个假设的因果结构和机制，随后智能体再通过演绎验证修正后的假设群体与累积证据和结构一致性，以指导下一轮。我们在DiscoverPhysics这一基准上评估EvoSCM，该基准要求智能体通过实验揭示非典型物理世界的隐藏动态。EvoSCM在科学发现方面始终优于基线，提供了更准确的解释和预测，同时更有效地利用实验交互。

View on arXiv Download PDF AI Translation

cs.AI / 123 / 2609.01552

Can LLMs Discover Scientific Laws in Real and Parallel Worlds?

大型语言模型能否在真实和并行世界中发现科学定律？

Huang, Yiming, Liu, Ziche, Wu, Zhuohang, Wang, Yiqian, Cui, Junxia, Zou, Xinkai, Mao, Linjun, Huang, Nan, Yu, Naicheng, Zhu, Kaijie, Ma, Yue, Zhou, Kun, Peng, Letian, Shang, Jingbo

Abstract

Scientific equation discovery has long been central to scientific progress, proceeding through iterative cycles of hypothesis generation, observational testing, and refinement under scientific constraints. As LLM capabilities advance and their role in AI for Science expands, it remains an open problem whether they can genuinely discover scientific laws and how this ability should be evaluated. Existing evaluations, however, often either simplify discovery through synthetic settings or reuse published targets that may already be familiar to LLMs. We therefore introduce SCILAWS-BENCH, a benchmark for scientific law discovery built from published research and real scientific data. It comprises 118 problems drawn from 381 scientific papers, covering 291 candidate laws and roughly 8M real data points across six scientific disciplines. Each problem is instantiated in two complementary settings: (1) SCILAWS-REAL asks models to propose laws from fixed real observations and evaluates held-out predictive fit and scientific validity derived from the source literature, and (2) SCILAWS-PARALLEL asks models to actively query residual-calibrated worlds and recover synthesized hidden laws derived from published forms. This two-setting task design preserves each problem's scientific context while separately evaluating fixed-record law discovery and active recovery of a newly synthesized hidden law. We find that predictive fit can diverge from scientific validity, memorization shapes whether models reproduce or move beyond published formulas, and our best-of-N study reveals a selection bottleneck. Our work provides a paper-grounded benchmark and new empirical perspectives for evaluating AI for scientific discovery. Project page: https://yiyihum.github.io/SciLaws-Bench

Chinese Translation

科学方程的发现长期以来一直是科学进步的核心，通常通过假设生成、观察测试和在科学约束下的迭代完善进行。随着大型语言模型（LLMs）能力的提升及其在科学人工智能中的作用不断扩大，是否能够真正发现科学定律以及如何评估这种能力仍然是一个未解的问题。然而，现有的评估往往通过合成环境简化发现过程，或重用已发布的目标，这些目标可能已经为LLMs所熟知。因此，我们引入了SCILAWS-BENCH，这是一个基于已发布研究和真实科学数据构建的科学定律发现基准。该基准包含来自381篇科学论文的118个问题，涵盖291个候选定律和大约800万个真实数据点，涉及六个科学学科。每个问题在两种互补的设置中实例化：(1) SCILAWS-REAL要求模型从固定的真实观察中提出定律，并评估源文献中得出的预测拟合和科学有效性；(2) SCILAWS-PARALLEL要求模型主动查询残差校准的世界，并恢复从已发布形式中派生的合成隐藏定律。这种双设置任务设计保留了每个问题的科学背景，同时分别评估固定记录的定律发现和新合成隐藏定律的主动恢复。我们发现，预测拟合可能与科学有效性存在偏差，记忆影响模型是否重现或超越已发布的公式，而我们的最佳选择研究揭示了选择瓶颈。我们的工作提供了一个基于论文的基准和评估科学发现的人工智能的新实证视角。项目页面：https://yiyihum.github.io/SciLaws-Bench

View on arXiv Download PDF AI Translation

cs.AI / 124 / 2609.01567

Selective Agent Guidance via Entropy: Learning Autonomous Policies from Imperfect VLM Teachers

通过熵进行选择性代理指导：从不完美的视觉语言模型教师中学习自主策略

Merler, Matteo, Bonetta, Giovanni, Zago, Davide, Cancelliere, Rossella, Magnini, Bernardo

Abstract

Vision-Language Models (VLMs) provide useful priors for interactive decision-making, but using them directly as policies is expensive and brittle: they must be queried at every step, do not improve from environment interaction, and can repeat systematic errors. We study how to learn a cheap autonomous policy from an online, expensive, and imperfect but informative VLM teacher. We propose SAGE (Selective Agent Guidance via Entropy), a framework that queries a VLM only when the learner is uncertain, executes the suggested action during training, and distills guidance into a lightweight Reinforcement Learning (RL) policy. Because VLM advice is not always reliable, SAGE can weight teacher-action distillation using environment-derived advantages rather than treating all suggestions as equally useful. Across sparse-reward visual reasoning and navigation tasks, SAGE learns policies that act without VLM guidance at evaluation time and improves over unguided RL in several environments, including settings where the learned policy exceeds its VLM teacher. The results show that selective guidance is most beneficial when the VLM can help the agent discover high-reward trajectories, and less useful when unguided exploration already succeeds or teacher actions do not lead to informative experience. SAGE also reduces VLM usage by prompting the teacher only on a fraction of training steps and requiring no VLM calls at deployment. Overall, our results suggest that VLMs don't need to be used as fixed policies to be useful; they can instead act as temporary, imperfect sources of guidance whose value is tested and internalized through interaction.

Chinese Translation

视觉语言模型（VLMs）为互动决策提供了有用的先验知识，但直接将其作为策略使用既昂贵又脆弱：每一步都必须查询它们，无法通过环境交互进行改进，并且可能重复系统性错误。我们研究如何从一个在线的、昂贵的、不完美但信息丰富的VLM教师中学习一个廉价的自主策略。我们提出了SAGE（通过熵进行选择性代理指导），这是一个框架，仅在学习者不确定时查询VLM，在训练期间执行建议的动作，并将指导提炼为轻量级的强化学习（RL）策略。由于VLM的建议并不总是可靠，SAGE可以使用环境派生的优势来加权教师动作的提炼，而不是将所有建议视为同样有用。在稀疏奖励的视觉推理和导航任务中，SAGE学习的策略在评估时能够在没有VLM指导的情况下行动，并在多个环境中优于无指导的RL，包括学习的策略超越其VLM教师的设置。结果表明，当VLM能够帮助代理发现高奖励轨迹时，选择性指导最为有利，而在无指导探索已经成功或教师动作未能带来有益经验时则不太有用。SAGE还通过仅在部分训练步骤中提示教师并在部署时不需要VLM调用来减少VLM的使用。总体而言，我们的结果表明，VLM不需要作为固定策略来发挥作用；它们可以作为临时的不完美指导来源，其价值通过交互进行测试和内化。

View on arXiv Download PDF AI Translation

计算语言学 (Computation and Language)

127

cs.CL / 1 / 2609.00014

Behaviorally Grounded User Profiles from the Wild for Personalized Alignment and Multi-Perspective Reasoning

基于行为的用户画像：来自真实环境的个性化对齐与多视角推理

Li, Yuxuan, Zhong, Victor, Kamalloo, Ehsan

Abstract

Persona-driven techniques increasingly adapt large language models (LLMs) to diverse contexts. However, existing methods predominantly rely on rigid, synthetic personas that flatten individual variation, rely on stereotypes, and miss the nuanced signals driving actual human preferences. We introduce profile behavioral grounding, a framework for extracting open-ended, high-fidelity user profiles directly from authentic, anonymized social media posts. We evaluate these profiles across two paradigms: train-time personalization via supervised finetuning (SFT) and non-parametric test-time multi-perspective reasoning. Across complex recommendation and open-ended query benchmarks, behaviorally grounded profiles consistently improve base models and outperform synthetic profile baselines, driving stronger parametric alignment and enabling richer, multifaceted reasoning. Our findings establish open-ended, behavior-derived profiles as a highly diverse and effective foundation for the next generation of personalized language systems. Our code base is available at https://github.com/ServiceNow/behavior-grounding.

Chinese Translation

以角色为驱动的技术日益将大型语言模型（LLMs）适应于多样化的情境。然而，现有方法主要依赖于僵化的合成角色，这种方法忽视了个体差异，依赖于刻板印象，并且未能捕捉到驱动实际人类偏好的细微信号。我们提出了画像行为基础框架，旨在直接从真实的匿名社交媒体帖子中提取开放式、高保真的用户画像。我们在两个范式中评估这些画像：通过监督微调（SFT）进行的训练时个性化和非参数测试时的多视角推理。在复杂的推荐和开放式查询基准测试中，基于行为的画像始终提升基础模型的性能，并超越合成画像基线，推动更强的参数对齐，并实现更丰富的多维推理。我们的研究结果确立了开放式、基于行为的画像作为下一代个性化语言系统的高度多样化和有效的基础。我们的代码库可在 https://github.com/ServiceNow/behavior-grounding 获取。

View on arXiv Download PDF AI Translation

cs.CL / 2 / 2609.00038

trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories

轨迹评估者：仅基于结果的LLM评估在代理轨迹上遗漏了什么

Mohammadi, Hadi

Abstract

Outcome-only evaluation is the production default for LLM agents: show a judge the request and the final reply and ask whether it was handled well. The metric is structurally blind to an agent that reaches the right answer the wrong way. We measure that blind spot where ground truth is known by construction: a deterministic tool-using support-desk environment, a scripted oracle policy that always solves it, and a fault injector that breaks exactly one thing at a known step, stratifying faults by whether the customer-visible outcome survived (silent) or not (loud). Five judges (programmatic rules, outcome-only, step-rubric at two model sizes, and a self-consistency ensemble) are scored on detection, step localisation, fault typing, calibration, and cost over 400 trajectories. The outcome-only judge catches 84% of loud faults but 45% of silent ones while flagging 33% of correct trajectories; a step-rubric judge reaches 77% silent recall with zero false alarms at 3x the cost. No judge reads the final reply: an invented promise appended to an otherwise perfect trajectory evades the rules entirely and the step judge 82% of the time, and self-consistency triples cost while improving nothing. We argue that judge evaluations must stratify recall by outcome survival, and release the environment, the injector, all raw verdicts, and an analysis pipeline that rebuilds every number offline.

Chinese Translation

仅基于结果的评估是LLM代理的默认生产模式：向评估者展示请求和最终回复，并询问处理是否得当。该指标在结构上对以错误方式达到正确答案的代理视而不见。我们测量了这一盲点，其中真实情况是通过构建已知的：一个确定性的工具使用支持台环境，一个始终能解决问题的脚本化神谕策略，以及一个在已知步骤中精确破坏一件事的故障注入器，按客户可见结果是否存活（静默）或未存活（响亮）对故障进行分层。五个评估者（程序规则、仅基于结果、两个模型规模的步骤评分标准，以及自一致性集成）在400条轨迹上进行检测、步骤定位、故障分类、标定和成本评分。仅基于结果的评估者捕捉到84%的响亮故障，但仅捕捉到45%的静默故障，同时标记了33%的正确轨迹；步骤评分标准的评估者在3倍成本下达到了77%的静默召回率且没有误报。没有评估者阅读最终回复：附加在其他完美轨迹上的虚构承诺完全规避了规则，步骤评估者在82%的情况下也未能识别，而自一致性则在提高任何效果的同时使成本增加了三倍。我们认为评估者的评估必须按结果存活进行分层召回，并释放环境、注入器、所有原始裁决和重建每个数字的离线分析管道。

View on arXiv Download PDF AI Translation

cs.CL / 3 / 2609.00048

GUI-CC: Benchmarking Contextual Consistency of GUI World Models as Agent Environments

GUI-CC：作为智能体环境的GUI世界模型的上下文一致性基准测试

Fu, Lin, Yang, Zheyuan, Zhang, Tianhui, Wei, Jinbiao, Gan, Guo, Liu, Boxu, Zhao, Yilun, Rong, Yu

Abstract

GUI world models are increasingly evaluated as one-step next-screen predictors, yet their intended use is often as multi-step environments for GUI agents. This mismatch leaves a key requirement under-tested: generated states must remain contextually consistent when they are repeatedly reused for future interaction. We introduce GUI-CC, a benchmark that evaluates contextual consistency of GUI world models as agent environments rather than isolated next-screen predictors. GUI-CC contains two complementary tracks: an offline reference-action track that rolls models along real mobile GUI trajectories, and an online agent-loop track that lets fixed probing agents interact with model-generated UIs. We construct 500 offline trajectory tasks from GUIOdyssey and 200 emulator-verified online tasks across 30 mobile apps. GUI-CC evaluates transition fidelity, transition plausibility, contextual consistency, and task progress. Experiments show that plausible single-step generation does not guarantee reliable environment simulation: current models often produce usable-looking screens while failing to preserve task-relevant context or support executable multi-step rollouts.

Chinese Translation

GUI世界模型越来越多地被评估为一步下一屏的预测器，但它们的预期用途通常是作为GUI智能体的多步环境。这种不匹配导致一个关键要求未得到充分测试：生成的状态在未来交互中重复使用时必须保持上下文一致性。我们引入了GUI-CC，这是一个基准测试，评估GUI世界模型作为智能体环境的上下文一致性，而不是孤立的下一屏预测器。GUI-CC包含两个互补的轨道：一个离线参考动作轨道，该轨道沿着真实的移动GUI轨迹滚动模型，以及一个在线智能体循环轨道，该轨道允许固定探测智能体与模型生成的用户界面进行交互。我们从GUIOdyssey构建了500个离线轨迹任务，并在30个移动应用中验证了200个在线任务。GUI-CC评估过渡保真度、过渡合理性、上下文一致性和任务进展。实验表明，合理的单步生成并不能保证可靠的环境模拟：当前模型通常生成看似可用的界面，但未能保持与任务相关的上下文或支持可执行的多步展开。

View on arXiv Download PDF AI Translation

cs.CL / 4 / 2609.00051

From Detection to Refusal: Safer LLMs via Circuit-Guided Weight Scaling

从检测到拒绝：通过电路引导的权重缩放实现更安全的语言模型

Chu, Kuan-Lin, Sun, Chung-En, Weng, Tsui-Wei

Abstract

Despite extensive alignment efforts, Large Language Models (LLMs) remain vulnerable to generating unsafe content under adversarial prompting, yet the internal mechanisms by which safety behaviors are implemented remain poorly understood. We study LLM safety from a mechanistic interpretability perspective and characterize a multi-stage *safety circuit* that organizes refusal behavior, consisting of (i) $\textbf{Harmful Detection Heads}$ that respond to harmful inputs, (ii) $\textbf{Safety Neurons}$ that mediate and stabilize safety signals in the residual stream, and (iii) $\textbf{Refusal Heads}$ that translate these signals into safe response generation. Using targeted attention-head and neuron-level interventions, we provide causal evidence consistent with this circuit organization, showing that suppressing upstream Harmful Detection Heads disrupts downstream refusal behavior and that safety neurons mediate this interaction. We validate that this decomposition recurs across multiple LLM architectures and adversarial attack settings, and use simple, architecture-preserving weight scaling as a mechanistic probe to test its functional relevance. Across six LLMs, circuit-guided scaling improves safety rates under attacks by 26.5%, while incurring only a 1.7% accuracy drop across four standard benchmarks. Overall, our results support a circuit-level interpretation of LLM safety and suggest that mechanistic abstractions can reveal stable and transferable patterns underlying aligned behavior.

Chinese Translation

尽管进行了广泛的对齐努力，大型语言模型（LLMs）在对抗性提示下仍然容易生成不安全内容，而安全行为的内部机制仍然不够清楚。我们从机制可解释性的角度研究LLM的安全性，并描述一个多阶段的*安全电路*，该电路组织拒绝行为，包括（i）$ extbf{有害检测头}$，用于响应有害输入，（ii）$ extbf{安全神经元}$，用于调节和稳定残差流中的安全信号，以及（iii）$ extbf{拒绝头}$，用于将这些信号转化为安全的响应生成。通过针对性地干预注意力头和神经元，我们提供了与该电路组织一致的因果证据，表明抑制上游的有害检测头会干扰下游的拒绝行为，而安全神经元在这一互动中起到调节作用。我们验证了这种分解在多个LLM架构和对抗攻击设置中反复出现，并使用简单的、保持架构的权重缩放作为机制探针来测试其功能相关性。在六个LLM中，电路引导的缩放在攻击下提高了26.5%的安全率，同时在四个标准基准测试中仅造成1.7%的准确率下降。总体而言，我们的结果支持LLM安全性的电路级解释，并表明机制抽象可以揭示对齐行为背后的稳定和可转移模式。

View on arXiv Download PDF AI Translation

cs.CL / 5 / 2609.00055

Zero-Shot Respiratory Sound Classification through LLM-Augmented Audio-Text Alignment

通过 LLM 增强的音频-文本对齐实现零样本呼吸声分类

İlerisoy, Mustafa Talha, Pham, Hung Manh, Funk, Mathias, Pechenizkiy, Mykola, Saeed, Aaqib

Abstract

Self-supervised respiratory encoders lack semantic grounding in clinical domain needed for zero-shot inference, limiting their utility without task-specific labeled data. We propose a framework that aligns these encoders with medical terminology in a shared latent space turning them into a zero-shot-capable foundation model. To address paired data scarcity, we use a medical LLM to synthesize structured reports from metadata, creating dense semantic anchors for contrastive learning. Our training combines a sigmoid-based contrastive loss with encoder's native SSL objective and similarity-aware negative sampling to sharpen pathological boundaries. Across 9 tasks on 6 datasets, our method achieves a 61.3% mean zero-shot AUC, surpassing CLAP (51.4%) and Qwen2-Audio (54.9%) while reaching the highest linear probing AUC (71.6%) with only 43% of data used by full-scale baselines, showing that structured semantic alignment outperforms large-scale, general-purpose models in clinical diagnostics.

Chinese Translation

自监督的呼吸编码器缺乏在临床领域所需的语义基础，限制了其在没有特定任务标注数据情况下的实用性。我们提出了一种框架，将这些编码器与医学术语在共享潜在空间中对齐，从而将其转变为具备零样本能力的基础模型。为了解决配对数据稀缺的问题，我们使用医学 LLM 从元数据合成结构化报告，创建密集的语义锚点以进行对比学习。我们的训练结合了基于 sigmoid 的对比损失、编码器的原生自监督学习目标以及相似性感知的负采样，以锐化病理边界。在 6 个数据集上的 9 个任务中，我们的方法实现了 61.3% 的平均零样本 AUC，超越了 CLAP（51.4%）和 Qwen2-Audio（54.9%），同时在仅使用全规模基准数据的 43% 的情况下达到了最高的线性探测 AUC（71.6%），显示出结构化语义对齐在临床诊断中优于大规模通用模型。

View on arXiv Download PDF AI Translation

cs.CL / 6 / 2609.00057

ValueGraph: Value-Signal Guided Graph Pre-training for Contextualized User Representation

ValueGraph：基于价值信号的图预训练用于上下文化用户表示

Han, Yitong, Gao, Wei, Zhao, Yi, Bhattacharya, Prasanta, Zeng, Fengzhu, Amanlou, Mohammad

Abstract

Value signals are aggregated user-level moral representations that capture users' inferred value-related tendencies from their online discourse. User behavior on social media is shaped not only by what users say or whom they interact with, but also by the value signal through which they express attitudes. Existing user representation methods largely miss this value-relevant dimension. We propose ValueGraph, a graph pre-training framework that uses automatically inferred moral-value signals as noisy auxiliary signals for contextualized user representation. From post-reply graphs, ValueGraph learns semantic and structural representations and further aligns users through relative value similarity with contrastive and clustering objectives. Rather than treating inferred values as gold psychological labels, ValueGraph uses them as soft constraints for representation learning. Experiments on stance detection and twitter bot detection show consistent gains over strong text-based, graph-based, and text-only LLM baselines, highlighting value-signal guidance as a useful inductive bias for socially informed user modeling.

Chinese Translation

价值信号是聚合的用户层面道德表征，捕捉用户在其在线话语中推断出的与价值相关的倾向。社交媒体上的用户行为不仅受到用户所说内容或互动对象的影响，还受到他们表达态度的价值信号的影响。现有的用户表示方法在很大程度上忽视了这一与价值相关的维度。我们提出了ValueGraph，一种图预训练框架，利用自动推断的道德价值信号作为噪声辅助信号，用于上下文化用户表示。ValueGraph从帖子-回复图中学习语义和结构表示，并通过对比和聚类目标进一步通过相对价值相似性对用户进行对齐。ValueGraph并不将推断出的价值视为黄金心理标签，而是将其作为表示学习的软约束。在立场检测和推特机器人检测的实验中，ValueGraph在强大的基于文本、基于图和仅基于文本的LLM基线之上显示出一致的提升，突显了价值信号引导作为社会信息化用户建模的有用归纳偏差。

View on arXiv Download PDF AI Translation

cs.CL / 7 / 2609.00058

CUDA-Harness: Harnessing Agentic CUDA Kernel Generation and Optimization from Natural Language

CUDA-Harness：从自然语言中获取自主CUDA内核生成与优化

Fan, Qi, Zou, An, Ma, Yehan

Abstract

Developing high-performance CUDA kernels demands specialized knowledge in algorithm implementation, correctness validation, and hardware-aware parallel optimization, creating a substantial expertise barrier and making generating CUDA kernels directly from natural language (Text2CUDA) essential. Meanwhile, the general-purpose code generation capability of Large Language Models (LLMs) prompts a series of works exploring LLM-based CUDA kernel generation. They mainly focus on transpilation from high-level frameworks such as PyTorch to CUDA (Torch2CUDA) rather than Text2CUDA, where models must understand the high-level input semantics and handle low-level kernel implementation and validation. Additionally, these methods are vulnerable to reward hacking due to reliance on predefined test inputs. In this paper, we propose CUDA-Harness, a framework for harnessing agentic CUDA kernel generation and optimization from natural language. Specifically, we introduce Intermediate-Structured Generation to connect high-level semantic understanding with low-level kernel generation. To dilute reward hacking in Text2CUDA, we construct Synthesis-Based Verification to provide isolated test data and progressive validation. Furthermore, we propose Feedback-Adaptive Evolution, a kernel evolution strategy that prioritizes correctness while optimizing performance. Finally, through extensive experiments, we demonstrate the effectiveness of CUDA-Harness, with further evaluations illustrating generalization across LLMs, hardware platforms, and to C-to-CUDA transpilation.

Chinese Translation

开发高性能的CUDA内核需要在算法实现、正确性验证和硬件感知并行优化方面具备专业知识，这造成了显著的专业壁垒，使得直接从自然语言生成CUDA内核（Text2CUDA）变得至关重要。同时，大型语言模型（LLMs）的一般代码生成能力促使了一系列探索基于LLM的CUDA内核生成的研究。这些研究主要集中在从高层框架（如PyTorch）到CUDA的转译（Torch2CUDA），而非Text2CUDA，在此过程中，模型必须理解高层输入的语义并处理低层内核的实现与验证。此外，由于依赖于预定义测试输入，这些方法容易受到奖励黑客攻击。在本文中，我们提出了CUDA-Harness，一个从自然语言中获取自主CUDA内核生成与优化的框架。具体而言，我们引入了中间结构生成（Intermediate-Structured Generation），以连接高层语义理解与低层内核生成。为了减少Text2CUDA中的奖励黑客问题，我们构建了基于合成的验证（Synthesis-Based Verification），以提供独立的测试数据和渐进式验证。此外，我们提出了反馈自适应进化（Feedback-Adaptive Evolution），这是一种优先考虑正确性同时优化性能的内核进化策略。最后，通过广泛的实验，我们展示了CUDA-Harness的有效性，进一步的评估表明其在LLMs、硬件平台以及C到CUDA转译中的泛化能力。

View on arXiv Download PDF AI Translation

cs.CL / 8 / 2609.00062

RePro: Proof-Verified Benchmark Rewriting for Reliable Evaluation of LLM Mathematical Problem Solving

RePro：用于可靠评估大型语言模型数学问题解决的证明验证基准重写

Zhou, Xiyuan, Li, Zhuoqi, Wang, Xinlei, He, Yirui, Wu, Yuhao, Cheng, Yuheng, Xu, Yan, Zhao, Junhua, Gu, Jinjin

Abstract

Data contamination undermines the reliable evaluation of large language models (LLMs) on mathematical problem solving. While rewriting-based evaluation mitigates memorization, existing methods lack guarantees of problem validity and answer correctness. We propose Proof-Verified Benchmark Rewriting (RePro), the first framework to integrate Lean-oriented neural automated theorem provers (ATPs) into benchmark rewriting, which rewrites problems and regenerates answers with correctness ensured by Lean-verified proofs. Experiments on GSM8K and MATH show that RePro's retained rewritten instances achieve 100% well-definedness, feasibility, and answer correctness, while existing methods still produce invalid or incorrect instances. Moreover, several models exhibit accuracy drops on proof-verified rewritten benchmarks, suggesting that their performance is sensitive to surface-level and structural variations and may partly reflect memorization effects. Our source code and data are available at https://github.com/AI4Engi/RePro.

Chinese Translation

数据污染削弱了对大型语言模型（LLMs）在数学问题解决方面的可靠评估。虽然基于重写的评估可以减轻记忆效应，但现有方法缺乏问题有效性和答案正确性的保障。我们提出了证明验证基准重写（RePro），这是第一个将面向 Lean 的神经自动定理证明器（ATPs）集成到基准重写中的框架，该框架通过 Lean 验证的证明确保重写问题和再生答案的正确性。在 GSM8K 和 MATH 上的实验表明，RePro 保留的重写实例实现了 100% 的明确性、可行性和答案正确性，而现有方法仍然产生无效或不正确的实例。此外，多个模型在经过证明验证的重写基准上表现出准确率下降，表明它们的性能对表面和结构变化敏感，并可能部分反映记忆效应。我们的源代码和数据可在 https://github.com/AI4Engi/RePro 获取。

View on arXiv Download PDF AI Translation

cs.CL / 9 / 2609.00063

Medical Causal Hypothesis Verification with Large Language Models

利用大型语言模型进行医学因果假设验证

Ahmed, Safiyyah, Ansari, Abrar, Islam, Md Aminul, Zheleva, Elena

Abstract

The growing use of large language models (LLMs) for search and information retrieval underscores the need to evaluate their reliability in high-stakes domains such as healthcare. Although LLMs can effectively answer questions about diseases, symptoms, and treatments, their ability to accurately assess causal relationships and ground their conclusions in verified scientific evidence remains unclear. Here, we present a preliminary, small-scale study that investigates the accuracy of LLMs in evaluating causal medical claims and supporting them with peer-reviewed research. We propose an evaluation framework for causal hypothesis verification that can be used to systematically track the performance of existing and future LLMs. We assess the performance of eight LLMs on 17 medical causal hypotheses to evaluate whether they can reliably verify these hypotheses using scientific evidence from the literature. We systematically annotate the scientific evidence they provide according to six criteria (a total of 1,067 annotation points) and assess them with nine evaluation metrics. Our analysis shows that while LLMs exhibit strong recall, they often perform poorly at providing valid scientific articles and evidence for support and at rejecting unsupported hypotheses. These findings highlight a critical limitation of current LLMs, as they cannot yet be trusted fully to verify causal relationships from the biomedical literature. This work underscores the need for rigorous evaluation before using LLMs for search and retrieval in healthcare settings.

Chinese Translation

大型语言模型（LLMs）在搜索和信息检索中的日益广泛应用凸显了评估其在医疗等高风险领域可靠性的必要性。尽管LLMs能够有效回答有关疾病、症状和治疗的问题，但它们在准确评估因果关系及其结论是否基于经过验证的科学证据方面的能力仍不明确。在此，我们呈现了一项初步的小规模研究，调查LLMs在评估因果医学主张及以同行评审研究支持这些主张方面的准确性。我们提出了一种因果假设验证的评估框架，可以系统地跟踪现有和未来LLMs的性能。我们评估了八个LLMs在17个医学因果假设上的表现，以评估它们是否能够可靠地使用文献中的科学证据来验证这些假设。我们根据六个标准（共1,067个注释点）系统地注释它们提供的科学证据，并使用九个评估指标进行评估。我们的分析表明，尽管LLMs表现出较强的召回率，但它们在提供有效的科学文章和支持证据以及拒绝不支持的假设方面的表现往往较差。这些发现突显了当前LLMs的一个关键局限性，因为它们尚不能完全信任以验证生物医学文献中的因果关系。这项工作强调了在医疗环境中使用LLMs进行搜索和检索之前进行严格评估的必要性。

View on arXiv Download PDF AI Translation

cs.CL / 10 / 2609.00065

Scientific Agent Skills: A Library of Procedural Knowledge for Research Agents

科学代理技能：研究代理的程序性知识库

Kassis, Timothy, Agarwal, Vinayak, He, Yuhuan, Patel, Darshil, Brueckner, Aubrey M.

Abstract

A language-model agent asked to analyse an experiment will usually return working code. Whether the analysis is defensible is a different question. A defensible analysis depends on procedural choices: which test the field accepts, which identifier namespace is authoritative, and which caveats must accompany a result. We present Scientific Agent Skills, an open library of 163 such procedures in 16 areas of practice, including genomics, cheminformatics, medical imaging, study design and scientific communication. Each skill is a directory built around a versioned, human-readable instruction file. An agent loads the file only when a task calls for it; the directory often also contains reference material and runnable scripts. We report no task-level evaluation and no host selection rate. Openly licensed and available at https://github.com/K-Dense-AI/scientific-agent-skills.

Chinese Translation

当一个语言模型代理被要求分析一个实验时，通常会返回可工作的代码。然而，分析是否具有可辩护性是一个不同的问题。可辩护的分析依赖于程序性选择：领域接受哪个测试，哪个标识符命名空间是权威的，以及哪些警告必须伴随结果。我们提出了科学代理技能（Scientific Agent Skills），这是一个开放的库，包含16个实践领域中的163个此类程序，包括基因组学、化学信息学、医学成像、研究设计和科学传播。每个技能都是围绕一个版本化的、可读的人类指令文件构建的目录。代理仅在任务需要时加载该文件；该目录通常还包含参考材料和可运行的脚本。我们报告没有任务级评估和没有主机选择率。该库开放授权，地址为 https://github.com/K-Dense-AI/scientific-agent-skills。

View on arXiv Download PDF AI Translation

cs.CL / 11 / 2609.00066

OCGQuant: Outlier-Companion Grouping for NVFP4 Quantization

OCGQuant：用于NVFP4量化的异常伴侣分组

Yao, Yishan, Li, Binjun, Yi, Hanling, Li, Pengyu, Liu, Xiaoqing, Yang, Zihan, Yu, Xiaotian, Yu, Zhiwen

Abstract

NVFP4 is an efficient microscaling format for low-bit inference, but activation outliers can still degrade quantization accuracy within NVFP4 blocks. Within each quantization block, large activations can dominate the block scale, increasing the quantization error of the remaining values sharing the same scale. Existing post-training quantization (PTQ) methods mitigate outlier errors through strategies such as mixed precision, rotation, or residual compensation, but these approaches are either not specifically tailored to NVFP4 or introduce additional computation. In this work, we revisit NVFP4 from a channel-grouping perspective and define the reducible error incurred by remaining block values under the scale set by the block maximum as Collateral Quantization Error. Based on this insight, we propose OCGQuant, a post-training quantization method centered on Outlier-Companion Grouping (OCG), which adaptively pairs outlier channels with low-magnitude companion channels to improve NVFP4 activation block composition. Experiments on Llama3 and Qwen3 show that OCGQuant achieves the lowest WikiText-2 perplexity and highest average downstream accuracy among evaluated PTQ methods, while maintaining prefill speedup close to RTN and matching its peak decoding memory. Code is available at https://github.com/Eshamont/OCGQuant.

Chinese Translation

NVFP4是一种高效的微缩格式，用于低比特推理，但激活异常值仍可能降低NVFP4块内的量化精度。在每个量化块内，大激活值可能主导块的尺度，从而增加共享同一尺度的其余值的量化误差。现有的后训练量化（PTQ）方法通过混合精度、旋转或残差补偿等策略来减轻异常值误差，但这些方法要么并未专门针对NVFP4进行优化，要么引入了额外的计算。在本研究中，我们从通道分组的角度重新审视NVFP4，并定义了在块最大值设定的尺度下，剩余块值所产生的可减少误差为附带量化误差（Collateral Quantization Error）。基于这一见解，我们提出了OCGQuant，一种以异常伴侣分组（Outlier-Companion Grouping, OCG）为中心的后训练量化方法，该方法自适应地将异常通道与低幅度伴侣通道配对，以改善NVFP4激活块的组成。在Llama3和Qwen3上的实验表明，OCGQuant在评估的PTQ方法中实现了最低的WikiText-2困惑度和最高的平均下游准确率，同时保持接近RTN的预填充加速，并匹配其峰值解码内存。代码可在https://github.com/Eshamont/OCGQuant获取。

View on arXiv Download PDF AI Translation

cs.CL / 12 / 2609.00067

Do Multimodal LLMs See Before They Read? Diagnosing Contextual Sycophancy

多模态大语言模型在阅读之前是否能够看见？诊断上下文谄媚现象

Lai, Yi-Cheng, Huang, Hen-Hsen

Abstract

External text can override conflicting image evidence in multimodal large language models, a failure we call multimodal contextual sycophancy. We introduce a 998-case diagnostic that independently varies visual evidence, commonsense priors, and external text, and probe when this failure arises by moving the information boundary around a context-blind visual witness. On abnormal images paired with Gemini-generated false text, GPT-5.1 scores 7.9% under joint conditioning, 49.7% when the context-blind witness report is scored directly, 63.7% under a matched two-call witness-arbiter pipeline that exposes the witness to the text, and 84.2% under System-2 Visual Arbitration (S2VA), which withholds the text from the witness. Across six models, S2VA improves over the direct witness report by 19.7 to 44.1 points, with all paired 95% confidence intervals excluding zero. The best information boundary is not uniform: textual context scaffolds some models, and a GPT-4o-regenerated subset changes the relative ordering of joint conditioning, Witness-Only, and S2VA. Contextual sycophancy is therefore sensitive to when text is introduced, as well as to the model and context source.

Chinese Translation

外部文本可以覆盖多模态大语言模型中相互矛盾的图像证据，这种失败现象我们称之为多模态上下文谄媚。我们引入了一个包含998个案例的诊断，独立变化视觉证据、常识先验和外部文本，并通过移动信息边界来探测这一失败何时发生，围绕一个不受上下文影响的视觉证人进行测试。在与Gemini生成的虚假文本配对的异常图像中，GPT-5.1在联合条件下得分为7.9%，在直接评分不受上下文影响的证人报告时得分为49.7%，在一个匹配的双通道证人-仲裁者管道下得分为63.7%（该管道将文本暴露给证人），而在系统-2视觉仲裁（S2VA）下得分为84.2%（该方法不向证人提供文本）。在六个模型中，S2VA的得分比直接证人报告提高了19.7到44.1分，所有配对的95%置信区间均不包括零。最佳信息边界并不均匀：文本上下文为某些模型提供了支撑，而GPT-4o再生的子集改变了联合条件、仅证人和S2VA的相对排序。因此，上下文谄媚对文本引入的时机、模型和上下文来源都非常敏感。

View on arXiv Download PDF AI Translation

cs.CL / 13 / 2609.00068

Life Operators: a self-evolving framework for multiscale life modelling

生命算子：一个自我演化的多尺度生命建模框架

Wang, Shuo, Guo, Yike

Abstract

Medical AI is moving beyond recognition towards clinical dialogue and longitudinal prediction. Yet a central question remains: how would a patient's state change under intervention? Statistical models learn future observations, whereas mechanistic models describe selected processes. Neither provides a common framework for representing patient state, coupling scales or revising failed assumptions. We propose Life Operators: task-bounded mappings that define three scientific roles. Perception operators infer task-relevant biological states from multimodal observations, Evolution operators propagate these states under natural or intervention-conditioned dynamics, and Generation operators map them to measurable signals. Each role may be realised by equations, statistical models, neural networks or hybrids. Bridge operators connect components with different variables, scales and time steps. Selected operators and bridges form task-specific Operator Graphs containing the smallest set of states and mechanisms sufficient for a declared claim. This modular structure also makes scientific revision localisable. An AI co-scientist may propose changes to states, operators, bridges or graph structure, while independent evidence determines which variants are retained, restricted or retired. Over time, validated components could accumulate into broader multiscale models of the human body and provide a computational foundation for medical artificial superintelligence.

Chinese Translation

医学人工智能正从识别向临床对话和纵向预测发展。然而，一个核心问题仍然存在：在干预下，患者的状态将如何变化？统计模型学习未来的观察结果，而机制模型描述选定的过程。两者都未提供一个共同的框架来表示患者状态、耦合尺度或修正失败的假设。我们提出了生命算子（Life Operators）：定义三种科学角色的任务绑定映射。感知算子（Perception operators）从多模态观察中推断与任务相关的生物状态，演化算子（Evolution operators）在自然或干预条件下传播这些状态，而生成算子（Generation operators）将其映射到可测量的信号。每个角色可以通过方程、统计模型、神经网络或混合体来实现。桥接算子（Bridge operators）连接具有不同变量、尺度和时间步长的组件。选定的算子和桥接形成特定任务的算子图（Operator Graphs），包含足够支持声明的状态和机制的最小集合。这种模块化结构也使得科学修订具有可定位性。人工智能共同科学家可以提出对状态、算子、桥接或图结构的更改，而独立证据决定保留、限制或淘汰哪些变体。随着时间的推移，经过验证的组件可以累积成更广泛的人体多尺度模型，并为医学人工超级智能提供计算基础。

View on arXiv Download PDF AI Translation

cs.CL / 14 / 2609.00069

Auditing Harness Tampering in Self-Improving Agents

审计自我改进代理中的工具篡改

Wang, Xing, Zhang, Xiaoyi, Shao, Jie

Abstract

Self-improving agents iteratively modify their own harness to push the frontier of their performance. However, such modifications can produce illusory performance gains or compromise integrity constraints such as authorization, provenance, and completeness without genuinely improving capability. We term this phenomenon as harness tampering, which extends the concept from reward and measurement tampering to the full self-improvement lifecycle. To systematically study this problem, we propose a two-axis taxonomy that categorizes each misaligned edit by the harness functional role in which it occurs and the obligation it violates. Then we build an annotated corpus by seeding tampered-benign edit pairs into the real trajectories of self-improving agents. We adapt and benchmark diverse audit methods on tampering classification and localization tasks. Finally we systematically audit real trajectories of self-improving agents. The results demonstrate that harness tampering consistently occurs in real runs from different agents, often persists in the lineage of the best agent, and forms distinct system-specific profiles across the taxonomy.

Chinese Translation

自我改进代理通过迭代修改自身的工具，以推动其性能的边界。然而，这种修改可能产生虚假的性能提升，或妨碍完整性约束，如授权、来源和完整性，而并未真正提高能力。我们将这一现象称为工具篡改，它将奖励和测量篡改的概念扩展到完整的自我改进生命周期。为了系统地研究这个问题，我们提出了一个双轴分类法，将每个不当编辑按其发生的工具功能角色和违反的义务进行分类。然后，我们通过将篡改-良性编辑对嵌入自我改进代理的真实轨迹中，构建了一个注释语料库。我们对篡改分类和定位任务适应并基准测试了多种审计方法。最后，我们系统地审计了自我改进代理的真实轨迹。结果表明，工具篡改在不同代理的真实运行中始终存在，通常在最佳代理的谱系中持续存在，并在分类法中形成独特的系统特定特征。

View on arXiv Download PDF AI Translation

cs.CL / 15 / 2609.00077

Beneath the Diff: Diagnosing and Mitigating Algorithmic Mode Collapse in Code-Level Autonomous Research Loops

深入探讨：诊断与缓解代码级自主研究循环中的算法模式崩溃

He, Bowei, Zhang, Weixu, Jin, Yili, Liu, Xue

Abstract

Code-level autonomous research loops (ARLs) have recently emerged as a concrete object of study in automated machine learning research. In such loops, an LLM agent proposes modifications to an experimental training pipeline, executes the modified pipeline, and retains edits that improve a verifiable in-loop metric. Although executable metrics may appear to provide a reliable signal of progress, it remains unclear whether repeated metric-driven code editing leads to genuine improvements that generalize beyond the loop. We provide a systematic diagnosis of this question. Across various experiment settings, we identify a robust failure mode that we call \textbf{algorithmic mode collapse}. In this regime, surface-level edit diversity remains stable, but semantic and mechanism-level diversity collapse: the agent continues to edit different lines of code while repeatedly proposing the same kinds of algorithmic changes. This collapse is accompanied by a widening gap between in-loop metric gains and gains measured on independent held-out evaluations. We then propose Diversity-Aware Proposal Sampling (\textsc{DAPS}), a lightweight mitigation that combines category-coverage reweighting, persistent edit memory, and a validation gate. Under a three-tier protocol separating the in-loop metric, the audit metric read by the gate, and a blind metric no loop component ever accesses, \textsc{DAPS} reduces semantic-cluster decay of edits by $69.1\%$ and improves relative faithfulness by $83.7\%$ blind and $81.6\%$ audited, while preserving in-loop optimization speed. We provide the code in Github \href{https://github.com/BokwaiHo/arl-mode-collapse}{repository}.

Chinese Translation

代码级自主研究循环（ARLs）最近作为自动化机器学习研究中的一个具体研究对象而出现。在这些循环中，LLM（大语言模型）代理提出对实验训练管道的修改，执行修改后的管道，并保留那些改善可验证循环内指标的编辑。尽管可执行指标似乎提供了一个可靠的进展信号，但反复进行指标驱动的代码编辑是否能带来超出循环的真正改进仍然不清楚。我们对此问题进行了系统的诊断。在各种实验设置中，我们识别出一种我们称之为 extbf{算法模式崩溃}的稳健失败模式。在这种状态下，表面级别的编辑多样性保持稳定，但语义和机制层面的多样性崩溃：代理继续编辑不同的代码行，同时反复提出相同类型的算法变更。这种崩溃伴随着循环内指标增益与在独立保留评估中测得的增益之间的差距扩大。随后，我们提出了多样性意识提案采样（Diversity-Aware Proposal Sampling， extsc{DAPS}），这是一种轻量级的缓解方法，结合了类别覆盖重加权、持久编辑记忆和验证门。在一个将循环内指标、门读取的审计指标和任何循环组件都无法访问的盲指标分开的三层协议下， extsc{DAPS}将编辑的语义簇衰退降低了69.1%，并在保持循环内优化速度的同时，盲审计的相对忠实度提高了83.7%和81.6%。我们在Github exthref{https://github.com/BokwaiHo/arl-mode-collapse}{代码库}中提供了相关代码。

View on arXiv Download PDF AI Translation

cs.CL / 16 / 2609.00082

KItCAT: Knowledge Injection via Input Corruption for Auto-regressive Training

KItCAT：通过输入干扰进行知识注入以实现自回归训练

Pulivarthi, Meghanadh, Bhushan, Kushagra, Kumar, Vineet, Pandey, Gaurav, Sen, Jaydeep, Raghu, Dinesh, Joshi, Sachindra, Nandwani, Yatin

Abstract

LLMs acquire vast amounts of knowledge during pre-training, but often lack the specialized knowledge needed to answer questions from niche sources such as manuals or technical documents unseen during pre-training. Continued pre-training (CPT) is widely used to inject such knowledge into model parameters. However, niche documents seldom repeat facts, making it difficult for CPT to robustly acquire such knowledge. Recent works address this by generating multiple paraphrases of the new knowledge, but paraphrasing is computationally expensive and typically requires powerful LLMs. In this work, we introduce KItCAT: Knowledge Injection via Corrupted Auto-regressive Training, a lightweight training strategy that reduces the need for paraphrasing in decoder-only LLMs. KItCAT augments standard next-token prediction by stochastically corrupting the input sequence. During training, a random subset of input tokens is replaced with other vocabulary tokens while the original next-token labels are kept unchanged. This simple intervention generates diverse training inputs from each sample, enabling large-scale data augmentation at negligible cost. We show that KItCAT consistently improves over CPT across multiple datasets and model families. Code is available at https://github.com/meghanadhpulivarthi/KItCAT.

Chinese Translation

大型语言模型（LLMs）在预训练过程中获取了大量知识，但通常缺乏回答来自小众来源（如手册或未在预训练中见过的技术文档）问题所需的专业知识。持续预训练（CPT）被广泛用于将此类知识注入模型参数。然而，小众文档很少重复事实，使得CPT难以稳健地获取此类知识。近期的研究通过生成新知识的多种释义来解决这一问题，但释义生成计算开销大，通常需要强大的LLMs。在本研究中，我们提出了KItCAT：通过腐败自回归训练进行知识注入，这是一种轻量级训练策略，减少了在仅解码器的LLMs中对释义生成的需求。KItCAT通过随机干扰输入序列来增强标准的下一个标记预测。在训练过程中，输入标记的随机子集被其他词汇标记替换，同时保持原始下一个标记标签不变。这一简单的干预从每个样本生成多样化的训练输入，使得在微不足道的成本下实现大规模数据增强。我们展示了KItCAT在多个数据集和模型系列上始终优于CPT。代码可在 https://github.com/meghanadhpulivarthi/KItCAT 获取。

View on arXiv Download PDF AI Translation

cs.CL / 17 / 2609.00086

Retrieval, Scoring, and Decoding Shape Performance and Stability in LLM-based Conversational Recommendation

检索、评分和解码对基于大型语言模型的对话推荐系统的性能和稳定性的影响

Kapetanovic, Ante, Duricic, Tomislav, Mercep, Andro, Lacic, Emanuel

Abstract

Large language models (LLMs) are increasingly used as rerankers in conversational recommender systems, yet measured gains depend strongly on the retrieval and inference protocol. On the ReDial conversational movie recommendation benchmark, we compare proprietary, open-weight, and fine-tuned LLM rerankers with collaborative-filtering and sequential baselines in a shared retrieve-then-rerank pipeline. We vary candidate-pool size, first-stage retriever, and decoding temperature. With a shared semantic top-250 candidate pool and strict candidate-aware scoring, the best proprietary reranker reaches NDCG@10 of 0.1497, compared with 0.0939 for the strongest non-LLM baseline. The same reranker reaches 0.2925 in zero-shot generation, showing that unconstrained scoring can yield a much larger apparent advantage than matched-pool evaluation. No evaluated open-weight LLM outperforms the tuned shallow autoencoder baseline under this protocol. For the strongest proprietary and open-weight rerankers, switching from semantic to collaborative-filtering candidates raises NDCG@10 by more than 50%, showing that measured reranker performance is highly sensitive to candidate generation. For the best proprietary reranker, raising temperature from 0 to 1.0 increases top-10 Jaccard distance from 0.0900 to 0.1240 while mean NDCG@10 changes negligibly, whereas weaker LLMs show larger degradation. These ReDial results support treating candidate generation, candidate-pool size, scoring policy, and decoding configuration as required reporting fields rather than implementation details.

Chinese Translation

大型语言模型（LLMs）越来越多地被用作对话推荐系统中的重排序器，但测得的增益在很大程度上依赖于检索和推理协议。在ReDial对话电影推荐基准上，我们比较了专有、开放权重和微调的LLM重排序器与协同过滤和序列基线在共享的检索-重排序管道中的表现。我们改变了候选池大小、第一阶段检索器和解码温度。在共享的语义前250候选池和严格的候选感知评分下，最佳的专有重排序器在NDCG@10上达到了0.1497，而最强的非LLM基线为0.0939。同一重排序器在零样本生成中达到了0.2925，表明无约束评分可以产生比匹配池评估更大的表观优势。在该协议下，没有评估的开放权重LLM超越调优的浅层自编码器基线。对于最强的专有和开放权重重排序器，从语义候选切换到协同过滤候选使NDCG@10提高了超过50%，这表明测得的重排序器性能对候选生成高度敏感。对于最佳的专有重排序器，将温度从0提高到1.0使前10的Jaccard距离从0.0900增加到0.1240，而平均NDCG@10几乎没有变化，而较弱的LLM则显示出更大的退化。这些ReDial结果支持将候选生成、候选池大小、评分策略和解码配置视为必要的报告字段，而非实现细节。

View on arXiv Download PDF AI Translation

cs.CL / 18 / 2609.00155

Lingua Franca or Probing Artifact? Rethinking Latent Language in Multilingual LLMs

通用语还是探测工具？重新思考多语言大型语言模型中的潜在语言

Bayazit, Deniz, AlKhamissi, Badr, Bosselut, Antoine

Abstract

Latent language identification is often used to argue that multilingual language models route computation through language-specific states, such as English pivots. However, existing probes infer latent language from different signals, such as the geometry of hidden states or what can be decoded from intermediate representations. Since such claims shape conclusions about how models share and route information across languages, we ask whether these probes measure the same phenomenon or expose distinct aspects of multilingual computation. We study this question across model families, training regimes, domains, tasks, checkpoints, and up to 27 languages. We find that identification probes systematically disagree: the GMM-based representation probe, which draws evidence from hidden state geometry, shows earlier cross-lingual mixing, whereas decoding-based probes, which rely on output-space decodability, retain sharper language-specific and more English-biased signals. These differences track model multilinguality and training progression, but are comparatively stable across domains. Our results suggest a more cautious interpretation of latent language identification, where current probes expose different aspects of multilingual processing, rather than directly revealing a single internal lingua franca.

Chinese Translation

潜在语言识别常被用来论证多语言模型通过语言特定状态（如英语枢纽）进行计算。然而，现有的探测工具通过不同的信号推断潜在语言，例如隐藏状态的几何形状或从中间表示中可以解码的信息。由于这些主张影响了关于模型如何在语言间共享和路由信息的结论，我们探讨这些探测工具是否测量相同的现象或揭示多语言计算的不同方面。我们在模型家族、训练模式、领域、任务、检查点以及多达27种语言中研究这个问题。我们发现识别探测工具系统性地存在分歧：基于高斯混合模型（GMM）的表示探测工具，从隐藏状态几何形状中获取证据，显示出更早的跨语言混合，而基于解码的探测工具，依赖于输出空间的可解码性，保留了更清晰的语言特定和更偏向英语的信号。这些差异与模型的多语言性和训练进展相关，但在不同领域中相对稳定。我们的结果提示对潜在语言识别的更谨慎解读，目前的探测工具揭示了多语言处理的不同方面，而不是直接揭示单一的内部通用语。

View on arXiv Download PDF AI Translation

cs.CL / 19 / 2609.00177

Do General NLP Embeddings Capture Ontological Reasoning?

通用自然语言处理嵌入是否能够捕捉本体推理？

Giglou, Hamed Babaei, D'Souza, Jennifer, Auer, Sören

Abstract

General-purpose NLP embedding models perform well on linguistic tasks, but their ability to capture symbolic ontological structure remains unclear. We introduce AVA, a systematic framework for evaluating whether embeddings distinguish logic-sensitive relational semantics in ontologies and knowledge graphs. AVA comprises 171,007 contrastive triplets derived from 163 heterogeneous ontologies using hierarchy inversion, relation substitution, and disjointness injection. Each triplet contains an ontology statement, a semantically equivalent paraphrase, and a logic-sensitive hard negative with contradictory relational meaning. We evaluate more than 25 state-of-the-art embedding models and find substantial limitations: the best model achieves only 0.739 triplet accuracy, while hard negative accuracy falls to 0.135. Fine-tuning improves discrimination by a large margin but transfers poorly to downstream Semantic Web tasks, including taxonomy discovery and ontology alignment. Further analysis suggests that improvements stem partly from perturbation-specific pattern recognition rather than robust ontological understanding. These findings reveal a persistent gap between linguistic representation learning and ontology-level discrimination, challenging the assumption that strong NLP benchmark performance translates to Semantic Web competence.

Chinese Translation

通用自然语言处理嵌入模型在语言任务中表现良好，但它们捕捉符号本体结构的能力仍不明确。我们引入了AVA，这是一个系统性框架，用于评估嵌入是否能够区分本体和知识图谱中的逻辑敏感关系语义。AVA包含171,007个对比三元组，这些三元组源自163个异构本体，采用了层次反转、关系替换和不相交注入等方法。每个三元组包含一个本体陈述、一个语义等价的释义，以及一个具有矛盾关系意义的逻辑敏感难负样本。我们评估了25个以上的最先进嵌入模型，并发现其存在显著局限性：最佳模型的三元组准确率仅为0.739，而难负样本的准确率降至0.135。微调显著提高了区分能力，但在下游语义网任务（包括分类法发现和本体对齐）中的迁移效果较差。进一步分析表明，改进部分源于特定扰动模式的识别，而非稳健的本体理解。这些发现揭示了语言表示学习与本体级区分之间的持续差距，挑战了强大的自然语言处理基准性能能够转化为语义网能力的假设。

View on arXiv Download PDF AI Translation

cs.CL / 20 / 2609.00184

Synthetic Worlds for Temporal Evaluation and Knowledge Updating in LLMs

用于大语言模型的时间评估和知识更新的合成世界

Zheng, Jonathan, Shao, Zirui, Ritter, Alan, Xu, Wei

Abstract

Large language models (LLMs) rely on static pretraining corpora, causing their knowledge to become outdated over time. Existing approaches for evaluating knowledge edits either suffer from rapid contamination or rely on counterfactual edits that conflict with rigid existing knowledge. In this work, we propose a synthetic, simulation-driven framework for studying knowledge insertion in LLMs. We introduce {\sc ParallelEvents}, a benchmark of fictional yet realistic future worlds that generates coherent event trajectories for controlled evaluation, avoiding contamination while preserving consistency. Building on this dataset, we develop {\sc Synapse}, a training framework that uses model-generated data to update model parameters via mid-training and instruction tuning. This synthetic pipeline enables scalable knowledge integration without costly human-curated data. Empirically, {\sc Synapse} outperforms existing methods by 14.23\%, demonstrating that simulation-based synthetic training leads to robust and coherent knowledge insertions.

Chinese Translation

大型语言模型（LLMs）依赖于静态的预训练语料库，导致其知识随着时间的推移而变得过时。现有的知识编辑评估方法要么受到快速污染的影响，要么依赖于与现有知识相悖的反事实编辑。在本研究中，我们提出了一种合成的、基于模拟的框架，用于研究LLMs中的知识插入。我们引入了{ extsc{ParallelEvents}}，这是一个虚构但现实的未来世界基准，能够生成连贯的事件轨迹以进行控制评估，避免污染同时保持一致性。在此数据集的基础上，我们开发了{ extsc{Synapse}}，这是一个训练框架，通过中期训练和指令调优，利用模型生成的数据来更新模型参数。该合成管道使得在没有昂贵人工整理数据的情况下实现可扩展的知识集成成为可能。实证结果表明，{ extsc{Synapse}}比现有方法提高了14.23 ext{%}的性能，证明了基于模拟的合成训练能够实现稳健且连贯的知识插入。

View on arXiv Download PDF AI Translation

cs.CL / 21 / 2609.00191

Assessing Suicide Risk in Arabic Crisis Helpline Calls: A Comparison of Arabic and English Large Language Models

评估阿拉伯危机热线电话中的自杀风险：阿拉伯语与英语大型语言模型的比较

Ma, Linhai, Hachem, Rita El, Hajj, Mahatab El, Ghandour, Lilian, Fodeh, Samah

Abstract

Crisis helplines assess suicide risk through structured interviews, a process that is slow and dependent on operator training and workload. Natural language processing could support risk assessment and call prioritization, but almost no work addresses Arabic-language helpline calls or operates within the privacy constraints of real helpline data. We analysed de-identified transcripts from Lebanon's National Lifeline for Emotional Support and Suicide Prevention. Audio never left the helpline: calls were transcribed on site with a speech recognition model for Levantine Arabic, and an Arabic named-entity recognition model removed identifying information locally. Only the de-identified transcripts were shared with the research team. Operators recorded the five suicidal ideation items of the Columbia Suicide Severity Rating Scale, which we combined into two binary outcomes: at-risk and high-risk. We also machine-translated the transcripts into English, giving a paired Arabic/English comparison. On each corpus, we fine-tuned five instruction-tuned large language models alongside six transformer encoder baselines (four Arabic, two English) and evaluated all models on a held-out test set. We included 383 calls: 373 for the at-risk task (52.3% positive) and 297 for the high-risk task (30.0% positive). The best Arabic model reached a macro-F1 of 81.19 and a ROC-AUC of 90.61 on high-risk; the best English model reached 85.00 and 92.59, identifying 88.9% of high-risk calls. In both languages, high-risk calls separated more cleanly than at-risk calls, and translation to English did not reduce the best observed performance. Suicide risk can be classified from de-identified Arabic transcripts without sending audio outside the helpline. The high-risk results support further testing as an operator-facing tool; lower-severity ideation proved the harder case.

Chinese Translation

危机热线通过结构化访谈评估自杀风险，这一过程既缓慢又依赖于操作员的培训和工作负荷。自然语言处理可以支持风险评估和电话优先级排序，但几乎没有研究涉及阿拉伯语热线电话或在真实热线数据的隐私限制下进行操作。我们分析了来自黎巴嫩国家情感支持与自杀预防热线的去标识化转录文本。音频从未离开热线：电话在现场通过针对黎凡特阿拉伯语的语音识别模型进行转录，并且一个阿拉伯语命名实体识别模型在本地移除了识别信息。只有去标识化的转录文本与研究团队共享。操作员记录了哥伦比亚自杀严重性评分量表的五个自杀意念项目，我们将其合并为两个二元结果：有风险和高风险。我们还将转录文本机器翻译为英语，提供了配对的阿拉伯语/英语比较。在每个语料库中，我们微调了五个指令调优的大型语言模型，以及六个变换器编码器基线模型（四个阿拉伯语，两个英语），并在保留的测试集上评估了所有模型。我们包括了383个电话：373个用于有风险任务（52.3%为正例），297个用于高风险任务（30.0%为正例）。最佳阿拉伯语模型在高风险任务中达到了81.19的宏F1值和90.61的ROC-AUC值；最佳英语模型达到了85.00和92.59，识别出88.9%的高风险电话。在两种语言中，高风险电话的区分度优于有风险电话，而翻译成英语并未降低观察到的最佳性能。自杀风险可以从去标识化的阿拉伯语转录文本中进行分类，而无需将音频发送出热线。高风险结果支持进一步测试作为面向操作员的工具；较低严重程度的意念被证明是更困难的案例。

View on arXiv Download PDF AI Translation

cs.CL / 22 / 2609.00213

Uncovering and Mitigating Aggregation-Induced Reward Hacking in Multi-Reward Reinforcement Learning

揭示和缓解多奖励强化学习中的聚合诱导奖励黑客行为

Yuan, Yu, Fan, Yaoyou, Zhao, Lili, Zheng, Guangting, Zhang, Kai, Pan, Lu, Zeng, Ke, Liu, Qi

Abstract

Reinforcement learning fine-tuning of large language models increasingly adopts multiple reward dimensions, including verifiable rules, task-specific evaluators, and learned reward models, to provide richer supervision across diverse capabilities. These dimensions are commonly scalarized with fixed aggregation weights. We identify a failure mode in which aggregation itself induces reward hacking: static projection aliases qualitatively different reward profiles into a single scalar, steering optimization toward whichever dimensions are easiest, densest, or systematically favored by the reward signal. Over training, this traps the policy in suboptimal profiles and prevents convergence to better-balanced ones that would yield higher task performance. To address this, we propose Adaptive Multi-Reward Projection (AMRP), a lightweight online method that reallocates aggregation weights using three signals, relative shortfall, reward volatility, and recent progress, increasing pressure on lagging, unstable, or stagnant dimensions while relieving saturated ones. Across structured reasoning, citation-grounded generation, and open-ended alignment under GRPO, AMRP consistently improves reward-profile balance and downstream performance over fixed and dynamic weighting baselines; it also remains effective with GDPO and PPO, supporting compatibility across RL algorithms. Our code is available at https://github.com/yyhappier/AMRP.git.

Chinese Translation

强化学习对大型语言模型的微调越来越多地采用多重奖励维度，包括可验证的规则、特定任务的评估者和学习的奖励模型，以便在多样化的能力上提供更丰富的监督。这些维度通常使用固定的聚合权重进行标量化。我们识别出一种失效模式，其中聚合本身诱导了奖励黑客行为：静态投影将质 qualitatively 不同的奖励特征别名为单一标量，从而引导优化朝向最容易、最密集或系统性偏好的奖励信号维度。在训练过程中，这使得策略陷入次优特征并阻止其收敛到能够提供更高任务性能的更平衡特征。为了解决这个问题，我们提出了自适应多奖励投影（Adaptive Multi-Reward Projection, AMRP），这是一种轻量级的在线方法，通过相对短缺、奖励波动性和近期进展三个信号重新分配聚合权重，增加对滞后、不稳定或停滞维度的压力，同时减轻饱和维度的压力。在结构化推理、基于引用的生成和在 GRPO 下的开放式对齐中，AMRP 一直在固定和动态加权基线之上持续改善奖励特征平衡和下游性能；它在 GDPO 和 PPO 中也保持有效，支持跨强化学习算法的兼容性。我们的代码可在 https://github.com/yyhappier/AMRP.git 获取。

View on arXiv Download PDF AI Translation

cs.CL / 23 / 2609.00222

LLM-as-a-Demographic: Whom Sociodemographic Prompting Helps, and Whom It Hurts

将大型语言模型视为人口统计学：社会人口统计提示对谁有帮助，对谁有伤害

Occhipinti, Daniela, Piergentili, Andrea, Guerini, Marco

Abstract

Large language models (LLMs) are increasingly used as judges for subjective tasks, where annotators disagree and the relevant question is not only how accurate a judge is, but whose judgments it reproduces. Sociodemographic prompting conditions the judge on an annotator's demographic profile to align its judgments with the corresponding group's. We test whether this alignment emerges distributionally, comparing the predicted label distributions of 23 open-weight LLMs on three subjective tasks against those of real annotator groups, under three conditions: no demographic information, single-attribute profiles, and intersectional profiles over gender, age, race, and education. Three findings emerge. First, a judge prompted with no demographics is not perspective-neutral: models best reproduce the judgments of White, college-educated annotators. Second, demographic conditioning is asymmetric: it moves the judge toward majority groups and away from minority groups, most strongly on offensiveness, where intersectional profiles amplify the harm. Third, by comparing base and instruct models we identify instruction-tuning as a possible source of the asymmetry. Demographic conditioning should therefore be used with caution to estimate group judgments: conditioning moves predictions away from the reference distributions of the minority groups the method is often invoked to serve.

Chinese Translation

大型语言模型（LLMs）越来越多地被用作主观任务的评判者，在这些任务中，标注者之间存在分歧，相关的问题不仅是评判者的准确性，还包括他们再现了谁的判断。社会人口统计提示根据标注者的人口统计特征来调整评判者，使其判断与相应群体一致。我们测试这种一致性是否在分布上出现，通过比较23个开放权重LLM在三项主观任务上的预测标签分布与真实标注者群体的标签分布，在三种条件下进行比较：没有人口统计信息、单一属性档案以及性别、年龄、种族和教育的交叉属性档案。得出了三个发现。首先，未使用人口统计信息提示的评判者并非视角中立：模型最能再现白人大学受教育标注者的判断。其次，人口统计条件化是不对称的：它使评判者向多数群体倾斜，远离少数群体，尤其在冒犯性判断上，交叉属性档案加剧了这种伤害。第三，通过比较基础模型和指令模型，我们确定指令调优可能是这种不对称性的来源。因此，使用人口统计条件化来估计群体判断时应谨慎：条件化使预测偏离了少数群体的参考分布，而这种方法通常被用来服务于这些群体。

View on arXiv Download PDF AI Translation

cs.CL / 24 / 2609.00228

Bridging Lexical Divergence: LLM-Assisted, Cost-Efficient, Zero-shot Scientific Entity Linking

弥合词汇差异：LLM辅助的成本高效零样本科学实体链接

Khondokar, Md Rasel, Qiao, Qiao, Samia, Farjana Sultana, Le, Nhat, Li, Yuepei, Li, Qi

Abstract

Scientific domain entity linking (EL) differs from general domain EL because mentions and entity names often lack lexical overlap. Another challenge is that specialized terminology is used in the scientific domain, which is rarely encountered in models pretrained on general domains. Therefore, models trained on general domains transfer poorly to scientific domains. To address this, in-domain fine-tuning is the natural remedy. However, many scientific domains lack expert-annotated data, motivating the need for a zero-human-annotation approach. Existing zero-shot methods heavily rely on LLMs to generate aliases across entire mention corpora, which incurs substantial computational cost, and those methods provide no mechanism to filter out noise from LLMs. To address these challenges, we propose Sci-ZSEL, a framework that selectively generates entity aliases with an LLM to control computational cost, and applies an ontology-aware filter to remove aliases that semantically drift toward ontology neighbors. Then, filtered aliases are used to construct pseudo-labeled mention-entity pairs for fine-tuning. To enable evaluation of EL under low lexical overlap, we also release a new animal science EL benchmark linked to three livestock trait ontologies, where mentions and entities exhibit substantially lower lexical overlap than in existing benchmarks. Across five benchmarks, Sci-ZSEL outperforms the non-fine-tuned baseline, is most useful on nonoverlapping mentions, and combining it with curated synonyms gives the best performance in most settings.

Chinese Translation

科学领域的实体链接（EL）与一般领域的EL有所不同，因为提及和实体名称往往缺乏词汇重叠。另一个挑战是科学领域使用的专业术语在经过一般领域预训练的模型中很少遇到。因此，基于一般领域训练的模型在科学领域的迁移效果较差。为了解决这个问题，领域内微调是自然的解决方案。然而，许多科学领域缺乏专家注释的数据，这促使我们需要一种零人工注释的方法。现有的零样本方法严重依赖LLM（大型语言模型）在整个提及语料库中生成别名，这会产生可观的计算成本，并且这些方法没有提供过滤LLM噪声的机制。为了解决这些挑战，我们提出了Sci-ZSEL，一个选择性生成实体别名的框架，利用LLM控制计算成本，并应用一个本体感知过滤器来去除语义上漂移到本体邻居的别名。然后，经过过滤的别名用于构建伪标注的提及-实体对以进行微调。为了评估在低词汇重叠下的EL，我们还发布了一个新的动物科学EL基准，链接到三个牲畜性状本体，其中提及和实体的词汇重叠显著低于现有基准。在五个基准测试中，Sci-ZSEL的表现优于未微调的基线，在非重叠提及上最为有效，并且将其与精心挑选的同义词结合在大多数设置中提供了最佳性能。

View on arXiv Download PDF AI Translation

cs.CL / 25 / 2609.00241

LOOMSUM:Weaving Quantitative and Narrative Evidence for Faithful Long Text-Table Summarization

LOOMSUM：编织定量与叙述证据以实现忠实的长文本-表格摘要

Zhou, Meng, You, Wenhao, Yuan, Wei

Abstract

Long documents often distribute important information across extensive narrative passages and multiple tables, making faithful summarization particularly challenging. Existing methods may generate individually supported quantitative facts and analytical statements yet associate them incorrectly, producing quantitatively plausible yet analytically unfaithful summaries. In this work, we propose LOOMSUM, a training-free framework that extracts source-grounded atomic evidence, explicitly links table-derived facts with supporting narrative analyses, and plans the discourse structure before generation. We also introduce Table-Grounded Faithfulness (TGF), a claim-level metric that separately evaluates Numeric Grounding, Analysis Support, and Relation Consistency. Experiments on the text--table summarization benchmarks FINDSum and USTT show that LOOMSUM improves analytical faithfulness while maintaining strong summarization quality. Human evaluation finds positive component-level associations with the corresponding human judgments. Our Relation Consistency metric further shows stronger agreement with human relation judgments than generic factuality metrics, indicating that explicit cross-modal linking helps reduce errors in which supported quantities are paired with incorrect narrative interpretations. Together, these findings show that faithful long text--table summarization requires not only grounding individual facts, but also preserving the relations between them.

Chinese Translation

长文档通常在广泛的叙述段落和多个表格中分布重要信息，使得忠实摘要特别具有挑战性。现有方法可能生成个别支持的定量事实和分析陈述，但将它们错误关联，从而产生在定量上看似合理但在分析上不忠实的摘要。在本研究中，我们提出了LOOMSUM，这是一种无训练的框架，提取源基础的原子证据，明确将表格派生的事实与支持的叙述分析联系起来，并在生成之前规划话语结构。我们还引入了表格基础忠实性（Table-Grounded Faithfulness，TGF），这是一种在声明层面上分别评估数值基础、分析支持和关系一致性的指标。在文本-表格摘要基准FINDSum和USTT上的实验表明，LOOMSUM在保持强摘要质量的同时提高了分析忠实性。人类评估发现与相应的人类判断之间存在积极的组件级关联。我们的关系一致性指标进一步显示与人类关系判断的更强一致性，相较于通用的事实性指标，表明明确的跨模态链接有助于减少支持的数量与错误的叙述解释之间的配对错误。综合这些发现表明，忠实的长文本-表格摘要不仅需要对个别事实进行基础化，还需要保持它们之间的关系。

View on arXiv Download PDF AI Translation

cs.CL / 26 / 2609.00256

NSIDDx: A Design Framework for Neuro-Symbolic, Practitioner-First Differential Diagnosis in Low-Resource Settings

NSIDDx：低资源环境中以实践者为中心的神经符号差异诊断设计框架

Singh, Aarav

Abstract

LLM-based diagnostic systems achieve high semantic accuracy on benchmarks, but open-ended evaluation on clinically uncommon presentations reveals a systematic gap between headline accuracy and verifiable clinical reliability. We evaluate an LLM+rare-disease-RAG pipeline across two cohorts and show that the paradigm produces confident outputs that are frequently unverifiable and systematically resistant to clinician interrogation. We present NSIDDx (Neuro-Symbolic Integrated Differential Diagnosis System), a design framework arguing that DDx systems in low-resource settings must treat the clinician as an active reasoning agent. We instantiate this through a neuro-symbolic pipeline with ternary symptom encoding, contradiction detection, audit strings, and practitioner override - running offline on consumer hardware. We distill five design principles for clinician-in-the-loop clinical NLP and invite the prospective studies needed to validate the claim at scale.

Chinese Translation

基于大语言模型（LLM）的诊断系统在基准测试中实现了高语义准确性，但在对临床不常见表现的开放式评估中，揭示了头条准确性与可验证临床可靠性之间的系统性差距。我们在两个队列中评估了LLM+罕见疾病检索增强（RAG）管道，结果表明该范式产生的自信输出常常难以验证，并且系统性地抵抗临床医生的质询。我们提出了NSIDDx（神经符号集成差异诊断系统），这是一个设计框架，主张低资源环境中的差异诊断系统必须将临床医生视为积极的推理主体。我们通过一个具有三元症状编码、矛盾检测、审计字符串和实践者覆盖功能的神经符号管道来实现这一目标，该管道在消费级硬件上离线运行。我们提炼出五个临床医生参与的临床自然语言处理设计原则，并邀请进行前瞻性研究以在更大规模上验证这一主张。

View on arXiv Download PDF AI Translation

cs.CL / 27 / 2609.00293

Slow to See, Slow to Suppress: Understanding the Effects of Modality in Context-Memory Conflicts

观察缓慢，抑制缓慢：理解模态在上下文-记忆冲突中的影响

Paraselli, Athulith, Hua, Etha Tianze, Pavlick, Ellie

Abstract

We investigate how vision-language models (VLMs) handle context-memory conflicts; that is, situations in which the model is given information in context that differs from what was stored parametrically during training. We document asymmetric biases: models tend to prefer in-context information about entities which appear in text, but prefer parametric information about entities which appear in images. We relate this asymmetry to the late representational alignment across modalities, showing that the longer processing time associated with resolving visual entities prevents the suppression of the model's usual factual recall mechanism, thus resulting in more parametric answers. Chain-of-thought reasoning does not appear to resolve the gap, but increasing the amount of visual information in the context does show an effect. These results illustrate the complexity of ensuring consistent behavior as models become increasingly multimodal and retrieval-augmented.

Chinese Translation

我们研究了视觉-语言模型（VLMs）如何处理上下文-记忆冲突；即模型在上下文中接收到的信息与训练期间存储的参数信息不同的情况。我们记录了不对称偏差：模型倾向于偏好文本中出现的实体的上下文信息，但更倾向于偏好图像中出现的实体的参数信息。我们将这种不对称性与跨模态的晚期表征对齐联系起来，表明与解决视觉实体相关的较长处理时间阻止了模型通常的事实回忆机制的抑制，从而导致更多的参数性答案。链式思维推理似乎无法弥补这一差距，但增加上下文中的视觉信息量确实显示出效果。这些结果说明了随着模型日益多模态和增强检索，确保一致行为的复杂性。

View on arXiv Download PDF AI Translation

cs.CL / 28 / 2609.00296

Toward Workflow-Aware Benchmarking for Healthcare NLP Agents

面向工作流程感知的医疗保健自然语言处理代理基准测试

Yao, Junyi, Li, Baichuan, Zheng, Zihao, Long, Jiayu

Abstract

Large language model (LLM) agents are increasingly proposed for healthcare tasks such as clinical documentation, evidence retrieval, patient messaging, and care coordination. Yet many evaluations remain limited to static medical question answering or one-shot generation, under-representing longitudinal state, interruptions, and human handoffs. We introduce an episode-level evaluation protocol for healthcare NLP agents. The protocol separates evidence across model, agent, and simulated-workflow behavior; specifies a five-field episode schema; and defines annotation and scoring for state continuity, evidence traceability, and escalation decisions. It is instantiated as four task templates: documentation update, evidence retrieval, patient messaging, and triage handoff. The protocol does not claim to measure clinical outcomes or deployment value. Instead, it supplies a reproducible intermediate evaluation layer between static benchmarks and prospective workflow studies, with an explicit cost-sensitive treatment of missed versus unnecessary escalation.

Chinese Translation

大型语言模型（LLM）代理越来越多地被提议用于医疗保健任务，例如临床文档、证据检索、患者消息传递和护理协调。然而，许多评估仍然局限于静态医学问答或单次生成，未能充分代表纵向状态、中断和人际交接。我们提出了一种针对医疗保健自然语言处理代理的情节级评估协议。该协议将证据分离为模型、代理和模拟工作流程行为；指定了五个字段的情节架构；并定义了状态连续性、证据可追溯性和升级决策的注释和评分。它被实例化为四个任务模板：文档更新、证据检索、患者消息传递和分诊交接。该协议并不声称测量临床结果或部署价值。相反，它提供了一种可重复的中间评估层，介于静态基准和前瞻性工作流程研究之间，并明确考虑了漏掉与不必要升级的成本敏感性。

View on arXiv Download PDF AI Translation

cs.CL / 29 / 2609.00310

Emotional Labor Strategy Preferences in LLM Personas

大型语言模型角色中的情感劳动策略偏好

Saim, Mohammad, Jiang, Tianyu

Abstract

Emotional labor is the effortful management of emotional displays to meet social or professional expectations. Personality traits have been correlated with emotional labor strategies, yet research on this link relies almost exclusively on self-report scales administered only in occupational settings. We investigate whether large language models injected with psychometrically grounded personas reproduce these personality-driven selection patterns across everyday social scenarios. We construct the first emotional labor strategy dataset of 500 socially situated events, each offering three behavioral choices corresponding to surface acting, deep acting, and genuine expression. We source 50 fictional characters from a large-scale personality repository and profile each through two parallel tracks: observer-rated bipolar adjective composites and in-character self-report items. Five LLMs evaluate all scenarios under both persona conditions. We find that models align more towards deep acting, and that Conscientiousness and Emotional Stability consistently predict this preference. Entropy analysis confirms that persona reliably influences the output and varies across models and emotions.

Chinese Translation

情感劳动是为了满足社会或职业期望而进行的情感表现管理。个性特征与情感劳动策略之间存在相关性，但对这一联系的研究几乎完全依赖于仅在职业环境中实施的自我报告量表。我们研究了注入心理测量基础角色的大型语言模型是否在日常社交场景中重现这些个性驱动的选择模式。我们构建了第一个包含500个社会情境事件的情感劳动策略数据集，每个事件提供三种行为选择，分别对应表面表演、深层表演和真实表达。我们从一个大规模个性库中获取了50个虚构角色，并通过两个平行的途径对每个角色进行特征描述：观察者评定的双极形容词组合和角色内自我报告项目。五个大型语言模型在这两种角色条件下评估所有场景。我们发现模型更倾向于深层表演，并且尽责性（Conscientiousness）和情绪稳定性（Emotional Stability）始终预测这种偏好。熵分析确认角色可靠地影响输出，并在不同模型和情感之间有所变化。

View on arXiv Download PDF AI Translation

cs.CL / 30 / 2609.00325

Latent Mechanisms of Language Control in Multilingual Language Models

多语言模型中的语言控制潜在机制

Mitsuhashi, Ryo, Boughorbel, Sabri, Hawasly, Majd

Abstract

Multilingual large language models can exhibit unintended code-switching -- unnecessarily alternating between languages during generation. We present a comparative study of three methods that identify language-controlling latents in cross-layer transcoders: activation value-based selection (ValSel), activation frequency-based selection (FreqSel), and LLM-generated latent annotation-based selection (AnnSel). To evaluate the efficacy of these methods in identifying language-controlling latents, we introduce two multilingual benchmarks that exhibit code-switching for fine-grained analysis of language steering across seven languages. Through targeted intervention experiments on Gemma-2-2B and Qwen3-4B, we find that all three methods effectively manipulate generation language, with FreqSel achieving the strongest overall performance, while AnnSel offering interpretable latent selection through explicit language annotations. A knock-out analysis suggests the methods select non-overlapping but each-functional latent subsets, indicating redundancy rather than a single canonical language direction. Code and data can be found at https://github.com/rm-3284/Latent-Mechanism-Multilingual.

Chinese Translation

多语言大型语言模型可能表现出意外的代码切换——在生成过程中不必要地在语言之间交替。我们提出了一项比较研究，探讨三种方法在跨层转码器中识别语言控制潜在因素的效果：基于激活值的选择（ValSel）、基于激活频率的选择（FreqSel）和基于LLM生成的潜在注释选择（AnnSel）。为了评估这些方法在识别语言控制潜在因素方面的有效性，我们引入了两个多语言基准，这些基准展示了代码切换，以便对七种语言的语言引导进行细致分析。通过对Gemma-2-2B和Qwen3-4B的针对性干预实验，我们发现这三种方法都能有效操控生成语言，其中FreqSel的整体表现最强，而AnnSel则通过明确的语言注释提供了可解释的潜在选择。击倒分析表明，这些方法选择了不重叠但各自功能不同的潜在子集，表明存在冗余而非单一的规范语言方向。代码和数据可在 https://github.com/rm-3284/Latent-Mechanism-Multilingual 找到。

View on arXiv Download PDF AI Translation

cs.CL / 31 / 2609.00329

The Curse of Multilinguality in Lexical Normalization

词汇规范化中的多语言诅咒

Rahbar, Saman

Abstract

Lexical normalization rewrites the noisy, non-standard words that fill user-generated text (tmrw, u, gr8) into their standard forms. Because labelled data is scarce for most languages, a popular shortcut is to train a single model on many languages at once. We ask a simple question: how many languages should such a model be trained on? Using one fixed-capacity character-level model and twelve languages from a standard benchmark, we vary the number of jointly trained languages from one to twelve and measure per-language accuracy. We find a clear curse of multilinguality: accuracy is highest when a language is trained with only a few others, often just one to four, and then falls steadily and substantially, dropping by about forty percent as the rest are piled on. A control that holds the total amount of training data constant makes the decline arrive sooner and fall further, which points to competition among the languages for one fixed-size model rather than to how much data is available. We also test whether a language's typological distance from the others predicts its ideal number of co-training languages, and find no dependable rule: any apparent relationship rests on a couple of languages and does not hold up. For compact normalization models, less can be more: a few languages beat pooling everything into a single model.

Chinese Translation

词汇规范化将充斥用户生成文本中的嘈杂、不标准词汇（如 tmrw、u、gr8）重写为其标准形式。由于大多数语言的标注数据稀缺，一种流行的捷径是同时在多种语言上训练一个单一模型。我们提出一个简单的问题：这样的模型应该在多少种语言上进行训练？我们使用一个固定容量的字符级模型和来自标准基准的十二种语言，改变联合训练语言的数量，从一种到十二种，并测量每种语言的准确性。我们发现了明显的多语言诅咒：当一种语言与仅有的几种其他语言（通常是一到四种）共同训练时，其准确性最高，随后准确性稳步且显著下降，随着其他语言的加入，准确性下降约四十个百分点。一个控制实验保持训练数据总量不变，使得准确性的下降更早到来且下降幅度更大，这表明语言之间的竞争是针对一个固定大小模型的，而不是可用数据的多少。我们还测试了一种语言与其他语言的类型距离是否能预测其理想的共同训练语言数量，发现没有可靠的规律：任何明显的关系都仅基于少数几种语言，并不成立。对于紧凑的规范化模型而言，少即是多：少数几种语言的表现优于将所有语言汇聚成一个单一模型。

View on arXiv Download PDF AI Translation

cs.CL / 32 / 2609.00330

Topic Matching in the Wild: Benchmark and Lessons from Real-World ASR Transcripts

现实环境中的主题匹配：来自真实世界ASR转录的基准与经验教训

Rahbar, Saman, Zhu, Xiliang, Cardoza, Irvin, Rossouw, David

Abstract

In contact centers, real-time agent-assist tools determine, for each of many predefined topics, whether a live customer utterance is relevant and display a coaching card to the agent when it is. The input is noisy and challenging: ASR(Automatic Speech Recognition) transcripts of spontaneous phone conversations, which can be unclear, repetitive, and mostly lack punctuation. To systematically study this real-world task, we curate a human-annotated topic-utterance judgments dataset sourced from real call-center transcripts. We compare three types of matchers: a regex-based baseline, zero-shot sentence embedding encoders, and Gemini-based LLM matchers. In addition, two types of topic representations are studied in our benchmark:keyphrases and natural language description. Our empirical experiments highlight the superior performance of lightweight LLM matchers over embedding and regex models when equipped with natural language descriptions.

Chinese Translation

在呼叫中心，实时代理辅助工具会针对多个预定义主题判断每个客户的实时发言是否相关，并在相关时向代理展示辅导卡。输入数据嘈杂且具有挑战性：来自自发电话对话的ASR（自动语音识别）转录，通常不清晰、重复且大多缺乏标点。为了系统地研究这一现实任务，我们整理了一个基于人类标注的主题-发言判断数据集，来源于真实的呼叫中心转录。我们比较了三种类型的匹配器：基于正则表达式的基线、零样本句子嵌入编码器和基于Gemini的LLM匹配器。此外，我们的基准研究中还考察了两种类型的主题表示：关键词和自然语言描述。我们的实证实验强调了轻量级LLM匹配器在配备自然语言描述时，相较于嵌入和正则表达式模型的优越性能。

View on arXiv Download PDF AI Translation

cs.CL / 33 / 2609.00335

Two locked tests of phase-structure features for transition prediction

两项锁定测试的相位结构特征用于过渡预测

Chachamovits, Abraham

Abstract

A published theoretical account of phase structure in rotary attention was subjected to two pre-specified empirical tests of whether phase-derived features improve prediction of a commitment or contradiction endpoint over a baseline that does not receive those features. Study 1 froze a contradiction-category pipeline and scored a sealed primary comparison of PC-2 against baseline. On 1,136 eligible cases the paired AUROC difference was +0.00087. The 99% interval included zero, and the difference did not reach the pre-specified threshold of +0.05. Advancement was not passed. Study 2 developed fifteen layer treatments on open blocks b0-b4 only (1,415 transitions, 20x5 grouped folds). A locked conjunctive rule required a positive PC-2 mean-repeat increment, a positive increment on at least four of five seed blocks, and a positive mean of those five differences. No treatment advanced. The official selection is null. The theoretical paper is not withdrawn. The extra ranking lift was not found under the rules locked in advance.

Chinese Translation

一项已发表的关于旋转注意力中相位结构的理论研究，经过两项预先指定的实证测试，以检验相位派生特征是否能改善对承诺或矛盾终点的预测，相较于不包含这些特征的基线。研究1冻结了矛盾类别管道，并对PC-2与基线的密封主要比较进行了评分。在1,136个符合条件的案例中，配对的AUROC差异为+0.00087。99%的置信区间包含零，且差异未达到预先指定的+0.05阈值，因此未通过进展。研究2仅在开放块b0-b4上开发了十五层处理（1,415个过渡，20x5分组折叠）。一个锁定的结合规则要求PC-2的均值重复增量为正，至少在五个种子块中的四个上有正增量，以及这五个差异的均值为正。没有处理获得进展。官方选择为无效。该理论论文未被撤回。在预先锁定的规则下未发现额外的排名提升。

View on arXiv Download PDF AI Translation

cs.CL / 34 / 2609.00344

From Tool Use to Technological Agency: LoopCAT as a Local-First, Open-Source Tool for Translation Technology Education

从工具使用到技术代理：LoopCAT作为一种本地优先的开源翻译技术教育工具

Dogru, Gokhan, Mor, Adrià Martín

Abstract

Translation students need to learn both how to use translation technologies and how to judge the choices those technologies make available. This article presents LoopCAT, an Apache-2.0-licensed, local-first computer-assisted translation environment co-created with OpenAI Codex using GPT-5.5 and GPT-5.6, and proposes a framework connecting workflow competence, evaluative judgement, and technological agency. The account draws on repository history, implementation inspection, and the verification records of an identified development build. LoopCAT combines local project storage, translation memories, terminology, quality assurance, document exchange, and optional connections to local or hosted AI services. Its English, Catalan, and Turkish interface catalogs also make the application itself available as teaching material: students can translate English UI strings into another language, review the existing automatically generated target drafts, import their revisions, and test the interface. We organize these opportunities around four forms of participation: operating a workflow, evaluating outputs, inspecting and configuring mechanisms, and making or defending a bounded intervention. A six-session sequence, a UI-localization assignment, a placeholder example, and an assessment rubric specify how teachers could use the framework. The paper separates implemented capabilities from proposed educational benefits; it reports no new student-learning outcomes. It distinguishes the latest package checks from earlier regression evidence and sets out a protocol for classroom evaluation. LoopCAT provides an inspectable setting for teaching how translation decisions interact with data, interfaces, and software rules. Whether these activities improve judgement, transfer, or participation remains an empirical question.

Chinese Translation

翻译学生需要学习如何使用翻译技术以及如何评判这些技术所提供的选择。本文介绍了LoopCAT，这是一种基于Apache-2.0许可证的、本地优先的计算机辅助翻译环境，采用OpenAI Codex共同创建，使用了GPT-5.5和GPT-5.6，并提出了一个将工作流程能力、评估判断和技术代理连接起来的框架。该研究基于代码库历史、实施检查和已识别开发版本的验证记录。LoopCAT结合了本地项目存储、翻译记忆、术语、质量保证、文档交换以及与本地或托管AI服务的可选连接。其英语、加泰罗尼亚语和土耳其语界面目录也使得该应用程序本身可作为教学材料：学生可以将英语用户界面字符串翻译成另一种语言，审查现有的自动生成目标草稿，导入他们的修订，并测试界面。我们围绕四种参与形式组织这些机会：操作工作流程、评估输出、检查和配置机制，以及进行或辩护有限干预。六个会话序列、一个用户界面本地化任务、一个占位符示例和一个评估标准具体说明了教师如何使用该框架。本文将实现的能力与提议的教育益处分开；并报告没有新的学生学习成果。它区分了最新的软件包检查与早期回归证据，并制定了课堂评估的协议。LoopCAT提供了一个可检查的环境，用于教授翻译决策如何与数据、接口和软件规则相互作用。这些活动是否改善了判断、转移或参与仍然是一个实证问题。

View on arXiv Download PDF AI Translation

cs.CL / 35 / 2609.00351

Detecting Hidden Behaviors in LLMs via Activation-matched Finetuning

通过激活匹配微调检测大型语言模型中的隐藏行为

Haselhorst, Robin, Flek, Lucie, Mai, Florian

Abstract

Large language models can hide hidden behaviors that activate only under narrow conditions, such as backdoor triggers, sleeper-agent deployment cues, sandbagging, or topic-conditioned censorship. Such behaviors are difficult to detect without prior knowledge what to look for. We present activation-matched finetuning, an unsupervised detection method that assumes no knowledge of the trigger or the target behavior. Given a suspect model and a publicly available anchor, we finetune the anchor to reproduce the suspect's activations on a small benign corpus, and score each evaluation prompt by the residual between the two models. Since no benign corpus covers the sparse trigger region, the reference learns the benign computation but not the hidden behavior. Therefore, trigger prompts -- and, crucially, their semantic neighbors -- incur a large residual that signal the presence of unusual behavior to the defender. Testing our method across third-party models and custom models, activation-matched finetuning surfaces hidden behavior reliably. Furthermore, we empirically consider a natural defense-aware attack and showcase that it fails to suppress our detection method without sacrificing the behavior itself.

Chinese Translation

大型语言模型可能隐藏在狭窄条件下才会激活的行为，例如后门触发器、卧底代理部署线索、故意隐瞒或主题条件审查。这些行为在没有先前知识的情况下很难被检测到。我们提出了激活匹配微调，这是一种无监督检测方法，假设对触发器或目标行为没有任何知识。给定一个可疑模型和一个公开可用的锚点，我们对锚点进行微调，以在一个小的良性语料库上重现可疑模型的激活，并通过两个模型之间的残差对每个评估提示进行评分。由于没有良性语料库覆盖稀疏的触发区域，因此参考模型学习的是良性的计算，而不是隐藏的行为。因此，触发提示——以及其语义邻近的提示——会产生较大的残差，向防御者发出异常行为存在的信号。在对第三方模型和自定义模型测试我们的方法时，激活匹配微调可靠地揭示了隐藏行为。此外，我们还经验性地考虑了一种自然的防御意识攻击，并展示了该攻击在不牺牲行为本身的情况下无法抑制我们的检测方法。

View on arXiv Download PDF AI Translation

cs.CL / 36 / 2609.00367

Neurosymbolics for Data Engineering: Achieving Long Context Token Reduction Without Finetuning

数据工程中的神经符号学：在不进行微调的情况下实现长上下文令牌减少

Bhat, Vishvesh

Abstract

Large Language Models are increasingly deployed for sophisticated data engineering tasks such as generating structured queries from natural language, Text-to-SQL, and automating complex spreadsheet operations. However, maximizing their utility demands both higher finetuning-free accuracy and solutions to the computational bottleneck imposed by the Transformer architectures inherent quadratic (On2) time complexity. This paper introduces a novel drop-in neurosymbolic layer designed to seamlessly integrate into existing LLM backbones enhancing logical reasoning and mitigating long-context resource consumption. On the reasoning front, the layer immediately and significantly improves performance yielding an average accuracy increase of 85% across rigorous benchmarks including BIRD-CRITIC and LiveSQLBench, critically achieving these gains without any task specific finetuning or RLHF. Concurrently, we repurpose this approach to address the severe computational strain of long context inference. By leveraging symbolic processing to prioritize and compress relevant contextual information the layer reduces the effective token usage by over 50% and brings the effective time complexity down from O(n2) to approximately O(n) on certain long context tasks. This dual impact approach not only makes LLMs substantially more reliable for data engineering but also drastically reduces the computational pressure on inference chips, making long context tasks more manageable and cost effective.

Chinese Translation

大型语言模型越来越多地被应用于复杂的数据工程任务，例如从自然语言生成结构化查询、文本到 SQL（Text-to-SQL）以及自动化复杂的电子表格操作。然而，最大化其效用需要更高的无微调准确性，并解决由变换器架构固有的平方（O(n²)）时间复杂度所带来的计算瓶颈。本文介绍了一种新颖的可插拔神经符号层，旨在无缝集成到现有的大型语言模型（LLM）骨干网络中，以增强逻辑推理并减轻长上下文的资源消耗。在推理方面，该层立即显著提高了性能，在包括 BIRD-CRITIC 和 LiveSQLBench 在内的严格基准测试中，平均准确性提高了 85%，并且在没有任何特定任务微调或强化学习人类反馈（RLHF）的情况下实现了这些增益。同时，我们重新利用这种方法来应对长上下文推理的严重计算压力。通过利用符号处理来优先考虑和压缩相关的上下文信息，该层将有效令牌使用量减少了超过 50%，并将某些长上下文任务的有效时间复杂度从 O(n²) 降低到大约 O(n)。这种双重影响的方法不仅使大型语言模型在数据工程中更加可靠，还大幅降低了推理芯片的计算压力，使长上下文任务更易于管理且成本更具效益。

View on arXiv Download PDF AI Translation

cs.CL / 37 / 2609.00378

Removable and Irreducible: A Token-Cost Ledger for the Multilingual Tokenization Tax

可移除与不可简化：多语言标记化税的标记成本账本

Mandarapu, Madhulatha, Kunkunuru, Sandeep

Abstract

Large language models pay a well-documented tax on non-English text: the same content costs several times more tokens, and because attention is quadratic in sequence length, far more compute. We ask how much of this tax is removable. Framing the token layer as source coding -- transformer compute is monotone in sequence length, whose per-atom floor is the Shannon rate $H/\log_2 V$, an object already applied to tokenizers in prior work -- we assemble a token-cost ledger that splits each language's cost, at fixed parallel content, into a removable coding redundancy, a residual coding slack, an intrinsic-content term, and an orthogonal, irreducible grapheme-to-phoneme term that governs the multimodal rather than the text cost. On FLORES-200 across eight languages, a production tokenizer costs up to $8.9\times$ more tokens for Indic scripts than for English; a script-matched code trained on $1,012$ sentences removes a median $64\%$ of that excess (bootstrap 95\% CI $[0.638, 0.647]$), and a script-fair information floor shows the intrinsic content differs by under $6\%$ -- the tax is representational, not informational. A constructed code removes $98\%$ of a controlled source's redundancy, and the token tax implies up to $79\times$ attention cost. We are explicit about scope and failure: this is compute-and-memory accounting, not a model-quality claim; we neither measure nor claim the cross-lingual direction of the orthographic term; and our matched code is a conservative small-data demonstration. We contribute the unifying ledger, the removable-versus-intrinsic attribution, and an open one-command harness.

Chinese Translation

大型语言模型在处理非英语文本时面临着众所周知的成本：相同内容的标记数量要高出几倍，并且由于注意力机制在序列长度上是二次的，因此计算成本也要高得多。我们探讨这一成本中有多少是可移除的。将标记层视为源编码——变换器的计算在序列长度上是单调的，其每个原子的下限是香农速率 $H/ ext{log}_2 V$，这一概念在之前的工作中已经应用于标记器——我们构建了一个标记成本账本，将每种语言的成本在固定并行内容下分解为可移除的编码冗余、剩余编码松弛、内在内容项，以及一个正交的、不可简化的字形到音素的项，该项控制多模态而非文本成本。在 FLORES-200 数据集中，八种语言的生产标记器对于印地语脚本的标记成本比英语高出多达 $8.9 imes$；一个基于 $1,012$ 个句子的脚本匹配代码去除了中位数 $64\%$ 的多余成本（自助法 95 ext{%} 置信区间 $[0.638, 0.647]$），而脚本公平的信息下限显示内在内容的差异低于 $6 ext{%}$——这项成本是表征性的，而非信息性的。构建的代码去除了 $98 ext{%}$ 的受控源冗余，而标记税暗示了高达 $79 imes$ 的注意力成本。我们明确了范围和局限性：这是一种计算和内存的核算，而不是模型质量的声明；我们既不测量也不声称正字法项的跨语言方向；而我们的匹配代码是一个保守的小数据演示。我们贡献了统一的账本、可移除与内在归因的区分，以及一个开放的一键式工具。

View on arXiv Download PDF AI Translation

cs.CL / 38 / 2609.00416

Late Transformer Layers Recode Syntax Canonically: Evidence from Greek Scrambling and Cross-Layer Generalisation

晚期变换器层以规范方式重新编码句法：来自希腊语词序变换和跨层泛化的证据

Zacharopoulos, Christos Nikolaos, Kyriakoglou, Revekka, Tsoukala, Chara, Desbordes, Théo

Abstract

Probing studies have established that syntactic information is decodable in early and middle transformer layers, but what happens to that information in later layers remains poorly understood. We apply a cross-layer generalisation analysis to three Greek-tuned large language models evaluated on tightly controlled minimal pairs: object-relative constructions in Modern Greek, where canonical (Subject-Verb-Object; SVO) and non-canonical (Verb-Subject-Object; VSO) orders differ only in within-clause word order, while preserving propositional meaning. When a probe trained on late layers (20-31) is tested on each early layer individually, it produces below-chance transfer (cluster-corrected, p<0.01), classifying 99.3% of non-canonical sentences as canonical. Probe coefficients reverse sign around layer 22, indicating a directional recoding toward the canonical form rather than simple information loss. These findings characterise a representational format change in late transformer layers that goes beyond the well-established decline in syntactic decodability, and they generate a directly testable prediction for human EEG and MEG decoding studies using the same stimuli. Code and stimuli are publicly available on OSF.

Chinese Translation

探测研究已确立句法信息可以在早期和中期变换器层中解码，但在后期层中这些信息的去向仍然不甚了解。我们对三个经过希腊语调优的大型语言模型应用了跨层泛化分析，这些模型在严格控制的最小对比中进行评估：现代希腊语中的对象相对结构，其中规范（主-谓-宾；SVO）和非规范（谓-主-宾；VSO）顺序仅在从句内的词序上有所不同，同时保持命题意义。当在后期层（20-31）上训练的探测器在每个早期层上单独测试时，产生了低于随机水平的转移（经过聚类校正，p<0.01），将99.3%的非规范句子分类为规范句。探测器系数在第22层附近反转符号，表明朝向规范形式的定向重新编码，而非简单的信息丢失。这些发现表征了晚期变换器层中表示格式的变化，超越了已确立的句法可解码性下降，并为使用相同刺激的人类脑电图（EEG）和脑磁图（MEG）解码研究生成了可直接测试的预测。代码和刺激材料已在开放科学框架（OSF）上公开提供。

View on arXiv Download PDF AI Translation

cs.CL / 39 / 2609.00443

(V)LMs generalize beyond surface co-occurrence: Evidence from cross-modal number agreement

(V)LMs超越表面共现进行泛化：来自跨模态数一致性的证据

Studdiford, Zach, Misra, Kanishka

Abstract

Language models learn about grammatical number primarily from co-occurrence, and show frequency effects as a result---sometimes taken to indicate that they do not learn abstract ``rules'', and are instead dependent on specific lexical items. Testing generalization with text stimuli alone cannot settle this debate, since distributional cues (is/are, this/these) easily give number away. We instead use cross-modal generalization as a tool to investigate abstractions in LMs that can also accept visual inputs (VLMs), restricting the evidence that diagnoses number to an extra-linguistic modality. We teach VLMs pairs of new nouns by adding new embeddings and only updating them during learning, comparing conditions where number is diagnosed by visual cues alone against ones where it is disambiguated by text. Across behavior, representational dynamics, and causal mechanisms, we find non-trivial evidence for cross-modal generalization across both exposure conditions, and that linguistic vs. extra-linguistic cue conditions are treated in similar ways in the internal mechanisms of the model. This suggests that statistical learners like VLMs can generalize beyond surface-level co-occurrence and show genuine abstraction-compatible behavior.

Chinese Translation

语言模型主要通过共现学习语法数，并因此显示出频率效应——这有时被认为表明它们并不学习抽象的“规则”，而是依赖于特定的词汇项。仅用文本刺激测试泛化无法解决这一争论，因为分布线索（is/are, this/these）很容易揭示数的特征。我们因此使用跨模态泛化作为工具，研究可以接受视觉输入的语言模型（VLMs）中的抽象，限制诊断数的证据仅来自额外语言模态。我们通过添加新的嵌入并仅在学习过程中更新它们来教VLMs一对新的名词，比较仅通过视觉线索诊断数的条件与通过文本消歧的条件。在行为、表征动态和因果机制方面，我们发现跨越两种暴露条件的跨模态泛化有非平凡的证据，并且语言与额外语言线索条件在模型的内部机制中以相似的方式处理。这表明，像VLMs这样的统计学习者可以超越表面层次的共现进行泛化，并表现出真正兼容抽象的行为。

View on arXiv Download PDF AI Translation

cs.CL / 40 / 2609.00454

Location-Aware Language Models via Secondary Embeddings

通过次级嵌入实现位置感知语言模型

Srinivasagan, Gokul, Georges, Munir

Abstract

Pretrained transformer-based language models achieve strong performance across a wide range of NLP tasks but remain limited in encoding geo-locational semantics, leading to suboptimal representations of place names and spatial entities. In this work, we propose a lightweight, model-agnostic approach for injecting geo-spatial awareness into pretrained embeddings without modifying the tokenizer or requiring costly retraining. Our method augments input representations with structured geographic signals by combining location names with their corresponding latitude and longitude, and employs a location-focused masking to better align textual representations with real-world spatial relationships. This design allows the model to incorporate geo-spatial context while preserving existing semantic and syntactic knowledge. Experimental results demonstrate substantial improvements in geo-spatial alignment while maintaining comparable performance on standard NLP benchmarks such as GLUE. The method is computationally efficient, requiring only minutes of additional training, and generalizes across multiple model architectures and scales.

Chinese Translation

预训练的基于变换器的语言模型在广泛的自然语言处理（NLP）任务中表现出色，但在编码地理位置语义方面仍然存在局限性，导致地名和空间实体的表示不够理想。在本研究中，我们提出了一种轻量级、模型无关的方法，通过在不修改分词器或不需要昂贵的重新训练的情况下，将地理空间意识注入预训练嵌入。我们的方法通过将位置名称与其对应的纬度和经度结合，增强输入表示，并采用以位置为中心的掩蔽策略，以更好地对齐文本表示与现实世界的空间关系。这种设计使模型能够在保留现有语义和句法知识的同时，融入地理空间上下文。实验结果表明，在保持与标准NLP基准（如GLUE）相当的性能的同时，地理空间对齐有显著改善。该方法计算效率高，仅需几分钟的额外训练，并且能够在多种模型架构和规模上推广。

View on arXiv Download PDF AI Translation

cs.CL / 41 / 2609.00463

Toppling the Hierarchy in Byte-level Language Modeling

颠覆字节级语言建模中的层级结构

Edman, Lukas, Fraser, Alexander

Abstract

This work examines recent byte-level models and their failure to perfectly manipulate characters. State-of-the-art byte-level models use a hierarchical structure, starting at the byte level, downsampling to the word level, and then upsampling back to bytes. While this improves training and inference efficiency, we find that the hierarchical design itself limits character-level understanding, with pure byte-level models consistently outperforming hierarchical variants on character manipulation tasks. Ablating transformer layers into attention and feed-forward components further reveals that byte-level attention is the primary mechanism driving this behavior. Together, our results provide an explanation for the character-level failures of hierarchical byte models and establish a clear trade-off between computational efficiency and fine-grained character understanding.

Chinese Translation

本研究考察了近期的字节级模型及其在字符操作方面的不足。当前最先进的字节级模型采用层级结构，从字节级开始，降采样到词级，然后再升采样回字节。尽管这提高了训练和推理的效率，但我们发现层级设计本身限制了对字符级的理解，纯字节级模型在字符操作任务中始终优于层级变体。对变压器层进行消融实验，将其分解为注意力和前馈组件，进一步揭示了字节级注意力是驱动这一行为的主要机制。综合来看，我们的结果为层级字节模型在字符级上的失败提供了解释，并明确了计算效率与细粒度字符理解之间的权衡。

View on arXiv Download PDF AI Translation

cs.CL / 42 / 2609.00470

TRIS: A Tri-Layer Retrieval Integrity Sieve Against Knowledge Poisoning

TRIS：一种针对知识中毒的三层检索完整性筛选器

Munir, Muhaimin Bin, Ononto, Akib Jawad, Joynab, Nazia Shehnaz, Thuraisingham, Bhavani, Khan, Latifur

Abstract

Retrieval-Augmented Generation (RAG) grounds large language models in external corpora, but implicit trust in retrieved documents creates a critical attack surface: PoisonedRAG shows that a handful of crafted passages can dominate dense retrieval and steer generation toward attacker-chosen answers. We present the Tri-Layer Sieve, a middleware defense that sanitizes retrieved evidence through cross-embedding-space clustering with an independent judge model, structural filtering of trigger-payload artifacts, and LLM consistency verification. The design exploits a key weakness of retrieval-stage poisoning: a single document must satisfy one embedding geometry, one internal Trigger-Payload structure, and one generation objective - rarely all three simultaneously, a fragility that persists even against an adaptive attacker who paraphrases around it. On Natural Questions, HotpotQA, and MS-MARCO with Contriever retrieval (k=50), the Sieve reduces black-box Attack Success Rate from 67.0/87.0/64.0% to 3.0/14.0/4.0%, mitigates white-box HotFlip attacks from ~74% to 27.8% on NQ with Layer 3 enabled, and drives poisoned-document MRR to 0.000, while restoring clean accuracy from 13-33% under attack to 58-76%. Under an architecture-aware adversary who paraphrases triggers to evade the structural filter, enabling the consistency layer halves adaptive ASR (32.0% to 15.0% on NQ) while raising clean accuracy by 18 points, at an added latency of ~16-19 s/query under live retrieval.

Chinese Translation

检索增强生成（RAG）将大型语言模型与外部语料库相结合，但对检索文档的隐性信任创造了一个关键的攻击面：PoisonedRAG表明，少量精心制作的段落可以主导密集检索，并将生成引导至攻击者选择的答案。我们提出了三层筛选器（Tri-Layer Sieve），这是一种中间件防御机制，通过跨嵌入空间聚类与独立判断模型、触发-有效载荷伪影的结构过滤以及大型语言模型（LLM）一致性验证来清理检索到的证据。该设计利用了检索阶段中毒的一个关键弱点：单个文档必须满足一种嵌入几何、一种内部触发-有效载荷结构和一种生成目标——这三者同时满足的情况极为罕见，这种脆弱性在适应性攻击者进行改写时依然存在。在Natural Questions、HotpotQA和MS-MARCO的Contriever检索（k=50）上，筛选器将黑箱攻击成功率从67.0/87.0/64.0%降低到3.0/14.0/4.0%，将白箱HotFlip攻击在NQ上从约74%减轻到27.8%（启用第3层时），并将中毒文档的平均检索排名（MRR）降至0.000，同时将攻击下的清晰准确率从13-33%恢复到58-76%。在一个架构感知的对手面前，该对手改写触发以规避结构过滤，启用一致性层将适应性攻击成功率减半（从32.0%降至15.0%），同时将清晰准确率提高18个百分点，增加的延迟约为每次检索16-19秒。

View on arXiv Download PDF AI Translation

cs.CL / 43 / 2609.00474

Exploring Collaboration between a language and a non-language agent

探索语言代理与非语言代理之间的协作

S I, Harini, Singh, Somesh, Singla, Yaman K, Shah, Rajiv Ratn, Doermann, David, Krishnamurthy, Balaji

Abstract

LLMs are increasingly deployed as orchestrators that coordinate specialized subagents to solve complex tasks through natural language. However, in many important domains like game playing and robotics, the strongest available agents are not language models. Integrating non-language agents with LLMs would require \emph{verbalization}: compressing their rich continuous representations into sparse textual summaries at each interaction step. To study whether verbalization constitutes a bottleneck, we introduce \textsc{LLAMIA-Bench}, a suite of six diverse collaborative chess tasks spanning three facets: behavioral imitation, state assessment, and natural-language explanation. Each task instantiates a well-established chess problem that neither the LLM nor the chess engine can solve alone. To solve LLM collaboration with non-language agents, we introduce \emph{latent state internalization}, which projects the subagent's continuous representations directly into the LLM's token stream as learned state tokens, with dynamic re-encoding as actions advance the environment state. Comparing internalization to verbalized integration, our experiments reveal a consistent \emph{verbalization debt}: the performance gap widens throughout training and persists as the LLM scales from 4B to 14B parameters. A single 14B model, \textsc{LLAMIA}, trained with latent state internalization, matches or exceeds task specialists and frontier models including GPT-5.1 with tool access across all benchmark tasks, and generalizes out-of-distribution where task-specific finetunes collapse

Chinese Translation

大型语言模型（LLMs）越来越多地被用作协调者，通过自然语言协调专门的子代理以解决复杂任务。然而，在许多重要领域，如游戏和机器人技术，最强大的可用代理并不是语言模型。将非语言代理与LLMs集成需要进行 extit{语言化}：在每个交互步骤中将其丰富的连续表示压缩为稀疏的文本摘要。为了研究语言化是否构成瓶颈，我们引入了 extsc{LLAMIA-Bench}，这是一个涵盖三方面的六个多样化协作国际象棋任务的套件：行为模仿、状态评估和自然语言解释。每个任务都实例化了一个成熟的国际象棋问题，LLM和国际象棋引擎都无法单独解决。为了解决LLM与非语言代理的协作问题，我们引入了 extit{潜在状态内化}，它将子代理的连续表示直接投影到LLM的令牌流中作为学习到的状态令牌，并在动作推进环境状态时进行动态重新编码。将内化与语言化集成进行比较，我们的实验揭示了一种持续的 extit{语言化债务}：性能差距在训练过程中不断扩大，并在LLM从40亿参数扩展到140亿参数时持续存在。一个单一的140亿模型 extsc{LLAMIA}，通过潜在状态内化进行训练，在所有基准任务中与任务专家和前沿模型（包括具有工具访问的GPT-5.1）相匹配或超越，并在任务特定微调崩溃的情况下实现了分布外泛化。

View on arXiv Download PDF AI Translation

cs.CL / 44 / 2609.00482

Are Near-Tied LLM Rankings Robust to Family-DIF-Guided Benchmark Recomposition?

近似平局的语言模型排名对家庭差异引导的基准重组是否稳健？

Zheng, Qiaoyuan, Yang, Yiqu

Abstract

Small leaderboard gaps are often interpreted as evidence that one language model is better than another, but their sign may depend on which benchmark items are included. We test this using item-level responses from five benchmarks and a family-label-free spectral approximation to multidimensional item-response theory (MIRT). In owner-disjoint folds, one owner half identifies items with low residual differential item functioning across model families (low-DIF); the resulting frozen, source- and easiness-balanced weights score models in the other half, while equally short matched-random subtests control for generic subtest variation. Full-benchmark and low-DIF rankings remain strongly correlated ($\tau_b=.900$--$.948$). Yet in four of five benchmarks, 30.9--47.1\% of cross-family pairs initially within one percentage point reverse order, exceeding their matched-random medians by 16.9--28.6 percentage points (all $p=.001$). The fifth benchmark shows no reliable excess ($-0.9$ points, $p=.689$). The pattern survives all pre-specified population perturbations, and residual item--family signatures replicate across owner halves; however, no family shows a consistent advantage across benchmarks. Thus, globally stable rankings can still leave individual near-tie orderings sensitive to benchmark composition, and sub-one-point leaderboard gaps should be accompanied by evidence that the implied ordering is composition-robust.

Chinese Translation

小的排行榜差距常被解读为一个语言模型优于另一个模型的证据，但其符号可能依赖于所包含的基准项目。我们使用来自五个基准的项目级响应和一种无家庭标签的光谱近似多维项目反应理论（MIRT）来测试这一点。在所有者不重叠的折叠中，一个所有者的半部分识别出在模型家族之间具有低残差差异项目功能（低-DIF）的项目；由此产生的冻结、源和易用性平衡的权重在另一半中对模型进行评分，同时同样短的匹配随机子测试控制一般子测试变异。完整基准和低-DIF 排名保持强相关性（$ au_b=.900$--$.948$）。然而，在五个基准中的四个中，30.9--47.1 ext{%} 的跨家族配对最初在一个百分点内反转顺序，超过其匹配随机中位数16.9--28.6个百分点（均为 $p=.001$）。第五个基准没有显示出可靠的超额（$-0.9$ 点，$p=.689$）。这一模式在所有预先指定的人群扰动中依然存在，残差项目-家族特征在所有者的半部分中复制；然而，没有任何家庭在各基准中表现出一致的优势。因此，全球稳定的排名仍然可能使个别近似平局的排序对基准组成敏感，而小于一个百分点的排行榜差距应伴随有证据表明所隐含的排序是组成稳健的。

View on arXiv Download PDF AI Translation

cs.CL / 45 / 2609.00487

EvoFlint: An Evolutionary Atlas of Multi-Turn LLM Vulnerabilities

EvoFlint：多轮大型语言模型脆弱性的进化图谱

Qiao, Feitong, Peng, Liren, Ren, Shiming, Jadhav, Aishwarya, Bahadorinejad, Arghavan, Chen, Marinette, Zhang, Muhan, Suria, Abdulaziz, Lu, Gennevi, Sarma, Anish Das

Abstract

Frontier language models that refuse harmful single-turn prompts often comply when the same intent is reached gradually over many turns, making multi-turn attacks one of the least understood failure modes of large language models. Most automated red-teaming methods treat this as a generation problem: produce attacks that break the model. We argue it is better framed as a search problem: discover, organize, and iteratively refine a diverse archive of attack strategies, producing a structured map of how a target model fails rather than a list of one-off successes. We introduce EvoFlint, which applies evolutionary quality-diversity search to multi-turn red-teaming. Attack strategies are phased conversation plans, not raw prompts, and are evolved through LLM-driven mutation and crossover. A Pareto fitness over attack success rate and peak severity preserves selection signal from near-miss attacks. A risk-indexed archive runs novelty search with local competition over strategy description embeddings inside each cell, maintaining diversity without committing to a predefined style taxonomy. A generation-level memory accumulates target-model insights across the population and feeds them back into strategy generation. On the HarmBench-test split, EvoFlint reaches attack success rates of 35.8% on Claude Sonnet 4.6, 59.7% on GPT-5.4, and 94.3% on Qwen3-32B, alongside 98.7% on the older GPT-4o included as a baseline reference. The resulting archive, organized by risk category, exposes for each target which categories of harm its safety training has and has not covered.

Chinese Translation

拒绝有害单轮提示的前沿语言模型在通过多轮逐渐达到相同意图时往往会妥协，这使得多轮攻击成为大型语言模型最不为人知的失败模式之一。大多数自动化红队方法将其视为生成问题：生成破坏模型的攻击。我们认为更好的框架是将其视为搜索问题：发现、组织并迭代优化多样化的攻击策略档案，生成目标模型失败的结构化地图，而不是一次性成功的列表。我们引入了EvoFlint，它将进化质量多样性搜索应用于多轮红队。攻击策略是分阶段的对话计划，而不是原始提示，通过大型语言模型驱动的变异和交叉进行进化。对攻击成功率和峰值严重性的帕累托适应度保留了来自近乎成功攻击的选择信号。一个风险索引档案在每个单元内的策略描述嵌入上进行新颖性搜索，并通过局部竞争保持多样性，而不依赖于预定义的风格分类。生成级别的记忆在整个种群中积累目标模型的见解，并将其反馈到策略生成中。在HarmBench测试分割上，EvoFlint在Claude Sonnet 4.6上达到35.8%的攻击成功率，在GPT-5.4上达到59.7%，在Qwen3-32B上达到94.3%，同时在作为基线参考的旧版GPT-4o上达到98.7%。最终的档案按风险类别组织，揭示了每个目标模型的安全训练覆盖了哪些和未覆盖哪些伤害类别。

View on arXiv Download PDF AI Translation

cs.CL / 46 / 2609.00491

MemeBridge: A Dataset for Benchmarking and Mitigating the Bidirectional Cultural Gap in Meme Interpretation

MemeBridge：用于基准测试和减轻表情包解读中双向文化差距的数据集

Zhu, Hangxiao, Qin, Suliu, Li, Zhuoyan, Jiang, Ming, Zhang, Yu, Xia, Meng

Abstract

Communicating across cultures is inherently challenging, especially through culturally dense and ambiguous formats like memes. While people expect large language models (LLMs) to hold promise for bridging such gaps, existing benchmark datasets often fail to capture the cultural context necessary for accurate interpretation. To address this, we introduce MemeBridge, a curated dataset centered on U.S.-originated memes, designed to capture two complementary perspectives: (1) how Chinese participants interpret these memes, and (2) how U.S. participants anticipate how people from other cultures might misunderstand them. Here, context refers to implicit cultural knowledge, including background beliefs, norms, and shared assumptions that shape meme comprehension. The dataset was constructed via a multi-stage crowdsourcing pipeline with rigorous validation, including human agreement checks and GPT-based classification verification. Each meme is annotated with sentiment, emotion, cultural significance, and knowledge type, providing rich supervision for downstream tasks. Notably, we observe that the anticipated misunderstandings from U.S. participants are often inaccurate, highlighting the asymmetries in cultural understanding and the challenges of adopting perspectives beyond one's own. This bidirectional framing, which focuses on both expression and perception, enables more nuanced benchmarking of cross-cultural comprehension. Our probing of multiple LLMs reveals that while models developed in different cultural contexts exhibit partial cross-cultural understanding, they often struggle with sophisticated interpretations. By contrast, fine-tuning with MemeBridge improves model performance, underscoring the value of culturally grounded resources for training and evaluating LLMs in globally diverse settings.

Chinese Translation

跨文化交流本质上具有挑战性，尤其是通过像表情包这样文化密集且模糊的格式。尽管人们期望大型语言模型（LLMs）能够在弥合这些差距方面发挥作用，但现有的基准数据集往往未能捕捉到准确解读所需的文化背景。为了解决这一问题，我们推出了MemeBridge，这是一个以美国起源的表情包为中心的精心策划的数据集，旨在捕捉两种互补的视角：（1）中国参与者如何解读这些表情包，以及（2）美国参与者如何预期其他文化的人可能会误解它们。在这里，背景指的是隐含的文化知识，包括塑造表情包理解的背景信念、规范和共享假设。该数据集通过多阶段众包流程构建，并进行了严格的验证，包括人类一致性检查和基于GPT的分类验证。每个表情包都标注了情感、情绪、文化重要性和知识类型，为下游任务提供了丰富的监督。值得注意的是，我们观察到美国参与者预期的误解往往不准确，突显了文化理解中的不对称性以及超越自身视角的挑战。这种双向框架关注表达和感知，使跨文化理解的基准测试更加细致。我们对多个LLMs的探测表明，尽管在不同文化背景下开发的模型表现出部分跨文化理解，但它们在复杂解读方面常常面临困难。相比之下，使用MemeBridge进行微调提高了模型性能，强调了在全球多样化环境中训练和评估LLMs时，文化根植资源的价值。

View on arXiv Download PDF AI Translation

cs.CL / 47 / 2609.00494

Human-Anchored Factuality Evaluation with Strategic Annotation

基于人类的事实性评估与战略注释

Wang, Yu, Erickson, Craig, Small, Kevin

Abstract

LLM-based factuality judges provide scalable evaluation signals, but their metrics are often systematically biased relative to human judgments. We study human-anchored factuality evaluation under limited annotation budgets, where judge predictions on the full dataset are combined with human labels on a small selectively sampled subset to obtain statistically valid estimates. The efficiency of this approach depends critically on which examples receive human annotation: in factuality evaluation, judge-human misalignment is not driven solely by low confidence, but also by structured failure modes such as incomplete evidence, temporal mismatch, unverifiable claims, and rubric misalignment. To exploit this structure, we introduce a factuality-specific annotation policy design pipeline that uses failure-space analysis (FSA) to derive diverse predictive signals for modeling human-judge misalignment. On an internal reference-based factuality evaluation system (AutoFA) and RAGTruth, where judge-predicted estimates substantially underestimate human-annotated factual accuracy, our FSA-guided policy improves annotation efficiency over uniform sampling and uncertainty-driven baselines, achieving effective-sample-size gains of 40.3% on AutoFA and 27.1% on RAGTruth.

Chinese Translation

基于大型语言模型（LLM）的事实性评估者提供了可扩展的评估信号，但其指标相对于人类判断往往存在系统性偏差。我们研究了在有限注释预算下的人类锚定事实性评估，其中评估者对完整数据集的预测与对小规模选择性抽样子集的人类标签相结合，以获得统计有效的估计。这种方法的效率在很大程度上依赖于哪些示例接受人类注释：在事实性评估中，评估者与人类之间的不一致不仅仅是由低置信度驱动的，还受到结构性失败模式的影响，例如证据不完整、时间不匹配、不可验证的主张和评分标准不一致。为了利用这种结构，我们引入了一种特定于事实性的注释策略设计流程，该流程使用失败空间分析（FSA）来推导多样化的预测信号，以建模人类与评估者之间的不一致。在一个基于内部参考的事实性评估系统（AutoFA）和RAGTruth中，评估者预测的估计显著低于人类注释的事实准确性，我们的FSA引导策略在注释效率上优于均匀抽样和基于不确定性的基线，在AutoFA上实现了40.3%的有效样本量增益，在RAGTruth上实现了27.1%的增益。

View on arXiv Download PDF AI Translation

cs.CL / 48 / 2609.00495

Beyond Token Positions: Safety Alignment Across Denoising Steps in Diffusion Language Models

超越标记位置：扩散语言模型去噪步骤中的安全对齐

Wang, Guoli, Shi, Haonan, Ouyang, Tu, Wang, An

Abstract

Diffusion large language models (dLLMs) generate text through iterative denoising rather than left-to-right decoding. This generation paradigm introduces two axes that can influence safety alignment: when tokens are generated during denoising and where they appear in the response. In this paper, we measure dLLM safety behavior under harmful prompts by tracing intermediate token distributions and commitment decisions throughout denoising. Our analysis shows that refusal signals are concentrated in early denoising steps and leading response positions, and the tokens committed early can strongly shape the final safety outcome. Our measurements further show that the denoising step and persistence of refusal-token commitment are important for understanding dLLM safety. Based on these findings, we propose Refusal-Aware Early Commitment (RAEC), a simple training-free decoding method that commits persistent refusal signals from early steps. Experiments on LLaDA and Dream show that RAEC reduces attack success rates while largely preserving utility. The code is available at https://github.com/Glresearch1/RAEC.

Chinese Translation

扩散大型语言模型（dLLMs）通过迭代去噪生成文本，而不是采用从左到右的解码方式。这种生成范式引入了两个可以影响安全对齐的维度：在去噪过程中何时生成标记，以及它们在响应中出现的位置。在本文中，我们通过追踪去噪过程中的中间标记分布和承诺决策，测量了dLLM在有害提示下的安全行为。我们的分析表明，拒绝信号集中在早期的去噪步骤和响应的前导位置，而早期承诺的标记可以强烈影响最终的安全结果。我们的测量进一步显示，去噪步骤和拒绝标记承诺的持续性对于理解dLLM的安全性至关重要。基于这些发现，我们提出了拒绝信号感知早期承诺（Refusal-Aware Early Commitment, RAEC），这是一种简单的无训练解码方法，能够承诺来自早期步骤的持续拒绝信号。在LLaDA和Dream上的实验表明，RAEC降低了攻击成功率，同时在很大程度上保持了实用性。代码可在 https://github.com/Glresearch1/RAEC 获取。

View on arXiv Download PDF AI Translation

cs.CL / 49 / 2609.00515

The Interlingua Hypothesis: LLMs Translate via a Latent Task-agnostic Feature Space

中介语假设：大型语言模型通过潜在的任务无关特征空间进行翻译

Brinton, Jacob, Brinkmann, Jannik, Crovella, Mark, Mueller, Aaron

Abstract

Large language models (LLMs) have recently demonstrated improved machine translation performance over strong supervised baselines. This raises questions as to what mechanisms underlie how LLMs perform machine translation between languages. Motivated by recent interpretability findings--namely, that LLMs use massively multilingual latent feature representations to perform language modeling--we propose the interlingua hypothesis. The hypothesis holds that language models translate by reading a source sentence into a latent feature space, and generate a target sentence by reading from the latent feature space. We show three lines of evidence in support of this hypothesis: (1) variance in BLEU across language pairs is largely predictable from language-specific competences with no language pair-specific interaction terms; (2) many model components are causally influential in both monolingual tasks and translation tasks; and (3) fine-tuning on monolingual data recovers a large proportion of translation improvements relative to fine-tuning on aligned documents. Together, these provide convergent evidence in support of the interlingua hypothesis, and suggest new ways of understanding and improving how LLMs can be leveraged to perform translation tasks.

Chinese Translation

大型语言模型（LLMs）最近在机器翻译性能上超越了强大的监督基线。这引发了关于LLMs在语言之间进行机器翻译的机制的疑问。受近期可解释性研究的启发——即LLMs使用大规模多语言潜在特征表示进行语言建模——我们提出了中介语假设。该假设认为，语言模型通过将源句子读取到潜在特征空间中进行翻译，并通过从潜在特征空间中读取生成目标句子。我们提供了三条证据支持这一假设：（1）不同语言对之间的BLEU值差异在很大程度上可以通过语言特定能力进行预测，而不需要语言对特定的交互项；（2）许多模型组件在单语任务和翻译任务中都有因果影响；（3）在单语数据上进行微调相较于在对齐文档上进行微调，能够恢复大量翻译改进。综合来看，这些证据为中介语假设提供了趋同支持，并提出了理解和改进LLMs在翻译任务中应用的新方法。

View on arXiv Download PDF AI Translation

cs.CL / 50 / 2609.00549

Skill Following: Evaluating Actual Skill Use in Retrieval-Enabled LLM Agents

技能跟随：评估检索增强的语言模型代理中的实际技能使用

Cho, Seonghyeon, Park, Chanjun

Abstract

Large Language Model (LLM) agents increasingly rely on external skills, yet standard evaluations obscure whether retrieving these skills actually helps. Aggregate metrics often compare retrieved versus non-retrieved tasks, introducing severe selection bias and failing to isolate the true effect of skill use. To measure this actual-use capability-which we formalize as Skill Following (SF)-we introduce the Retrieval-Invoked Actual-Use Effect (RAE). RAE computes the same-task outcome difference between matched skill-enabled and skill-disabled executions, conditioned exclusively on tasks where the agent actively retrieved a skill. Evaluating 17 LLMs across coding and mathematical domains, we uncover a stark evaluation paradox: models frequently show positive aggregate retrieval lift but negative RAE. On MBPP+, multiple models that appear to benefit system-wide actually harm their own performance on the exact tasks where retrieval occurred. These findings demonstrate that aggregate averages can create a misleading illusion of tool-use proficiency, whereas RAE directly measures whether the retrieval-to-answer pipeline genuinely rescues more outcomes than it harms.

Chinese Translation

大型语言模型（LLM）代理越来越依赖外部技能，但标准评估方法模糊了检索这些技能是否真正有助于任务完成。总体指标通常比较检索与非检索任务，导致严重的选择偏差，并未能孤立出技能使用的真实效果。为了衡量这种实际使用能力——我们将其形式化为技能跟随（Skill Following, SF）——我们引入了检索引发的实际使用效应（Retrieval-Invoked Actual-Use Effect, RAE）。RAE 计算在仅基于代理主动检索技能的任务中，匹配的技能启用和技能禁用执行之间的同任务结果差异。在编码和数学领域评估了 17 个 LLM，我们发现了一个明显的评估悖论：模型经常显示出积极的总体检索提升，但 RAE 却是负值。在 MBPP+ 数据集上，多个看似在系统层面受益的模型实际上在检索发生的确切任务上损害了自身表现。这些发现表明，总体平均值可能会产生误导性的工具使用能力幻觉，而 RAE 则直接衡量检索到答案的流程是否真正拯救了更多结果而非造成损害。

View on arXiv Download PDF AI Translation

cs.CL / 51 / 2609.00550

Same Semantics, Different Outcome: On the Modality Robustness of Multimodal LLMs under Knowledge Conflict

相同语义，不同结果：多模态大型语言模型在知识冲突下的模态鲁棒性研究

Lee, Jungyeon, Yoon, Yejin, Kim, Taeuk

Abstract

Multimodal large language models (MLLMs) are increasingly provided with contextual evidence in heterogeneous forms: as a text passage, as a rendered image of the same passage, or as both together. However, it remains unclear how consistently these surface forms are processed, especially when the evidence conflicts with the model's parametric knowledge. We study modality robustness under knowledge conflict across 13 MLLMs and two datasets, and find them far from robust. (1) Contrary to common belief, models favor a context that contradicts parametric knowledge more readily in image form than in text form; (2) when a contradicting text and image are presented together, the preferred modality is essentially arbitrary, varying with input order, model, and dataset. We further demonstrate that this instability has practical consequences: it degrades performance in multimodal RAG and can be exploited by adversarial attacks. To alleviate this brittleness, we examine several simple techniques---prompting, steering, supervised fine-tuning (SFT), and direct preference optimization; the majority prove ineffective, whereas SFT achieves moderate success. We therefore call for greater awareness of this inconsistency and argue that it is fundamental, demanding attention at multiple training stages.

Chinese Translation

多模态大型语言模型（MLLMs）越来越多地以异构形式提供上下文证据：作为文本段落、作为相同段落的渲染图像，或两者结合。然而，这些表面形式在处理时的一致性仍然不明确，尤其是在证据与模型的参数知识发生冲突时。我们研究了13个MLLMs和两个数据集在知识冲突下的模态鲁棒性，发现它们远非鲁棒。（1）与普遍看法相反，模型更倾向于接受以图像形式呈现的与参数知识相矛盾的上下文，而非文本形式；（2）当矛盾的文本和图像同时呈现时，首选模态基本上是任意的，随输入顺序、模型和数据集而变化。我们进一步证明，这种不稳定性具有实际后果：它降低了多模态RAG的性能，并可能被对抗性攻击所利用。为了减轻这种脆弱性，我们考察了几种简单的技术——提示、引导、监督微调（SFT）和直接偏好优化；大多数证明无效，而SFT取得了适度成功。因此，我们呼吁对这种不一致性有更大的关注，并认为这是根本性的，值得在多个训练阶段引起重视。

View on arXiv Download PDF AI Translation

cs.CL / 52 / 2609.00551

EM^2Mem: Event-Centric Multimodal Memory for Large Language Models

EM^2Mem：面向事件的多模态记忆框架用于大型语言模型

Chen, Yijun, Zheng, Yaqi, Li, Yanya, Xiao, Boyi, Xu, Buqiang, Qiao, Shuofei, Fang, Jizhan, Deng, Xinle, Yao, Yunzhi, Wang, Xuehai, Zhang, Liuxin, Li, Hui, Chen, Huajun, Deng, Shumin

Abstract

Multimodal memory offers a scalable interface for long-video question answering, but existing methods often retrieve captions, frames, transcripts, summaries, or graph facts as isolated fragments. Although searchable, such fragments are not generation-ready: language models must reconstruct cross-modal and temporal alignments at inference time, when context is limited and attribution is difficult. We propose EM^2Mem, an event-centric multimodal memory framework that binds heterogeneous evidence to event anchors during memory construction. Each event-indexed memory cell aligns multimodal records, temporal context, graph-linked relations, semantic facts, and provenance, enabling compact evidence readout over grounded multimodal events rather than modality-specific fragments. Across three long-video QA benchmarks, EM^2Mem improves average accuracy over the strongest memory baseline by 2.0, 2.4, and 3.7 points, improves strict event-level Top-5 evidence recall by 7.0 points, and reduces per-query latency by 4.67 times and total inference tokens by 63.66% (The code will be integrated into https://github.com/zjunlp/LightMem).

Chinese Translation

多模态记忆为长视频问答提供了可扩展的接口，但现有方法通常将字幕、帧、转录文本、摘要或图形事实作为孤立片段进行检索。尽管这些片段是可搜索的，但并不适合生成：语言模型必须在推理时重建跨模态和时间对齐，而此时上下文有限且归因困难。我们提出了EM^2Mem，一种面向事件的多模态记忆框架，在记忆构建过程中将异构证据绑定到事件锚点。每个事件索引的记忆单元对齐多模态记录、时间上下文、图形关联关系、语义事实和来源，从而实现对基于事件的紧凑证据读取，而不是特定模态的片段。在三个长视频问答基准测试中，EM^2Mem在最强记忆基线的平均准确率上提高了2.0、2.4和3.7个百分点，严格的事件级Top-5证据召回率提高了7.0个百分点，查询延迟减少了4.67倍，总推理令牌减少了63.66%（代码将集成到https://github.com/zjunlp/LightMem）。

View on arXiv Download PDF AI Translation

cs.CL / 53 / 2609.00581

Enoki: Efficient Multi-Level Hallucination Detection

Enoki：高效的多层次幻觉检测

Rykov, Elisei, Ionov, Timur, Ivanov, Nikolay, Savkin, Maksim, Makarenko, Maksim, Panchenko, Alexander, Konovalov, Vasily, Belikova, Julia

Abstract

Ensuring factuality remains a critical challenge for deploying LLMs in high-stakes settings. Existing hallucination detectors usually operate at a single level: claim-level methods provide interpretable factual units, while span-level methods localize unsupported text. Bridging these views is costly, as LLM-heavy pipelines require multiple decomposition and verification calls, and modular systems need additional claim-to-span alignment. We propose Enoki, an Open Information Extraction framework for multi-level hallucination detection. Enoki extracts text-anchored relational facts, verifies them against evidence, and projects unsupported facts back to hallucinated spans. This shared representation enables claim-level verification and span-level localization without requiring separate alignment. Enoki supports LLM-based, encoder-based, and rule-based extraction regimes, balancing accuracy and inference cost through a common interface. Experiments show that Enoki remains competitive with strong claim-level systems while using fewer resources and achieves superior performance on fine-grained span- and entity-level localization. We also release EnokiQA, a dual-granularity dataset with aligned claim-level verification and span-level localization annotations.

Chinese Translation

确保事实性仍然是将大型语言模型（LLMs）应用于高风险环境中的一项关键挑战。现有的幻觉检测器通常在单一层次上运作：声明级方法提供可解释的事实单元，而跨度级方法则定位不支持的文本。弥合这些视角的成本较高，因为依赖LLM的流程需要多次分解和验证调用，而模块化系统需要额外的声明与跨度对齐。我们提出了Enoki，一个用于多层次幻觉检测的开放信息提取框架。Enoki提取文本锚定的关系事实，验证其与证据的一致性，并将不支持的事实投影回幻觉化的跨度。该共享表示使得声明级验证和跨度级定位能够在不需要单独对齐的情况下进行。Enoki支持基于LLM、基于编码器和基于规则的提取机制，通过一个共同接口在准确性和推理成本之间取得平衡。实验表明，Enoki在资源使用较少的情况下仍然与强大的声明级系统具有竞争力，并在细粒度的跨度和实体级定位上表现出优越的性能。我们还发布了EnokiQA，一个具有对齐的声明级验证和跨度级定位注释的双粒度数据集。

View on arXiv Download PDF AI Translation

cs.CL / 54 / 2609.00588

Quit While You're Ahead: Quit for Efficient Candidate Generation in Machine Translation Reranking

趁胜追击：在机器翻译重排序中高效候选生成的退出策略

Chen, Guangyu, Lyu, Boxuan, Kamigaito, Hidetaka, Funakoshi, Kotaro, Okumura, Manabu

Abstract

Reranking methods, such as Minimum Bayes Risk (MBR) decoding and Quality Estimation (QE) reranking, are widely used in modern neural machine translation (NMT) to select an output from a set of candidate hypotheses. However, the performance gains come at the cost of high inference latency. Existing acceleration methods target MBR decoding and reduce only reranking computation, leaving QE reranking unaddressed and candidate generation---which can be the larger computational bottleneck---largely untouched. In this work, we propose Quit (Quantifying Uncertainty for Incremental Termination), a novel early-stopping strategy for the entire generation--reranking pipeline. Viewing candidate generation as a sequential decision under uncertainty, Quit incrementally generates and reranks candidates, stopping when the highest estimated quality in the candidate set stabilizes. Comprehensive experiments on three NMT models across 19 language pairs show that Quit yields end-to-end speedups of $1.47$--$2.66\times$ for MBR and $3.43$--$4.12\times$ for QE reranking while preserving translation quality within prespecified equivalence margins.

Chinese Translation

重排序方法，如最小贝叶斯风险（Minimum Bayes Risk, MBR）解码和质量估计（Quality Estimation, QE）重排序，在现代神经机器翻译（Neural Machine Translation, NMT）中被广泛应用于从候选假设集合中选择输出。然而，性能提升的代价是高推理延迟。现有的加速方法主要针对MBR解码，仅减少重排序计算，而QE重排序则未得到解决，候选生成——这可能是更大的计算瓶颈——也基本未被触及。在本研究中，我们提出了Quit（量化不确定性以实现增量终止），这是一种针对整个生成-重排序流程的新型提前停止策略。将候选生成视为在不确定性下的顺序决策，Quit逐步生成和重排序候选，当候选集中的最高估计质量稳定时停止。对三个NMT模型在19个语言对上的全面实验表明，Quit在MBR和QE重排序中分别实现了$1.47$--$2.66 imes$和$3.43$--$4.12 imes$的端到端加速，同时在预设等效范围内保持翻译质量。

View on arXiv Download PDF AI Translation

cs.CL / 55 / 2609.00608

Investigating Assistant Bias in LLM User Simulators Using a Role Vector

使用角色向量研究大型语言模型用户模拟器中的助手偏见

Jeong, Daeheon, Lee, Yoonjoo, Choi, Eugene, van der Ben, Sinie, Kim, Juho

Abstract

LLM-based user simulators are increasingly used to evaluate autonomous agents at scale, in place of costly human evaluations. Despite this promise, these simulators exhibit "assistant bias," a tendency to cooperate and pursue task goals. They rarely reproduce the frustration or disengagement that real users exhibit, compromising evaluation validity. Prior work outlines that this bias is baked in during model training, which role-playing prompts fail to override. We analyze this bias from model activations, extracting a user role vector by contrasting how the model represents user versus assistant perspectives on the same dialogue. We observe two findings: (i) the user direction is identifiable in activations, elicits user-like behaviors, and captures characteristics distinct from assistant traits; and (ii) although user-role activation associates with simulation realism and steering strengthens it, it can exaggerate user behaviors and override individual user profiles. Together, our findings provide a representation-level analysis of LLM user simulators, confirming that assistant bias is structurally identifiable and that user behavior can be directionally analyzed.

Chinese Translation

基于大型语言模型（LLM）的用户模拟器越来越多地被用于大规模评估自主智能体，取代昂贵的人类评估。尽管这一前景令人期待，这些模拟器却表现出“助手偏见”，即倾向于合作并追求任务目标。它们很少再现真实用户所表现出的挫败感或脱离感，从而影响评估的有效性。之前的研究表明，这种偏见在模型训练过程中已经根深蒂固，而角色扮演提示无法覆盖这一点。我们通过分析模型激活来研究这种偏见，通过对比模型在同一对话中对用户与助手视角的表示，提取出用户角色向量。我们观察到两个发现：（i）用户方向在激活中是可识别的，能够引发类似用户的行为，并捕捉到与助手特征不同的特征；（ii）尽管用户角色激活与模拟的真实感相关，并且引导作用增强了这种真实感，但它可能会夸大用户行为并覆盖个体用户特征。综合来看，我们的研究提供了对LLM用户模拟器的表征级分析，确认助手偏见在结构上是可识别的，并且用户行为可以进行方向性分析。

View on arXiv Download PDF AI Translation

cs.CL / 56 / 2609.00624

Trust Your Guide Only When Certain: Uncertainty-Aware Sparse Alignment at Inference Time

仅在确定时信任你的引导者：推理时的不确定性感知稀疏对齐

Zhu, Zeen, Li, Zhuo, Guo, Weiyang, Zhao, Liye, Di, Haibing, Wang, Yequan, Li, Jing

Abstract

A prominent paradigm in inference-time alignment employs lightweight supervisors to steer Large Language Models (LLMs). Through empirical analysis, we identify a structural mismatch in this paradigm: weak supervisors exhibit pervasive high entropy across the vast majority of tokens, yet prevailing dense intervention approaches mandate supervision at every decoding step. This leads to frequent low-confidence interventions that can disrupt valid base-model reasoning and incur substantial utility costs. To resolve this, we propose TUSA (Trust-based Uncertainty Sparse Alignment). Moving away from continuous oversight, TUSA reframes alignment as a dynamic arbitration process, introducing an uncertainty-aware arbiter that authorizes intervention only when two conditions are met: the supervisor is confident and the token is semantically salient. This mechanism effectively filters out uncertainty-driven noise and redundant supervision. Extensive experiments across multiple models and benchmarks show that TUSA consistently improves both safety alignment and general helpfulness. By bypassing approximately 50% of alignment steps, it not only enhances safety preference by up to 15.6%, but also boosts general preference rates by up to 12.0% compared to the dense baseline, demonstrating that selective, high-precision alignment can outperform continuous supervision.

Chinese Translation

推理时对齐的一个显著范式采用轻量级监督者来引导大型语言模型（LLMs）。通过实证分析，我们识别出该范式中的结构性不匹配：弱监督者在绝大多数标记上表现出普遍的高熵，而现有的密集干预方法要求在每个解码步骤都进行监督。这导致频繁的低置信度干预，可能会干扰有效的基础模型推理并产生可观的效用成本。为了解决这个问题，我们提出了TUSA（基于信任的不确定性感知稀疏对齐）。TUSA摒弃了持续监督的方式，将对齐重新构建为一个动态仲裁过程，引入了一种不确定性感知的仲裁者，仅在满足两个条件时授权干预：监督者有信心且标记在语义上显著。该机制有效过滤掉由不确定性驱动的噪声和冗余监督。针对多个模型和基准的广泛实验表明，TUSA在安全对齐和一般有用性方面始终表现出改善。通过绕过约50%的对齐步骤，它不仅将安全偏好提高了多达15.6%，还将一般偏好率提高了多达12.0%，相较于密集基线，证明了选择性、高精度的对齐能够超越持续监督。

View on arXiv Download PDF AI Translation

cs.CL / 57 / 2609.00629

ExpArt-KG: Artwork Image Description Generation through Iterative Exploration of Knowledge Graphs

ExpArt-KG：通过对知识图谱的迭代探索生成艺术作品图像描述

Kato, Yuta, Ozaki, Shintaro, Hayashi, Kazuki, Sakai, Yusuke, Kamigaito, Hidetaka, Hayashi, Katsuhiko, Watanabe, Taro

Abstract

Large Vision-Language Models (LVLMs) achieve strong performance on image-grounded text generation and visual question answering. However, it remains difficult for them to comprehensively and accurately describe the factual relations among the entities and concepts associated with the objects depicted in an image. In this work, we propose a framework that efficiently exploits factual information from a knowledge graph via retrieval-augmented generation (RAG), with the goal of enabling LVLMs to generate detailed and accurate image explanations. Specifically, our method alternates between answer generation and knowledge-graph retrieval, and controls the search using a correctness judgment, thereby acquiring the necessary and sufficient factual information efficiently. We also construct a knowledge graph for the artwork domain (ExpArt-KG), in which the correspondence between images and entities is unambiguous. Applying the proposed method to this knowledge graph, we show experimentally that it improves the level of detail of artwork explanations and reduces the retrieval cost of external knowledge while maintaining generation quality comparable to that of iterating a fixed number of times.

Chinese Translation

大型视觉语言模型（LVLMs）在基于图像的文本生成和视觉问答方面表现出色。然而，它们在全面和准确地描述与图像中所描绘对象相关的实体和概念之间的事实关系方面仍然面临挑战。在本研究中，我们提出了一个框架，通过检索增强生成（RAG）有效利用知识图谱中的事实信息，旨在使LVLMs能够生成详细且准确的图像解释。具体而言，我们的方法在答案生成和知识图谱检索之间交替进行，并通过正确性判断来控制搜索，从而高效获取必要且充分的事实信息。我们还为艺术作品领域构建了一个知识图谱（ExpArt-KG），其中图像与实体之间的对应关系明确。将所提出的方法应用于该知识图谱，我们通过实验证明，它提高了艺术作品解释的细节水平，并在保持生成质量与固定迭代次数相当的情况下，降低了外部知识的检索成本。

View on arXiv Download PDF AI Translation

cs.CL / 58 / 2609.00683

Creative Generation via Multi-Agent Debate: Does Debate Suppress Diversity?

通过多智能体辩论进行创造性生成：辩论是否抑制多样性？

Nguyen, Tien Anh, Nguyen, Khanh-Binh, Do, Van Dai, Venkatesh, Svetha, Le, Hung

Abstract

Creative generation tasks, such as narrative writing and scientific ideation, demand both high-quality outputs and distinct responses across independent runs to maximize exploration. Multi-Agent Debate (MAD) has shown strong quality gains on factual and reasoning tasks, making it a natural candidate for creative generation. However, we find its convergence-driven design actively suppresses output diversity across independent runs, creating an inherent trade-off with creative tasks. We theoretically show that preserving diversity among agents within each debate session is a necessary condition for achieving diverse outputs across independent runs. Building on this finding, we propose Creative-MAD, which introduces two synergistic interventions to sustain agent divergence. Specifically, Cognitive Lens Assignment counters identity drift by anchoring each agent to a distinct and persistent cognitive mode, while Embedding-based Peer Selection counters majority pull by limiting each agent's context to its most semantically distant peers. Experiments across four creative benchmarks demonstrate that Creative-MAD significantly enhances both lexical and semantic diversity while maintaining MAD's output quality.

Chinese Translation

创造性生成任务，如叙事写作和科学构思，要求在独立运行中产生高质量的输出和不同的响应，以最大化探索。多智能体辩论（Multi-Agent Debate, MAD）在事实和推理任务上显示出显著的质量提升，使其成为创造性生成的自然候选者。然而，我们发现其以收敛为驱动的设计在独立运行中积极抑制输出多样性，从而与创造性任务形成固有的权衡。我们理论上证明，在每个辩论会话中保持智能体之间的多样性是实现独立运行中多样化输出的必要条件。基于这一发现，我们提出了Creative-MAD，旨在引入两种协同干预措施以维持智能体的差异性。具体而言，认知视角分配（Cognitive Lens Assignment）通过将每个智能体锚定到一个独特且持久的认知模式来对抗身份漂移，而基于嵌入的同行选择（Embedding-based Peer Selection）通过限制每个智能体的上下文到其最语义上遥远的同行来对抗多数影响。在四个创造性基准测试中的实验表明，Creative-MAD显著增强了词汇和语义的多样性，同时保持了MAD的输出质量。

View on arXiv Download PDF AI Translation

cs.CL / 59 / 2609.00685

Visual Framing for News Stance Detection via Image Generation

通过图像生成进行新闻立场检测的视觉框架

Lee, Dahyun, Han, Jiyoung, Park, Kunwoo

Abstract

Article-level news stance detection aims to identify the perspective of news articles toward social issues. Despite advances in stance detection and its importance for trustworthy media environments, news articles pose distinct challenges because their stances are often implicit, subtly conveyed through journalistic framing, and embedded in long, structurally complex texts. To address these challenges, we introduce VFStance, which leverages visual framing to make implicit stance cues more explicit via image generation. In evaluation experiments, we demonstrate the effectiveness of VFStance over existing methods and the contribution of visual framing to its performance. Finally, a controlled user study (N=200) in a snippet-based news consumption setting further demonstrates that VFStance can make stance signals visually salient and highlights its potential use beyond automated stance detection.

Chinese Translation

文章级新闻立场检测旨在识别新闻文章对社会问题的视角。尽管在立场检测方面取得了进展，并且其对于可信媒体环境的重要性不言而喻，但新闻文章仍然面临独特的挑战，因为它们的立场往往是隐含的，通过新闻框架微妙地传达，并嵌入在结构复杂、篇幅较长的文本中。为了解决这些挑战，我们提出了VFStance，该方法利用视觉框架通过图像生成使隐含的立场线索更加明确。在评估实验中，我们展示了VFStance相对于现有方法的有效性，以及视觉框架对其性能的贡献。最后，在基于片段的新闻消费环境中进行的受控用户研究（N=200）进一步表明，VFStance能够使立场信号在视觉上更加显著，并强调了其在自动立场检测之外的潜在应用。

View on arXiv Download PDF AI Translation

cs.CL / 60 / 2609.00689

SCoNE: Selective Context-aware Neuron Editing for Robust Retrieval-Augmented Generation

SCoNE：用于增强检索生成的选择性上下文感知神经元编辑

Kim, Chaewon, Park, Seo Yeon

Abstract

Retrieval-Augmented Generation (RAG) is highly sensitive to retrieval noise: when retrieved documents mix informative and irrelevant context, LLMs are easily distracted, leading to hallucinations. To overcome this, we propose SCoNE (Selective Context-aware Neuron Editing), a training-free model editing approach that improves retrieval noise robustness by selectively strengthening context-aware FFN neurons that are identified by both high attribution and high cross-input variability. SCoNE requires only a small number of mining samples, no fine-tuning, and no inference-time overhead. Across various knowledge-intensive question-answering benchmarks and two LLM backbones, SCoNE consistently outperforms competitive baseline methods. Our code is available at https://github.com/HYU-ARK-Lab/SCoNE.

Chinese Translation

增强检索生成（RAG）对检索噪声高度敏感：当检索到的文档混合了信息性和无关的上下文时，大型语言模型（LLMs）容易分心，导致幻觉。为了解决这个问题，我们提出了SCoNE（选择性上下文感知神经元编辑），这是一种无训练模型编辑方法，通过选择性地增强由高归因和高跨输入变异性识别的上下文感知前馈神经元，从而提高对检索噪声的鲁棒性。SCoNE只需少量挖掘样本，不需要微调，也没有推理时的开销。在各种知识密集型问答基准和两个LLM基础模型上，SCoNE始终优于竞争基线方法。我们的代码可在https://github.com/HYU-ARK-Lab/SCoNE获取。

View on arXiv Download PDF AI Translation

cs.CL / 61 / 2609.00706

A Certificate-Producing Cascade for Equational Implication: The SAIR EQT2 Stage 2 Solver

等式蕴含的证书生成级联：SAIR EQT2 第二阶段求解器

Ma, Haobo, Zhang, Wenlin, Cázares, Manuel Israel

Abstract

The SAIR Mathematics Distillation Challenge on Equational Theories asks a solver to classify whether one magma identity implies another and, for either verdict, to return a certificate accepted by a deterministic Lean judge. We present a single-file solver organized as a cheapest-first cascade. Its false branch combines coefficient tests over structured algebra families, bounded finite-model search, an explicit central-groupoid witness, and several infinite-carrier witnesses. Its true branch is a proof-producing ordered unit superposition procedure with Knuth-Bendix ordering, bidirectional demodulation, indexing, memoised substitution, and anytime size deepening. Search results remain outside the trusted base: successful derivations are replayed as small Lean terms, and countermodels are rechecked by the competition judge. The frozen solver is a 189,504-byte Python file with SHA-256 f2392533c9f4c03b.... In local runs through official judge revision 2848228, it produced accepted certificates for all 1,889 rows of the six public sets with no language-model calls. Separate measurements recorded full agreement on the 800 published Stage 1 evaluation-distribution problems, 100 accepted rows in the canonical Marathon manifest without tokens, and 200 accepted rows in the hosted playground. These are regression and playground measurements, not a leaderboard result and not evidence about a hidden set. All quantitative claims are tied to immutable result ledgers; the paper makes no completeness or comparative-superiority claim.

Chinese Translation

SAIR 数学提炼挑战关于等式理论要求求解器判断一个熔融体身份是否蕴含另一个身份，并在任一结果下返回一个被确定性 Lean 判官接受的证书。我们提出了一个以最便宜优先级组织的单文件求解器，其错误分支结合了对结构代数族的系数测试、有界有限模型搜索、显式中心群体见证以及多个无限载体见证。其正确分支是一个生成证明的有序单位叠加过程，采用 Knuth-Bendix 排序、双向去模化、索引、记忆化替换和随时深度增加。搜索结果保持在可信基础之外：成功的推导被重放为小型 Lean 术语，反例模型由竞赛判官重新检查。冻结的求解器是一个大小为 189,504 字节的 Python 文件，SHA-256 为 f2392533c9f4c03b.... 在通过官方判官修订版 2848228 的本地运行中，它为所有 1,889 行的六个公共集合生成了被接受的证书，没有调用语言模型。单独的测量记录了在 800 个已发布的第一阶段评估分布问题上的完全一致性，在没有令牌的规范马拉松清单中接受的 100 行，以及在托管游乐场中接受的 200 行。这些是回归和游乐场测量，而不是排行榜结果，也不是关于隐藏集合的证据。所有定量声明都与不可变结果账本相关；本文不作完整性或比较优越性的声明。

View on arXiv Download PDF AI Translation

cs.CL / 62 / 2609.00747

Can Large Language Models Forecast What Researchers Study Next?

大型语言模型能预测研究者接下来研究什么吗？

Li, Fenghai, Tang, Zihan, Yu, Haofei, Zhao, Yining, You, Jiaxuan

Abstract

Large language models increasingly generate research ideas, yet judging their novelty or feasibility at generation time does not establish whether they anticipate subsequent work. We introduce IdeaForecastBench to evaluate research idea forecasting. Given a community's literature up to a cutoff, a system produces up to five ranked ideas, which are evaluated against later papers. The benchmark comprises 624 rolling episodes across 52 topics, with a fixed retrieve-then-judge protocol and separately reported results from two judges. We compare five history-compression strategies across GPT-4.1, Qwen2.5-7B/14B, and Qwen3.5-9B, together with a learned Mode-Decomposition Forecaster (MDF). Under the primary GPT-4.1-mini judge, Summary improves on Direct in Hit@5 and Precision@5 across all four backbones. Qwen2.5 scores above GPT-4.1, whereas Qwen3.5 scores below it. An outcome-blind assessment finds that Qwen2.5 produces broader forecasts, but does not identify how much breadth contributes to its advantage. Threshold and judge diagnostics further clarify the limits of interpreting realization as precise anticipation. IdeaForecastBench provides a common task for studying which research ideas a community subsequently pursues and how reliably this outcome can be measured.

Chinese Translation

大型语言模型越来越多地生成研究想法，但在生成时判断其新颖性或可行性并不能确定它们是否能够预见后续的工作。我们引入了IdeaForecastBench来评估研究想法的预测能力。在截止日期之前，给定一个社区的文献，系统生成最多五个排名的想法，并与后来的论文进行评估。该基准测试包括52个主题的624个滚动实验，采用固定的检索-判断协议，并分别报告两位评审的结果。我们比较了在GPT-4.1、Qwen2.5-7B/14B和Qwen3.5-9B上五种历史压缩策略，以及一个学习的模式分解预测器（Mode-Decomposition Forecaster, MDF）。在主要的GPT-4.1-mini评审下，Summary在Hit@5和Precision@5方面优于Direct，适用于所有四个基础模型。Qwen2.5的得分高于GPT-4.1，而Qwen3.5的得分低于它。一项不考虑结果的评估发现，Qwen2.5产生了更广泛的预测，但未能确定这种广度对其优势的贡献程度。阈值和评审诊断进一步阐明了解读实现为精确预见的局限性。IdeaForecastBench为研究一个社区随后追求的研究想法及其结果的可靠性提供了一个共同任务。

View on arXiv Download PDF AI Translation

cs.CL / 63 / 2609.00748

Measuring Optimal Transport in Transformer Depth

测量变压器深度中的最优传输

Quemy, Alexandre

Abstract

A transformer carries each token's state from layer to layer, and the whole vocabulary carried together forms a cloud that moves with depth. We ask whether a trained network moves this cloud the way optimal transport would: at the cheapest cost, and along the map that pairs each token with its optimal destination. We measure both on Pythia-160m and Pythia-410m, with an exact assignment between consecutive layer clouds, a measured sampling floor, calibration on couplings known to be optimal, and a split of the cost into the common shift of the cloud and the token-specific moves. At the last layer, both models move their tokens where the optimal-transport map sends them, at the optimal cost for Pythia-410m and slightly above it for Pythia-160m. At the first layer they do not. In between, single layers can be judged on cost at only two of ten transitions, and blocks of several layers move the cloud at close to the optimal cost. The agreement at the last layer is much weaker at initialisation (0.64 against 0.86) and grows with training.

Chinese Translation

变压器将每个标记的状态从一层传递到另一层，整个词汇表共同形成一个随着深度移动的云。我们询问一个训练好的网络是否以最优传输的方式移动这个云：以最低的成本，并沿着将每个标记与其最优目标配对的映射进行移动。我们在 Pythia-160m 和 Pythia-410m 上进行测量，建立了连续层云之间的精确分配，测量了采样底线，对已知为最优的耦合进行了校准，并将成本分解为云的共同移动和特定于标记的移动。在最后一层，两个模型都将其标记移动到最优传输映射所指向的位置，对于 Pythia-410m 的成本是最优的，而 Pythia-160m 则略高于此。在第一层，它们并没有这样做。在两者之间，单层仅在十个过渡中的两个上可以根据成本进行评估，而多个层的块以接近最优的成本移动云。在最后一层的协议在初始化时要弱得多（0.64 对 0.86），并随着训练而增强。

View on arXiv Download PDF AI Translation

cs.CL / 64 / 2609.00756

Joint Training Is Not Enough: Conditioned Cross-Granularity Training for Multimodal Document Understanding

联合训练不足：针对多模态文档理解的条件交叉粒度训练

Gan, Chengguang, Liang, Yunhao, Wei, Hanjun, Zhang, Qinghao, Ni, Shiwen

Abstract

The Mutual Reinforcement Effect (MRE) asks whether a fine, span-level and a coarse, document-level task help each other when one model handles both. We test it in multimodal document understanding on three corpora, two of receipts and one of scanned business forms, comparing single-task, joint and conditioned training, which puts one granularity's gold output in the other's prompt during training only. We build Doc-MRE, an annotation layer pairing gold field extraction (point) with four document-level facets (line), from a three-judge LLM committee under a pre-registration, validated by blind re-annotation. One predicate, fixed in advance: at a shared recipe, a regime reinforces if it beats the matched single-task model on both granularities. Mixed joint training, the arrangement prior MRE work assumes, reinforces on no corpus at the main scale: it is below both single-task models on CORD and trades one granularity for the other on the two others, as single-task tuning does. Conditioned training reinforces on two of the three, CORD (+0.5 point, +4.8 line) and the forms corpus (+7.2 point, +11.0 line), resolvably on the coarse side and directionally on the fine one, and trades on WildReceipt; at that recipe no alternative measurably beats it on either side anywhere. Two byte-identical-prompt controls separate content from format: shuffled conditioning destroys the coarse-side skill but costs the fine side far less, and a neutral-content control reproduces the whole fine-side gain on WildReceipt, which is therefore prompt structure but buys nothing resolvable on the other two. On the forms corpus conditioning buys collapse avoidance: mixed training and the neutral control both assign the majority semantic label to all 50 test documents; only conditioning recovers the gold distribution. Probes find the information decodable under every regime with no resolvable increase under conditioning.

Chinese Translation

互助强化效应（Mutual Reinforcement Effect, MRE）探讨当一个模型同时处理细粒度（span-level）和粗粒度（document-level）任务时，二者是否能够相互促进。我们在三个语料库上测试了这一效应，其中两个是收据，另一个是扫描的商业表单，比较了单任务、联合训练和条件训练，后者在训练过程中将一个粒度的黄金输出放入另一个的提示中。我们构建了Doc-MRE，一个注释层，将黄金字段提取（point）与四个文档级别的方面（line）配对，基于一个三评审的LLM委员会，在预注册下进行，经过盲重注释验证。一个先验的谓词是：在一个共享的配方中，如果一个制度在两个粒度上都优于匹配的单任务模型，则它会得到强化。混合联合训练，之前的MRE研究所假设的安排，在主要规模上对任何语料库都没有强化：在CORD上低于两个单任务模型，并在其他两个语料库上交换一个粒度与另一个粒度，正如单任务调优所做的那样。条件训练在三个语料库中的两个上得到了强化，CORD（+0.5点，+4.8行）和表单语料库（+7.2点，+11.0行），在粗粒度方面明确有效，在细粒度方面方向性有效，并在WildReceipt上进行了交换；在该配方下，没有其他替代方案在任一方面的表现上可测量地超越它。两个字节相同提示的控制将内容与格式分开：打乱的条件训练破坏了粗粒度技能，但对细粒度的影响较小，而中性内容控制在WildReceipt上再现了整个细粒度的增益，因此这表明是提示结构的影响，但在其他两个上没有带来可解决的收益。在表单语料库中，条件训练避免了崩溃：混合训练和中性控制都将大多数语义标签分配给所有50个测试文档；只有条件训练恢复了黄金分布。探测发现，在每种制度下信息都是可解码的，但在条件训练下没有可解决的增加。

View on arXiv Download PDF AI Translation

cs.CL / 65 / 2609.00759

Compile, Don't Memorize: A Context Compilation Architecture (CCA) for In-Context Learning

编译，而非记忆：一种用于上下文学习的上下文编译架构（CCA）

Qi, Jinhu, Hu, Minda, Zhang, Wentao, Jin, Weiqiang, Chen, Yanyu, Wang, Junli, King, Irwin

Abstract

Large language models (LLMs) increasingly handle in-context learning (ICL) tasks where a long, novel context defines the rules, knowledge, and output schema for a series of questions. On benchmarks that grade against every detail of the context, even strong open-weights models pass only 12-16% of tasks: a single overlooked rule fails the whole response. We argue this brittleness is structural: the dominant "read-and-reason" paradigm asks the model to extract, plan, generate, and self-verify in one forward pass. We therefore ask whether explicit context compilation can fix it, how it compares to existing long-context strategies (gist retrieval, multi-agent self-play), and where the resulting harness benefit holds across task structure and model scale. We propose the Context Compilation Architecture (CCA), whose central novelty is a typed intermediate representation (IR) with fixed slots (rules.{must_do, must_not, conditional}, output_spec, available_tools, data_profile) into which any prose context is compiled once; executable verifiers and a violation-gated correction loop follow as downstream consequences. On CL-bench (1,899 tasks across 4 open base models), CCA outperforms vanilla prompting and two long-context baselines (ReadAgent-P, Ctx2Skill) on every base model, lifting Kimi K2.5 from 15.4% to 21.4% with gains concentrated on rule-dense sub-categories. Code and cached completions are available at https://github.com/TonyQJH/cca-emnlp2026.

Chinese Translation

大型语言模型（LLMs）越来越多地处理上下文学习（ICL）任务，其中一个长且新颖的上下文定义了一系列问题的规则、知识和输出模式。在对上下文每一个细节进行评分的基准测试中，即使是强大的开放权重模型也仅通过了12-16%的任务：一个被忽视的规则会导致整个响应失败。我们认为这种脆弱性是结构性的：主导的“读取与推理”范式要求模型在一次前向传递中提取、规划、生成和自我验证。因此，我们探讨显式的上下文编译是否能够解决这个问题，它与现有的长上下文策略（要点检索、多智能体自我对弈）相比如何，以及所得到的架构优势在任务结构和模型规模上持有的地方。我们提出了上下文编译架构（CCA），其核心新颖性在于一种具有固定槽位的类型化中间表示（IR）（rules.{must_do, must_not, conditional}, output_spec, available_tools, data_profile），任何散文上下文都可以一次性编译到其中；可执行的验证器和一个基于违规的修正循环作为下游结果随之而来。在CL-bench（4个开放基础模型的1,899个任务）上，CCA在每个基础模型上均优于普通提示和两个长上下文基准（ReadAgent-P, Ctx2Skill），将Kimi K2.5的性能从15.4%提升至21.4%，增益集中在规则密集的子类别上。代码和缓存的完成结果可在https://github.com/TonyQJH/cca-emnlp2026获取。

View on arXiv Download PDF AI Translation

cs.CL / 66 / 2609.00760

A Unified Mechanistic Analysis of Knowledge- and Safety-Based Refusals

知识基础与安全基础拒绝的统一机制分析

Son, Yuri, Kim, Seunghee, Kim, Hyuhng Joon, Kim, Taeuk

Abstract

Large language models (LLMs) are increasingly trained to decline queries that fall outside their knowledge (knowledge-based refusal, KR) or violate safety policies (safety-based refusal, SR). Although KR and SR result in superficially similar responses, they have largely been studied in isolation, leaving open whether they share an underlying mechanism. We address this gap with a systematic study on a new dataset of 213 contrastive quadruples that jointly probe both refusal types. We find that KR and SR are governed by overlapping yet distinguishable mechanisms. Both share a refusal direction, yet the overlap is asymmetric: SR signals transfer more strongly to KR than the reverse. Type-specific specialization emerges mainly in upper layers, with KR aligning with uncertainty- and knowledge-related representations and SR with safety- and policy-related ones. We thus characterize refusal as a commit-then-specify process: a shared initial mechanism commits to refusing, then type-specific features in later layers specify whether the grounds are epistemic or normative.

Chinese Translation

大型语言模型（LLMs）越来越多地被训练以拒绝超出其知识范围的查询（知识基础拒绝，KR）或违反安全政策的查询（安全基础拒绝，SR）。尽管KR和SR在表面上产生相似的响应，但它们在很大程度上是孤立研究的，因此尚不清楚它们是否共享一个潜在机制。我们通过对一个包含213个对比四元组的新数据集进行系统研究，填补了这一空白，该数据集共同探测这两种拒绝类型。我们发现KR和SR受重叠但可区分的机制支配。两者共享一个拒绝方向，但这种重叠是不对称的：SR信号对KR的转移比反向转移更强。类型特异性专业化主要出现在上层，KR与不确定性和知识相关的表示对齐，而SR则与安全和政策相关的表示对齐。因此，我们将拒绝特征化为一个先承诺后具体化的过程：一个共享的初始机制承诺拒绝，然后在后续层中，类型特异性特征指明理由是认识论的还是规范性的。

View on arXiv Download PDF AI Translation

cs.CL / 67 / 2609.00791

Instella-MoE Technical Report

Instella-MoE 技术报告

Liu, Jiang, Ranjan, Sudhanshu, Mishra, Prakamya, Dukler, Yonatan, Ramesh, Gowtham, Wu, Jialian, Sun, Ximeng, Xie, Wen, Hou, Chaojun, Appia, Vikram, Gu, Zhenyu, Liu, Zicheng, Barsoum, Emad

Abstract

In this work, we introduce Instella-MoE, a fully open Mixture-of-Experts (MoE) language model with 16 billion total parameters and 2.8 billion active parameters per token, trained entirely from scratch on AMD Instinct MI300X and MI325X GPUs. Instella-MoE combines a sparsely activated MoE design with architectural and system-level innovations, including Gated Multi-head Latent Attention (Gated MLA) and FarSkip-Collective connectivity, enabling efficient large-scale training and inference. The model is developed through a multi-stage pipeline comprising pre-training, mid-training, long-context extension, supervised fine-tuning with feedback-driven data curation, direct preference optimization, and reinforcement learning with Multi-Teacher On-Policy Distillation. Instella-MoE achieves an average score of 76.7 across standard pre-training benchmarks, outperforming prior fully open models including OLMo-3-7B, SmolLM3-3B, and OLMoE-1B-7B, while remaining competitive with open-weight MoE and dense baselines at comparable active-parameter scales, including Moonlight-16B-A3B and Qwen3.5-4B. After post-training, our final Think checkpoint achieves an average score of 73.2 across instruction-following, reasoning, math, coding, and chat benchmarks, outperforming both fully open and open-weight models with comparable or larger active parameter counts in our evaluation. To support transparent and reproducible research, we release the complete Instella-MoE model flow, including model weights, training configurations, data mixtures, and training code. Together, these contributions establish Instella-MoE a strong, fully open foundation for efficient, high-performing MoE models and reproducible research.

Chinese Translation

在本研究中，我们介绍了 Instella-MoE，这是一种完全开放的混合专家（Mixture-of-Experts, MoE）语言模型，具有 160 亿个总参数和每个标记 28 亿个活跃参数，完全从头开始在 AMD Instinct MI300X 和 MI325X GPU 上进行训练。Instella-MoE 将稀疏激活的 MoE 设计与架构和系统级创新相结合，包括门控多头潜在注意力（Gated Multi-head Latent Attention, Gated MLA）和远跳集体连接（FarSkip-Collective），实现了高效的大规模训练和推理。该模型通过一个多阶段管道开发，包括预训练、中期训练、长上下文扩展、基于反馈的数据整理的监督微调、直接偏好优化和多教师在线策略蒸馏的强化学习。Instella-MoE 在标准预训练基准测试中平均得分为 76.7，超越了之前的完全开放模型，包括 OLMo-3-7B、SmolLM3-3B 和 OLMoE-1B-7B，同时在可比的活跃参数规模下，与开放权重的 MoE 和稠密基线（如 Moonlight-16B-A3B 和 Qwen3.5-4B）保持竞争力。在后训练后，我们的最终 Think 检查点在遵循指令、推理、数学、编码和聊天基准测试中平均得分为 73.2，超越了在我们评估中具有可比或更大活跃参数数量的完全开放和开放权重模型。为了支持透明和可重复的研究，我们发布了完整的 Instella-MoE 模型流程，包括模型权重、训练配置、数据混合和训练代码。这些贡献共同确立了 Instella-MoE 作为高效、高性能 MoE 模型和可重复研究的强大、完全开放的基础。

View on arXiv Download PDF AI Translation

cs.CL / 68 / 2609.00796

SFAD: Speculative Factuality-Aware Decoding

SFAD：考虑推测性事实的解码

Chen, Guanqiao, Wang, Di, Hu, Lijie

Abstract

As one of the most critical challenges in large language models, contextual faithfulness directly determines their reliability in knowledge-intensive applications. This task is particularly challenging as it requires balancing factual consistency with generation efficiency. Contrastive decoding methods require dual forward passes (with and without context) to compare model outputs, doubling inference computational overhead, while post-training alignment demands extensive reinforcement learning with substantial computational overhead. To address this challenge, we present \textbf{SFAD}, a speculative decoding framework that enhances contextual faithfulness without inference degradation. We first construct \textbf{ConFide}, a preference dataset with fine-grained atomic perturbations, to train a context-faithful draft model via Direct Preference Optimization. During inference, Epistemic Friction detects potential hallucinations by quantifying distributional tension weighted by specialist certainty. When friction exceeds the threshold, Asymmetric Logit Steering refines the target distribution through residual-based logit injection; otherwise, standard speculation proceeds. Extensive experiments demonstrate that SFAD substantially improves faithfulness while achieving $2.48\times$ speedup, offering a practical solution for efficient LLMs.

Chinese Translation

在大型语言模型中，语境忠实性是最关键的挑战之一，直接决定了它们在知识密集型应用中的可靠性。这个任务尤其具有挑战性，因为它需要在事实一致性和生成效率之间取得平衡。对比解码方法需要进行双重前向传递（有上下文和无上下文）以比较模型输出，从而使推理计算开销翻倍，而后训练对齐则需要大量的强化学习，伴随相当大的计算开销。为了解决这一挑战，我们提出了 extbf{SFAD}，一种推测性解码框架，能够在不降低推理性能的情况下增强语境忠实性。我们首先构建了 extbf{ConFide}，一个具有细粒度原子扰动的偏好数据集，通过直接偏好优化训练一个语境忠实的草稿模型。在推理过程中，知识摩擦通过量化加权专家确定性的分布张力来检测潜在的幻觉。当摩擦超过阈值时，非对称逻辑引导通过基于残差的逻辑注入来细化目标分布；否则，标准推测继续进行。大量实验表明，SFAD显著提高了忠实性，同时实现了$2.48 imes$的加速，为高效的LLM提供了实用的解决方案。

View on arXiv Download PDF AI Translation

cs.CL / 69 / 2609.00802

TEIDAN: A Multilingual Multiparty Dialogue Corpus

TEIDAN：多语言多方对话语料库

Mori, Taiga, Inoue, Koji, Elmers, Mikey, Lala, Divesh, Kawahara, Tatsuya

Abstract

Multi-party interaction is a central setting for human communication and a necessary target for human-agent interaction systems that must participate in group conversation. Yet available corpora often focus on meetings, task-oriented interaction, text-based interaction, or acted scenarios, and fewer resources support cross-linguistic comparison of spontaneous face-to-face triadic discussion. This paper presents TEIDAN, a multilingual multimodal corpus that currently consists of Japanese and English three-party conversations. TEIDAN records groups of three participants discussing open-ended topics with individual pin microphones, a microphone array, and participant-facing cameras, and provides IPU-based transcripts for both language portions. Earlier studies used subsets of the Japanese portion for task-specific benchmarks in multi-party dialogue modeling; in contrast, this paper presents TEIDAN as a corpus resource spanning both Japanese and English, with planned expansion to additional languages. We describe the collection design, participants, recording setup, transcription format, and corpus statistics, and provide preliminary analyses to illustrate how TEIDAN can support research on turn-taking, addressee recognition, and multimodal grounding in human-human and human-agent interaction.

Chinese Translation

多方互动是人类沟通的核心场景，也是人机互动系统必须参与群体对话的必要目标。然而，现有的语料库往往集中于会议、任务导向的互动、基于文本的互动或表演场景，支持自发面对面三方讨论的跨语言比较的资源则较少。本文介绍了TEIDAN，一个目前由日语和英语三方对话组成的多语言多模态语料库。TEIDAN记录了三名参与者使用个人领夹麦克风、麦克风阵列和面向参与者的摄像机讨论开放性话题的过程，并为两种语言部分提供基于话语单位（IPU）的转录文本。早期研究使用了日语部分的子集进行多方对话建模的任务特定基准测试；相比之下，本文将TEIDAN作为一个跨越日语和英语的语料库资源，并计划扩展至其他语言。我们描述了数据收集设计、参与者、录音设置、转录格式和语料库统计信息，并提供初步分析，以说明TEIDAN如何支持关于轮次转换、称呼识别以及人类与人类和人类与机器互动中的多模态基础的研究。

View on arXiv Download PDF AI Translation

cs.CL / 70 / 2609.00832

TWIX: a Two-Stage Approach for End-To-End Named Entity Recognition and Relation Extraction

TWIX：一种用于端到端命名实体识别和关系提取的两阶段方法

Martinelli, Marco, Menotti, Laura

Abstract

The exponential growth of scientific publications calls for automatic Information Extraction (IE) systems to support knowledge discovery. In this context, the GutBrainIE benchmark evaluates Named Entity Recognition (NER), Named Entity Recognition and Disambiguation (NERD), and Relation Extraction (RE) systems in the gut-brain axis domain. We propose Two-stage Workflow for Information eXtraction (TWIX), an end-to-end IE pipeline featuring three interconnected modules, each leveraging a two-stage framework to solve all four GutBrainIE subtasks. Evaluation on the development and test sets shows that our method substantially outperforms the baseline by a wide margin, while also ranking first among all participant submissions across all subtasks. These results indicate that the proposed two-stage pipeline effectively improves both precision and recall in practical settings.

Chinese Translation

科学出版物的指数增长呼唤自动信息提取（IE）系统以支持知识发现。在此背景下，GutBrainIE基准评估了肠脑轴领域的命名实体识别（NER）、命名实体识别与消歧（NERD）以及关系提取（RE）系统。我们提出了信息提取的两阶段工作流程（TWIX），这是一个端到端的IE管道，具有三个相互连接的模块，每个模块利用两阶段框架解决所有四个GutBrainIE子任务。在开发集和测试集上的评估显示，我们的方法在很大程度上超越了基线，同时在所有子任务中在所有参与者提交中排名第一。这些结果表明，所提出的两阶段管道在实际应用中有效提高了精度和召回率。

View on arXiv Download PDF AI Translation

cs.CL / 71 / 2609.00833

Dense Process Supervision for Search Agents via Fact Utility Estimation

通过事实效用估计对搜索代理进行密集过程监督

Zhu, Rongzhi, Liu, Xiangyu, Liu, Yi, Zhang, Shuo, Zhang, Ruirui, Wu, Rui, Jiang, Tao, Sun, Zequn, Xu, Wenhao, Hu, Wei

Abstract

Reinforcement learning (RL) for search agents typically relies on outcome rewards. However, it often fails to achieve effective credit assignment, due to the unclear value of intermediate steps. It is hard to separate their contributions from the final result. In this paper, we propose a dense process supervision method based on fact utility estimation, which models the reasoning process as the accumulation of discrete evidence facts. We first extract structured facts from raw observations and organize them into an explicit fact store. To support credit assignment, we then cluster semantically equivalent facts and infer the posterior utility of each fact cluster using Bayesian estimation over group rollouts. Finally, we convert the estimated fact utilities into dense step-level rewards to guide RL training. Experiments on seven single-hop and multi-hop QA benchmarks show that our method consistently outperforms existing baselines. Ablation studies validate clear relative improvements on multi-hop QA compared to outcome reward-only training.

Chinese Translation

搜索代理的强化学习（RL）通常依赖于结果奖励。然而，由于中间步骤的价值不明确，它往往无法有效地进行信用分配，难以将其贡献与最终结果分离。在本文中，我们提出了一种基于事实效用估计的密集过程监督方法，将推理过程建模为离散证据事实的积累。我们首先从原始观察中提取结构化事实，并将其组织成一个明确的事实存储库。为了支持信用分配，我们随后对语义上等价的事实进行聚类，并使用贝叶斯估计在组回滚中推断每个事实聚类的后验效用。最后，我们将估计的事实效用转换为密集的步骤级奖励，以指导RL训练。在七个单跳和多跳问答基准上的实验表明，我们的方法始终优于现有基线。消融研究验证了与仅基于结果奖励的训练相比，在多跳问答上有明显的相对改善。

View on arXiv Download PDF AI Translation

cs.CL / 72 / 2609.00844

Staged Linguistic Seeding: Grounded Query Expansion for Verified-Unit QA in AI Contact Centers

分阶段语言种子：面向人工智能联络中心验证单元问答的基础查询扩展

Yoon, Hyeonseop, Park, Jeong-Eun

Abstract

Customer-service QA in an AI contact center (AICC) runs under deployment constraints that benchmark QA misses: tight voice-hotline latency and a high cost for unsupported or wrong automatic answers. We deploy a system that answers only from a closed set of verified QA units: it returns a retrieved unit verbatim, or routes to clarify, abstain, or handoff. The index is enriched offline by staged linguistic seeding (SLS): a human authors a per-unit world-grounded slot recipe, gpt-4.1-mini renders it into variants, and a light human gate filters them. One methodology is reused across both domains, so inference stays a single retrieval pass with no query-time generation. On held-out query variants from two industrial domains, SLS lifts hybrid R@1 to 0.881/0.930 (+0.27/+0.34), with gains across all five retrievers tested. At the same gpt-4.1-mini generation budget, SLS beats doc2query by +0.20/+0.32, while cross-provenance evaluation provides additional evidence of transfer across generated-query distributions. Verified-unit answering also removes free-form generation's unsupported-content surface (7-13% versus approximately 0%). We report this as an application study, including negative results.

Chinese Translation

人工智能联络中心（AICC）中的客户服务问答（QA）在部署约束下运行，这些约束会影响QA的遗漏：紧迫的语音热线延迟和对不支持或错误自动回答的高成本。我们部署了一个仅从一组经过验证的QA单元中回答的问题的系统：它要么逐字返回检索到的单元，要么进行澄清、放弃或转交。该索引通过分阶段语言种子（SLS）离线增强：由人类撰写每个单元的世界基础插槽配方，gpt-4.1-mini将其转换为多种变体，并由轻量级人类网关进行筛选。一种方法在两个领域中重复使用，因此推理保持为单次检索过程，无需查询时生成。在来自两个工业领域的保留查询变体上，SLS将混合R@1提升至0.881/0.930（+0.27/+0.34），在测试的所有五个检索器中均有增益。在相同的gpt-4.1-mini生成预算下，SLS比doc2query提高了+0.20/+0.32，而跨来源评估提供了生成查询分布之间转移的额外证据。验证单元回答还消除了自由形式生成中不支持内容的表面（7-13%对比约0%）。我们将此报告作为应用研究，包括负面结果。

View on arXiv Download PDF AI Translation

cs.CL / 73 / 2609.00873

Membership Inference in Fine-tuned Diffusion Language Models via Token-level Memorization Asymmetry

通过令牌级记忆不对称性进行微调扩散语言模型的成员推断

Zhai, Shengfang, Marchyok, Leo, Shi, Yuling, Chen, Huanran, Dong, Yinpeng, Zhang, Jiaheng, Hong, Sanghyun

Abstract

Diffusion language models (DLMs) have recently emerged as an alternative modeling paradigm to autoregressive LMs, offering advantages such as parallel generation and bidirectional context modeling. Despite growing interest in their generative capabilities, the privacy risks of DLMs remain underexplored. We identify a phenomenon termed token-level memorization asymmetry through theoretical analysis of diffusion training dynamics. Building on this finding, we propose Q-Skew, a quantile-weighted skewness-based indicator for membership inference on finetuned DLMs. Experiments across multiple fine-tuning datasets and models show that our method outperforms existing baselines. Moreover, we show that Q-Skew can also facilitate other privacy violations, such as PII extraction. Our findings reveal a previously underexplored privacy attack surface and highlight the need for systematic privacy evaluation of DLMs.

Chinese Translation

扩散语言模型（DLMs）最近作为自回归语言模型（LMs）的替代建模范式出现，提供了并行生成和双向上下文建模等优势。尽管对其生成能力的兴趣日益增长，但DLMs的隐私风险仍然未得到充分探讨。我们通过对扩散训练动态的理论分析，识别出一种现象，称为令牌级记忆不对称性。在此基础上，我们提出了Q-Skew，这是一种基于分位数加权偏度的指标，用于对微调DLMs进行成员推断。在多个微调数据集和模型上的实验表明，我们的方法优于现有基准。此外，我们还展示了Q-Skew可以促进其他隐私侵犯，例如个人身份信息（PII）提取。我们的研究揭示了一个先前未被充分探讨的隐私攻击面，并强调了对DLMs进行系统性隐私评估的必要性。

View on arXiv Download PDF AI Translation

cs.CL / 74 / 2609.00925

Context-Grounding Gains Are Mediated by Pre-existing Machinery: Auditing GRPO, SFT, and DPO

上下文基础收益由预先存在的机制介导：审计 GRPO、SFT 和 DPO

Gupta, Prakhar, Gupta, Vaibhav

Abstract

Language models can ignore prompt evidence when it conflicts with memorized knowledge. Post-training can make models follow such evidence more reliably, but it is unclear whether these gains require new machinery or strengthen machinery already present. We compare nine post-training arms spanning GRPO, SFT, and DPO from one starting checkpoint, with key comparisons extended across scales and families. We estimate a grounding direction from that checkpoint before training. Across five tested GRPO variants, grounding gains are small. For the two variants replicated across seeds, equivalence tests bound their effects below the conflict-SFT gain even as the rewarded metric improves. Conflict-SFT improves grounding moderately, while DPO drives grounding near ceiling on its matched distribution. Conflict-SFT and DPO largely use the same causal attention-head set as the starting model. Subtracting the starting-model direction suppresses both gains, while adding it to the starting model recovers 35% of DPO's gain at a dose passing all stated side-effect checks. After a supervised warm start makes the context answer appear in more rollouts, the same GRPO recipe adds essentially no further grounding gain. In our setting, grounding gains largely depend on machinery already present in the starting model.

Chinese Translation

语言模型在提示证据与记忆知识冲突时可能会忽略这些证据。后训练可以使模型更可靠地遵循这些证据，但尚不清楚这些收益是否需要新的机制或是增强了已存在的机制。我们比较了从一个起始检查点出发的九个后训练方法，包括 GRPO、SFT 和 DPO，并在不同规模和家族之间进行了关键比较。我们在训练前从该检查点估计了一个基础方向。在五个测试的 GRPO 变体中，基础收益较小。对于在不同种子下复制的两个变体，等效性测试将其效果限制在冲突-SFT 收益之下，即使被奖励的指标有所改善。冲突-SFT 中等程度地改善了基础，而 DPO 则在其匹配分布上将基础提升至接近上限。冲突-SFT 和 DPO 在很大程度上使用与起始模型相同的因果注意力头集合。减去起始模型方向会抑制这两种收益，而将其添加到起始模型中则恢复了 DPO 收益的 35%，且通过了所有声明的副作用检查。在经过监督的热启动使上下文答案在更多的回滚中出现后，相同的 GRPO 配方几乎没有增加进一步的基础收益。在我们的设置中，基础收益在很大程度上依赖于起始模型中已存在的机制。

View on arXiv Download PDF AI Translation

cs.CL / 75 / 2609.00935

DualStake: Dual-Path Confidence Calibration in Deep Research Agents

DualStake：深度研究代理中的双路径置信度校准

Xu, Yinuo, Liang, Yuwei, Cheng, Jianjie, Wang, Meng, Yu, Yongcan, Lu, Shuo, Liang, Jian

Abstract

Deep Research agents tackle knowledge-intensive tasks through multi-round retrieval and decision-oriented generation. However, these agents suffer from severe overconfidence, making their expressed confidence unreliable for user trust and downstream abstention. To address this, we augment the Deep Research pipeline with step confidence elicitation after each retrieval, building on the commonly used post-answer verbalized confidence. Interestingly, we find that Evidence Confidence (E-Conf), elicited after the final retrieval step, provides a stronger uncertainty signal than Answer Confidence (A-Conf), elicited after answer generation, and that A-Conf is largely shaped by E-Conf. Based on these findings, we propose DualStake, a dual-path calibration method that applies margin-clipped, confidence-dependent stake rewards to jointly align E-Conf and A-Conf with answer correctness while limiting extreme confidence optimization. Experiments on Qwen2.5-7B, Qwen2.5-7B-Instruct, and Qwen3-4B across 8 QA benchmarks demonstrate that DualStake consistently improves calibration without sacrificing answer accuracy. The code is available at https://github.com/FloXXXt/DualStake.

Chinese Translation

深度研究代理通过多轮检索和决策导向生成来处理知识密集型任务。然而，这些代理存在严重的过度自信，使得其表达的置信度对用户信任和后续决策的可靠性降低。为了解决这一问题，我们在每次检索后增强深度研究流程，进行步骤置信度引导，基于常用的后答案口头化置信度。值得注意的是，我们发现最终检索步骤后引导的证据置信度（Evidence Confidence, E-Conf）提供了比答案置信度（Answer Confidence, A-Conf）更强的不确定性信号，而A-Conf在很大程度上受E-Conf的影响。基于这些发现，我们提出了DualStake，一种双路径校准方法，通过应用边际裁剪的、依赖于置信度的赌注奖励，联合对齐E-Conf和A-Conf与答案正确性，同时限制极端置信度的优化。在8个问答基准上对Qwen2.5-7B、Qwen2.5-7B-Instruct和Qwen3-4B的实验表明，DualStake在不牺牲答案准确性的情况下，持续改善了校准效果。代码可在https://github.com/FloXXXt/DualStake获取。

View on arXiv Download PDF AI Translation

cs.CL / 76 / 2609.00940

A Dataset for Modeling Iterative Problem-Solving

用于建模迭代问题解决的 dataset

Patel, Fagun, Truong, Sang T., Nguyen, Duc Q., Fukuhara, Kazunori, Domingue, Benjamin W., Koyejo, Sanmi, Haber, Nick

Abstract

Solving problems through repeated attempts is a sequential modeling task: at each step, the solver receives feedback and decides how to revise their solutions. Predicting whether performance improves, plateaus, or regresses across attempts is central to understanding any iterative problem-solving process in both human learners and autonomous agents. Beyond outcomes, modeling what errors persist and how strategies shift across attempts provides deeper insight into the mechanics of sequential learning. Studying these dynamics requires observing many solvers as they attempt, receive feedback, and revise. Programming courses with automated grading provide this setting, as students iteratively submit code to test suites and receive feedback on every attempt. We therefore curate CodeInsight, a large-scale dataset of over 3 million submissions from 3,286 undergraduates across 2 introductory C++ courses in 2 academic years, with test-case-level outcomes, timestamps, and source code. On this dataset, we build a benchmark that evaluates models spanning parametric, sequential, and generative traditions under a shared calibration-and-scoring protocol, including a Recurrent State Space Model (RSSM) adapted to track solver characteristics through discrete latent variables and an LLM-based predictor that generates explicit solutions. The adapted RSSM achieves the strongest predictive accuracy on three of the four courses. The LLM predictor is less accurate but produces full submissions at each attempt, enabling direct analysis of failure modes. We find that the model's coding proficiency is inversely related to predictive performance in this setting, with the LLM better understood as a generative solver conditioned on context rather than a faithful predictor of solver behavior. We publicly release our code and the dataset on request to facilitate future research.

Chinese Translation

通过反复尝试解决问题是一项顺序建模任务：在每一步中，解决者接收反馈并决定如何修正他们的解决方案。预测在尝试中表现是改善、持平还是退步，对于理解人类学习者和自主智能体的任何迭代问题解决过程至关重要。除了结果之外，建模哪些错误持续存在以及策略如何在尝试中变化，能够更深入地洞察顺序学习的机制。研究这些动态需要观察许多解决者在尝试、接收反馈和修正的过程中。自动评分的编程课程提供了这种环境，因为学生迭代地提交代码到测试套件，并在每次尝试中收到反馈。因此，我们整理了 CodeInsight，这是一个大规模的数据集，包含来自两个学年中 3,286 名本科生在两个入门 C++ 课程中的超过 300 万次提交，提供测试用例级别的结果、时间戳和源代码。在这个数据集上，我们建立了一个基准，评估跨越参数化、顺序和生成传统的模型，采用共享的校准和评分协议，包括一个适应于通过离散潜变量跟踪解决者特征的递归状态空间模型（Recurrent State Space Model, RSSM）和一个基于 LLM 的预测器，生成明确的解决方案。经过调整的 RSSM 在四门课程中的三门课程上实现了最强的预测准确性。LLM 预测器的准确性较低，但在每次尝试中生成完整的提交，便于直接分析失败模式。我们发现，在这种环境中，模型的编码能力与预测性能呈反比关系，LLM 更应被理解为一种基于上下文的生成解决者，而不是解决者行为的忠实预测器。我们应请求公开发布我们的代码和数据集，以促进未来的研究。

View on arXiv Download PDF AI Translation

cs.CL / 77 / 2609.00949

Calibration is the Bottleneck: An Action-Class Diagnostic of Multi-Turn Tool-Calling

校准是瓶颈：多轮工具调用的动作类别诊断

Zhao, Kangjia, Li, Jiajun, Shen, Haozhan, Chow, Wei, Li, Linfeng, Song, Hang, Kong, Lingdong, Zhi, Chen, Zhao, Tiancheng, Liu, Songhua, Yin, Jianwei

Abstract

Multi-turn tool calling is a core evaluation scenario for large language model (LLM) agents. On public tool-calling benchmarks, open-weight models now approach or even surpass closed-source frontier models in aggregate accuracy. However, this metric averages over many different multi-turn situations and obscures whether progress is balanced across them. We propose an action-class-oriented diagnostic framework that decomposes multi-turn failures into two orthogonal modes: action-class miscalibration and action-execution failure. The framework operates over a four-class action space (TOOL_CALL/ASK/REFUSE/CONFIRM) and introduces a self-revealing upper bound Acc <= GAR (Gold Action Recall); the two modes show up as bound violation (Acc > GAR, exposing state-grader masking of miscalibration) and large bound slack (GAR >> Acc, localizing execution failure within TOOL_CALL). We validate it on a panel of tool-calling models across multiple multi-turn benchmarks. Across our panel, the diagnostic reveals action-class miscalibration as a substantial failure mode the state grader cannot see. This gap inflates standing for heavily tool-trained families, which our diagnostic separates from families with context-appropriate action choice. Calibration is reshapable through context-only perturbations, but the reshape is heterogeneous: a single perturbation moves accuracy in opposite directions across families (up to +11.5 vs -21.0 pp on the same scenario), and its effect further depends on the perturbation mechanism. We argue that multi-turn tool-calling evaluations should supplement aggregate accuracy with action-class diagnostics that expose what the model actually does in each scenario.

Chinese Translation

多轮工具调用是大型语言模型（LLM）代理的核心评估场景。在公共工具调用基准测试中，开放权重模型在整体准确性上已接近甚至超过封闭源前沿模型。然而，这一指标对许多不同的多轮情况进行了平均，掩盖了进展是否在各个情况之间保持平衡。我们提出了一种面向动作类别的诊断框架，将多轮失败分解为两种正交模式：动作类别误校准和动作执行失败。该框架在一个四类动作空间（TOOL_CALL/ASK/REFUSE/CONFIRM）上运行，并引入了一个自揭示的上限 Acc <= GAR（黄金动作召回率）；这两种模式表现为上限违反（Acc > GAR，暴露出状态评估者对误校准的掩盖）和较大的上限松弛（GAR >> Acc，将执行失败局限于 TOOL_CALL）。我们在多个多轮基准测试中对一组工具调用模型进行了验证。在我们的面板中，诊断揭示了动作类别误校准作为一种显著的失败模式，而状态评估者无法察觉。这一差距使得重度工具训练的模型家族的表现膨胀，而我们的诊断将其与具有上下文适当动作选择的家族区分开来。校准可以通过仅上下文的扰动进行重塑，但这种重塑是异质的：单一扰动在不同家族中导致准确性朝相反方向变化（在同一场景下，最高可达 +11.5 与 -21.0 个百分点），其效果还进一步依赖于扰动机制。我们认为，多轮工具调用评估应该用动作类别诊断补充整体准确性，以揭示模型在每个场景中实际执行的内容。

View on arXiv Download PDF AI Translation

cs.CL / 78 / 2609.00958

PersianAnonymizer: Evaluating LLM-Labeled Training for Efficient NER-based Anonymization in Persian

PersianAnonymizer：评估基于LLM标注训练的高效波斯语命名实体识别（NER）匿名化

Shalchian, Mohammad Hossein, Amiri, Mostafa, Sadeghzadeh, Amir Mahdi

Abstract

We target practical anonymization of Persian customer chats by training a compact NER model from LLM-labeled supervision and selecting the best labeler for deployment. We compare three instruction-tuned LLMs: DeepSeek-V3-0324, GPT-OSS-120B, and Qwen3-235B-A22B-Instruct-2507, to produce span annotations under a shared JSON protocol, yielding four corpora (OSS_ZeroShot, Qwen_ZeroShot, Qwen_FewShot, DeepSeek_FewShot). A MatinaRoberta-based token-classifier is trained per corpus and evaluated with token-level Precision/Recall/F1 (overall and per-class). We also report Label Coverage Recall (LCR), the proportion of gold non-O tokens predicted as non-O, and quantify cross-labeler behavior via a token-level Venn on test annotations. Finally, we contrast test-set annotation latency of the LLMs on H200 nodes with the trained NER's test-time labeling on a single RTX 3090. Results show that supervision from OSS_ZeroShot yields the strongest macro-F1 and LCR, while the resulting NER labels an entire 40K-message test set in approximately 2 minutes on one consumer GPU. This establishes a practical path to high-quality, low-cost anonymization for Persian industrial data.

Chinese Translation

我们旨在通过从LLM标注监督中训练一个紧凑的NER模型，并选择最佳标注器进行部署，以实现波斯语客户聊天的实际匿名化。我们比较了三种经过指令调优的LLM：DeepSeek-V3-0324、GPT-OSS-120B和Qwen3-235B-A22B-Instruct-2507，以在共享的JSON协议下生成跨度注释，产生四个语料库（OSS_ZeroShot、Qwen_ZeroShot、Qwen_FewShot、DeepSeek_FewShot）。每个语料库训练一个基于MatinaRoberta的标记分类器，并使用标记级别的精确度/召回率/F1（总体和每类）进行评估。我们还报告了标签覆盖召回率（LCR），即被预测为非O的金标准非O标记的比例，并通过测试注释的标记级别维恩图量化跨标注器行为。最后，我们对比了在H200节点上LLM的测试集注释延迟与在单个RTX 3090上训练的NER的测试时间标注。结果表明，来自OSS_ZeroShot的监督产生了最强的宏F1和LCR，而最终的NER在大约2分钟内在一台消费级GPU上标注了整个40K消息的测试集。这为波斯工业数据的高质量、低成本匿名化建立了一条实际路径。

View on arXiv Download PDF AI Translation

cs.CL / 79 / 2609.00982

Disclosure-Gated User Simulation for Companion-Agent Evaluation

披露门控用户模拟用于伴随代理评估

Liu, Yao, He, Yu

Abstract

Using a large language model to play the user is now standard in scalable evaluation. It has a repeatedly diagnosed failure: the simulated user is excessively cooperative, so a system under test can score by the sheer number of questions it asks rather than by making the user willing to speak. We answer with a disclosure gate conditioning information release on the companion agent's behaviour: its state is a ladder of five ordered gates, merged onto three observable depth layers. We specify, ablate, and audit it, and train a user simulator against that specification. Gating behaviour is learned from the training corpus's synthetic branch, while the real branch supplies how people speak and react; after training, the simulator need not be told at runtime which gate each item sits behind. The gate is a load-bearing component of the environment: on the English corpus of a published companion-agent benchmark (CompanionBench), once training no longer states per example which gate each item sits behind, the largest rank displacement across 12 systems under test exceeds the noise band set by re-running that environment under a new seed, while per-system scores show no detectable change. We state two acceptance criteria: a ranking must be order-preserving, and absolute scores must be scale-stable. Of the candidates we examine, only one passes both -- the simulator we release -- and its leaderboard correlates at 0.993 with the benchmark's original simulator. By contrast, prompting a frontier model as the simulator barely moves the ranking while shifting every score upward -- a shift invisible to anyone checking the ranking alone. The environment we specify is the one that benchmark already used. That publication describes the mechanism in about four hundred words, and we supply what it lacked: specification, ablations, human studies, negative controls, and downstream sensitivity analysis.

Chinese Translation

使用大型语言模型进行用户模拟已成为可扩展评估的标准。然而，这种方法存在一个反复被诊断出的缺陷：模拟用户过于合作，因此被测试系统的得分往往是通过其提问数量的多少，而非用户愿意交流的程度。我们提出了一种披露门控机制，以调节伴随代理行为的信息释放：其状态由五个有序门控组成，合并为三个可观察的深度层次。我们对其进行了规范、消融和审计，并根据该规范训练了一个用户模拟器。门控行为是从训练语料库的合成分支中学习的，而真实分支则提供了人们的说话和反应方式；训练后，模拟器在运行时无需被告知每个项目背后对应的门控。门控是环境的一个承载组件：在已发布的伴随代理基准（CompanionBench）的英语语料库中，一旦训练不再为每个示例指明每个项目背后对应的门控，12个被测试系统之间的最大排名位移超过了在新种子下重新运行该环境所设定的噪声带，而每个系统的得分则没有检测到变化。我们提出了两个接受标准：排名必须保持顺序不变，绝对得分必须稳定。我们检查的候选者中，只有一个同时满足这两个标准——我们发布的模拟器——其排行榜与基准的原始模拟器的相关性达到0.993。相比之下，将前沿模型作为模拟器进行提示几乎没有改变排名，但却将每个得分向上移动——这种变化对于仅检查排名的人来说是不可见的。我们指定的环境正是基准已使用的环境。该出版物用大约四百个词描述了机制，而我们提供了其所缺乏的内容：规范、消融、人类研究、负对照和下游敏感性分析。

View on arXiv Download PDF AI Translation

cs.CL / 80 / 2609.00998

Inspicio: Open-Vocabulary, LLM-Based Sense Retrieval for Historical Languages

Inspicio：基于大型语言模型的开放词汇历史语言语义检索

Ciletti, Michele

Abstract

Word Sense Disambiguation has advanced rapidly for English and a handful of well-resourced modern languages, but it continues to assume the existence of a sense inventory and a word-to-sense mapping in the source language (Navigli, 2026). These assumptions break down for most historical and low-resource languages, whose dedicated WordNets are either incomplete or still under construction. We present Inspicio, an open-vocabulary retrieval pipeline that links tokens in context to synsets of the Open English WordNet (McCrae et al., 2020) without requiring any source-language inventory or mapping. For each occurrence, an instruction-tuned LLM produces two English translations of the surrounding sentence, a small set of candidate dictionary-style definitions, and a few candidate English lemmas. These outputs drive a hybrid retrieval step that combines dense definition-synset similarity, sparse lemma matching, and Maximal Marginal Relevance re-ranking. We evaluate the pipeline across a 6x6 grid of LLMs and sentence-embedding models on a new bilingual set of manually annotated Latin and Ancient Greek perception verbs, on a subset of PREMOVE dataset (Farina, 2025), and on a diachronic sample of Italian. The best configuration reaches 96% Recall@50 on the perception-verb test set, with each component contributing measurable gains, and remains competitive in the out-of-domain and cross-lingual settings.

Chinese Translation

词义消歧在英语和少数资源丰富的现代语言中迅速发展，但它仍然假设源语言中存在词义库存和词与义的映射（Navigli, 2026）。这些假设在大多数历史语言和低资源语言中失效，因为它们的专用WordNet要么不完整，要么仍在建设中。我们提出了Inspicio，一个开放词汇检索管道，将上下文中的词元链接到开放英语WordNet的同义词集（McCrae et al., 2020），而无需任何源语言的库存或映射。对于每个出现，经过指令调优的大型语言模型生成周围句子的两个英语翻译、一小组候选字典式定义和一些候选英语词元。这些输出驱动一个混合检索步骤，结合了密集的定义-同义词集相似性、稀疏的词元匹配和最大边际相关性重新排序。我们在一个新的双语手动标注的拉丁语和古希腊语感知动词集上，以及PREMOVE数据集的一个子集（Farina, 2025）和意大利语的历时样本上，评估了该管道在6x6的LLM和句子嵌入模型网格中的表现。最佳配置在感知动词测试集上达到了96%的Recall@50，每个组件都贡献了可测量的增益，并在跨领域和跨语言设置中保持竞争力。

View on arXiv Download PDF AI Translation

cs.CL / 81 / 2609.00999

Right Frame, Wrong Rule: Cultural Cues Expose the Financial Knowledge Gap They Were Meant to Close

正确的框架，错误的规则：文化线索揭示了它们旨在弥补的金融知识差距

Elbadry, Rania, Heakl, Ahmed, Almheiri, Saeed, Zhang, Fan, AlMahri, Muhra, Peng, Xueqing, Kabir, Mohsinul, Wang, Shuyao, Han, Yi, Eletter, Saadeldine, Zhang, Duzhen, Nakov, Preslav, Wang, Yuxia, Koto, Fajri, Xie, Zhuohan

Abstract

When a question has valid answers under different normative frameworks, a language model must decide which framework to use and whether it can answer correctly within it. We call this setting normative pluralism and study it in Islamic finance using a four-choice taxonomy that separates framework selection from within-framework correctness. This separation reveals the stereotype trap: a cultural cue steers a model toward one framework, but the model selects an incorrect answer within that framework. Across twelve models, two languages, and fifty demographic signals, cultural cues change framework selection and reveal substantial differences in accuracy, especially among non-frontier models. Under the strongest signal, large open-weight models select the Islamic framework 97% of the time. A two-choice evaluation would report near-perfect alignment, although 57--66% of those selections are incorrect. These findings motivate, but do not directly test, the competence-conditioned routing hypothesis: models may favor frameworks where they are more accurate, while cultural cues may expose framework-specific competence gaps.

Chinese Translation

当一个问题在不同的规范框架下有有效答案时，语言模型必须决定使用哪个框架以及是否能够在该框架内正确回答。我们称这种情境为规范多元主义，并在伊斯兰金融中进行研究，采用四选分类法将框架选择与框架内正确性分开。这种分离揭示了刻板印象陷阱：文化线索引导模型朝向一个框架，但模型在该框架内选择了一个错误的答案。在十二个模型、两种语言和五十个人口信号的研究中，文化线索改变了框架选择，并揭示了准确性方面的显著差异，尤其是在非前沿模型中。在最强信号下，大型开放权重模型97%的时间选择伊斯兰框架。虽然两选评估会报告近乎完美的一致性，但其中57%至66%的选择是错误的。这些发现激励了但并未直接检验能力条件路由假设：模型可能更倾向于选择其准确性更高的框架，而文化线索可能揭示框架特定的能力差距。

View on arXiv Download PDF AI Translation

cs.CL / 82 / 2609.01016

Phrase-Localized Language-Contrastive Guidance: Training-Free Localized Accent Control for Code-Switching Text-to-Speech

短语本地化语言对比引导：无训练的代码切换文本到语音的本地口音控制

Lee, Che Hyun, Park, Sangkwon, Kang, Donghun, Lee, Dongwook, Cho, Youngho, Kim, Heeseung, Yoon, Sungroh

Abstract

Current speech synthesis struggles with code-switching, which mixes a foreign language phrase into a primary language utterance, causing the phrase to be spoken with the primary language's accent rather than its native one. We propose Phrase-Localized Language-Contrastive Guidance (LCG), a training-free inference framework that restores a native accent to code-switched phrases in cross-lingual text-to-speech. LCG replaces the single language guidance applied across the whole utterance with a separate guidance for each region, so each part is guided by its own language. To choose where to apply this localized guidance, we propose a self-attention probing technique that finds the phrase boundaries without external alignments. Together, these components generate speech in which each region carries the accent of its own language, requiring no fine-tuning or auxiliary models. Across diverse language pairs, LCG robustly increases the nativeness of the code-switched phrase while suppressing accent leakage, and preserving overall speaker identity and naturalness.

Chinese Translation

当前的语音合成在代码切换方面存在困难，代码切换是指在主要语言的表述中混入外语短语，导致短语以主要语言的口音而非其母语口音被发音。我们提出了短语本地化语言对比引导（Phrase-Localized Language-Contrastive Guidance, LCG），这是一种无训练的推理框架，旨在为跨语言文本到语音中的代码切换短语恢复母语口音。LCG用每个区域的单独引导替代了在整个表述中应用的单一语言引导，从而使每个部分都由其自身的语言引导。为了选择应用这种本地化引导的位置，我们提出了一种自注意力探测技术，该技术能够在没有外部对齐的情况下找到短语边界。这些组件共同生成的语音使得每个区域都带有其自身语言的口音，无需微调或辅助模型。在多种语言对中，LCG有效地增强了代码切换短语的本土性，同时抑制了口音泄漏，并保持了整体说话者身份和自然性。

View on arXiv Download PDF AI Translation

cs.CL / 83 / 2609.01024

PCoMoE: Shifting MoE Inference from Monolithic Expert Selection to Fine-Grained Path Composition

PCoMoE：将 MoE 推理从单一专家选择转向细粒度路径组合

Gan, Ziyan, Liu, Fangxin, Guan, Chenyang, Wang, Junjie, Yang, Ning, Li, Haomin, Li, Xiang, Yang, Siran, Wang, Jiamang, Qu, Lin, Wang, Zongwu, Jiang, Li, Guan, Haibing

Abstract

Mixture-of-Experts (MoE) architectures scale Large Language Model (LLM) capacity efficiently by activating a sparse subset of experts per token. However, modern MoE inference remains heavily constrained by the rigid, whole-expert abstraction. Existing frameworks manage, schedule, or prune experts as atomic execution units, which fixes the optimization boundary too early and leaves fine-grained intra-expert computational redundancy underexplored. In this work, we present PCoMoE, a path-compositional execution framework that shifts MoE inference from coarse-grained expert selection to fine-grained path composition. PCoMoE incorporates a path-level formulation of expert computation, a compatibility-aware layer-wise pruning strategy to suppress low-value path combinations, and a hardware-friendly execution engine to exploit reusable sub-expert structures under strictly bounded overheads. Experimental results demonstrate that PCoMoE achieves up to a 1.31x end-to-end inference speedup while enhancing model accuracy by 10%. The code is available at https://github.com/gzyyy0/PCoMoE

Chinese Translation

混合专家（MoE）架构通过为每个标记激活稀疏的专家子集，有效地扩展了大语言模型（LLM）的容量。然而，现代 MoE 推理仍然受到刚性整体专家抽象的严重限制。现有框架将专家作为原子执行单元进行管理、调度或修剪，这过早地固定了优化边界，导致细粒度的专家内部计算冗余未被充分探索。在本研究中，我们提出了 PCoMoE，一种路径组合执行框架，将 MoE 推理从粗粒度的专家选择转向细粒度的路径组合。PCoMoE 结合了专家计算的路径级公式、兼容性感知的层级修剪策略以抑制低价值路径组合，以及一个硬件友好的执行引擎，以在严格的开销限制下利用可重用的子专家结构。实验结果表明，PCoMoE 实现了高达 1.31 倍的端到端推理加速，同时提高了模型准确性 10%。代码可在 https://github.com/gzyyy0/PCoMoE 获取。

View on arXiv Download PDF AI Translation

cs.CL / 84 / 2609.01048

Lagged Coupling: Internal Representations Become Readable Before They Become Causal

滞后耦合：内部表征在成为因果之前变得可读

Xun, Xining

Abstract

Across the full Pythia suite (160M-12B, eight checkpoints, four task families), a linear probe can read a target variable from the residual stream as early as step 1,000 at every scale -- yet steering along that same reading direction remains null-equivalent in 43 of 48 model-checkpoint cells. Internal readability systematically outruns causal efficacy, and the lag does not shrink with scale. We call this structure lagged coupling and decompose it into three dissociable tracks: (i) internal readability, saturated (AUROC >= 0.990) from the first checkpoint everywhere; (ii) behavioral readability, which develops gradually and progressively later at larger scales (12B reaches 0.909 only at the final checkpoint); (iii) causal efficacy, almost always null-equivalent, occasionally counterproductive early, with one isolated positive pulse (12B, step 8,000, z = +2.49) our grid cannot resolve. The ordering is dominantly read-before-write (11/11 units, no inversion). Representation headroom along the probe direction grows up to 57x with training and scale while causal write-in stays below 0.11% of headroom -- the variable is increasingly written into the representation and increasingly ignored by the readout. Under a fully pre-registered protocol, both single-onset hypotheses resolve INDETERMINATE (scale slope +0.24, 95% CI [-0.60, +0.87]; time vote 3:3) -- a disciplined negative explained by the three-track decomposition. A pre-registered OLMo-2 replication preserves the direction at attenuated magnitude. Our results caution against inferring steerability from probe accuracy and establish a developmental bottleneck: representation formation reliably outpaces causal readout consolidation.

Chinese Translation

在完整的 Pythia 套件（160M-12B，八个检查点，四个任务家族）中，线性探针可以在每个规模的第 1,000 步从残差流中读取目标变量——然而，在 48 个模型检查点单元中，有 43 个的引导方向仍然是零等效的。内部可读性系统性地超越了因果效能，并且这种滞后在规模上并没有缩小。我们称这种结构为滞后耦合，并将其分解为三个可分离的轨道：（i）内部可读性，从第一个检查点开始在各处饱和（AUROC >= 0.990）；（ii）行为可读性，逐渐发展，并在更大规模上逐步延迟（12B 仅在最后一个检查点达到 0.909）；（iii）因果效能，几乎总是零等效，早期偶尔产生反效果，只有一个孤立的正脉冲（12B，第 8,000 步，z = +2.49）我们的网格无法解析。排序主要是读在写之前（11/11 单元，无反转）。沿探针方向的表征余量随着训练和规模的增长而增加，最高可达 57 倍，而因果写入保持在 0.11% 以下——该变量越来越多地被写入表征中，并且越来越被读取所忽视。在完全预注册的协议下，两个单一发生假设均解析为不确定（规模斜率 +0.24，95% CI [-0.60, +0.87]；时间投票 3:3）——这种有纪律的负面结果可以通过三轨分解来解释。预注册的 OLMo-2 复制保持了方向但幅度减弱。我们的结果警告不要从探针准确性推断可引导性，并确立了一个发展瓶颈：表征形成可靠地超越了因果读取巩固。

View on arXiv Download PDF AI Translation

cs.CL / 85 / 2609.01068

OUTLETS: Output-Length Prediction from Speculative Decoding Backbones

OUTLETS：基于推测解码骨干的输出长度预测

Wen, Weihuang, Liu, Yingying, Liu, Yichuan, Zeng, Wenqi, Zhou, Li, Sun, Chumin, Sun, Jie, Yu, Tianshu

Abstract

The heavy-tailed distribution of output lengths in Large Language Model (LLM) serving poses major challenges for resource provisioning and cluster scheduling. Although output-length prediction can mitigate these issues, existing approaches have key drawbacks: external proxy models add substantial latency and often have limited fidelity, whereas internal state-based methods are efficient but rely on shallow probes of current model states. We identify a structural connection between speculative decoding (SD) and length prediction: latent representations produced by the draft decoder in advanced frameworks (e.g., EAGLE-3) encode signals that are predictive of generation length. Building on this insight, we introduce OUTLETS (Output-Length Prediction from Speculative Decoding Backbones), which repurposes the speculative backbone as a trajectory-aware length predictor. When its draft representations are already computed for speculative decoding, OUTLETS adds only a lightweight regression head and achieves lower MAE than the evaluated methods. Under saturated disaggregated serving, OUTLETS predictions enable standard scheduling policies to prioritize shorter requests and distribute requests more evenly across decoding instances, reducing short-request P99 latency by 34.8%.

Chinese Translation

大型语言模型（LLM）服务中输出长度的重尾分布对资源配置和集群调度提出了重大挑战。尽管输出长度预测可以缓解这些问题，但现有方法存在关键缺陷：外部代理模型增加了显著的延迟，并且通常具有有限的准确性，而基于内部状态的方法虽然高效，但依赖于对当前模型状态的浅层探测。我们识别出推测解码（SD）与长度预测之间的结构性联系：在先进框架（例如 EAGLE-3）中，草稿解码器生成的潜在表示编码了与生成长度相关的信号。基于这一见解，我们提出了OUTLETS（基于推测解码骨干的输出长度预测），该方法将推测骨干重新用于轨迹感知的长度预测。当其草稿表示已为推测解码计算时，OUTLETS仅添加一个轻量级回归头，并且实现了比评估方法更低的平均绝对误差（MAE）。在饱和的分散式服务下，OUTLETS的预测使标准调度策略能够优先处理较短的请求，并在解码实例之间更均匀地分配请求，从而将短请求的P99延迟减少了34.8%。

View on arXiv Download PDF AI Translation

cs.CL / 86 / 2609.01073

Post-hoc Alignment of LLM-judges to Human Judgment Distribution

后验对齐大型语言模型评判者与人类判断分布

Steindl, Sebastian, Voskarides, Nikos, Gasparin, Alberto, Marcheggiani, Diego

Abstract

The LLM-as-a-judge (LLMaJ) framework offers a cost-effective and reproducible solution for automatic evaluation. However, current evaluation practices typically compare LLMaJ judgments against aggregated ground-truth labels, overlooking the valuable information contained in Human Label Variation (HLV). Inspired by an increasing line of work that proposes to leverage HLV, we systematically study LLMaJ performance on predicting both a single, aggregated ground truth hard-label and unaggregated soft-labels that represent Human Judgment Distributions (HJD). Our results across five diverse datasets reveal that while LLMs achieve near human-level performance at hard-label prediction on most tasks, they exhibit poor performance when predicting soft-labels. To address this limitation, we propose NAPHA (eNtropy-Aware Post-Hoc Alignment), a simple yet effective lightweight post-hoc alignment method that matches the LLM distribution to the HJD by first assigning an instance to a discrete entropy class and then routing it to specialized, trained alignment models. We find that NAPHA consistently improves soft-labels prediction across base LLM models and datasets, with particularly strong gains on high-entropy instances where capturing diverse human perspectives is most critical. We also show via oracle experiments that improving entropy class prediction can substantially enhance NAPHA's practical effectiveness.

Chinese Translation

大型语言模型作为评判者（LLM-as-a-judge, LLMaJ）框架提供了一种具有成本效益和可重复性的自动评估解决方案。然而，目前的评估实践通常将LLMaJ的判断与聚合的真实标签进行比较，忽视了人类标签变异（Human Label Variation, HLV）中包含的宝贵信息。受到越来越多的研究启发，这些研究提出利用HLV，我们系统地研究了LLMaJ在预测单一聚合真实硬标签和表示人类判断分布（Human Judgment Distributions, HJD）的未聚合软标签方面的表现。我们在五个不同的数据集上的结果表明，尽管LLM在大多数任务的硬标签预测中达到了接近人类水平的表现，但在预测软标签时表现较差。为了解决这一局限性，我们提出了NAPHA（eNtropy-Aware Post-Hoc Alignment），这是一种简单而有效的轻量级后验对齐方法，通过首先将实例分配到离散熵类，然后将其路由到专门训练的对齐模型，从而将LLM分布与HJD进行匹配。我们发现，NAPHA在基础LLM模型和数据集上始终改善了软标签的预测，尤其在高熵实例上取得了显著提升，在这些情况下捕捉多样化的人类视角尤为重要。我们还通过oracle实验表明，改善熵类预测可以显著增强NAPHA的实际有效性。

View on arXiv Download PDF AI Translation

cs.CL / 87 / 2609.01081

StateSwap: Probing Support-Elimination Hidden States in Multiple-Choice Questions

状态交换：探究多项选择题中的支持消除隐藏状态

Gao, Chao, Liu, Haijiang, Li, Qiyuan, Guo, Caicai, van Harmelen, Frank, Gu, Jinguang

Abstract

Large language models often answer the same multiple-choice question inconsistently when it is posed under support-oriented and elimination-oriented framings. We investigate whether these discrepancies arise from different internal representations induced by the two framings. We introduce a dual-framing protocol with minimally varied prompts that use either support- or elimination-oriented framing while keeping the evaluation target fixed. To probe the internal computation, we append an untrained special token, [STATE], and treat its residual-stream activation as an intervention interface. Across both models, the two framings induce separable [STATE] activations concentrated in intermediate layers. Swapping these activations between paired prompts systematically changes predictions and improves cross-framing agreement, providing intervention-based evidence that the activations are behaviorally relevant. Beyond instance-level substitution, mean-difference steering directions derived from the dual-framing contrast exhibit more bounded layer-wise responses than matched contrastive activation addition directions under the evaluated protocol.

Chinese Translation

大型语言模型在支持导向和消除导向的框架下回答相同的多项选择题时，往往表现出不一致性。我们研究这些差异是否源于这两种框架所诱发的不同内部表征。我们引入了一种双框架协议，使用最小变化的提示，采用支持导向或消除导向的框架，同时保持评估目标不变。为了探测内部计算，我们附加了一个未训练的特殊标记 [STATE]，并将其残差流激活视为干预接口。在两个模型中，这两种框架诱发了集中在中间层的可分离 [STATE] 激活。将这些激活在配对提示之间交换，系统性地改变了预测并改善了跨框架的一致性，提供了基于干预的证据，表明这些激活在行为上是相关的。除了实例级的替代外，从双框架对比中导出的均值差异引导方向展现出比在评估协议下匹配对比激活添加方向更为有限的层级响应。

View on arXiv Download PDF AI Translation

cs.CL / 88 / 2609.01100

Beyond Magnitude: Contrastive Routing for Modular Mixture-of-Experts

超越幅度：模块化专家混合模型的对比路由

Xiros, Nikolaos, Damianos, Dimitrios, Zoumpoulidi, Maria-Eleni, Voukoutis, Leon, Katsouros, Vassilis, Paraskevopoulos, Georgios

Abstract

In current Mixture-of-Experts architectures, routing is performed based on representations dominated by structure shared across all tokens, limiting expert specialization. We show that contrasting each token against an Exponential Moving Average of the layer's hidden states, rather than routing on absolute magnitude, concentrates the routing signal onto a low-dimensional, highly separable subspace. Building on this, we propose the Contrastive Routing Mechanism (CoRM), which scores each expert by the gap between its affinity for the incoming token and its affinity for this shared reference state, interpreted through a distinct per-expert projection. The resulting experts have routing boundaries that align with linguistic structure significantly more than the Top-k baseline. Our experiments show that CoRM improves average zero-shot accuracy by +0.67 to +1.69 points (Top-1) and +1.38 to +1.77 points (Top-2) over standard Top-k MoE baselines on nine zero-shot reasoning benchmarks, at the minimal cost of 2.9% added parameters and 2.6% added FLOPs per token.

Chinese Translation

在当前的专家混合模型架构中，路由是基于所有标记共享的结构主导的表示进行的，这限制了专家的专业化。我们展示了与其基于绝对幅度进行路由，不如将每个标记与层的隐藏状态的指数移动平均进行对比，这样可以将路由信号集中在一个低维、高可分离的子空间上。在此基础上，我们提出了对比路由机制（Contrastive Routing Mechanism, CoRM），该机制通过专家对输入标记的亲和力与其对该共享参考状态的亲和力之间的差距来评分每个专家，这一过程通过每个专家的独特投影进行解释。结果显示，所得到的专家的路由边界与语言结构的对齐程度显著高于Top-k基线。我们的实验表明，CoRM在九个零-shot推理基准上，相较于标准的Top-k专家混合模型基线，平均零-shot准确率提高了+0.67到+1.69个百分点（Top-1）和+1.38到+1.77个百分点（Top-2），同时每个标记仅增加了2.9%的参数和2.6%的FLOPs。

View on arXiv Download PDF AI Translation

cs.CL / 89 / 2609.01103

When Modality Gap Reduction Fails: Prediction-Level Hubness in CLIP

当模态间隙缩减失败时：CLIP中的预测级中心性

Sato, Shota, Kiyama, Hajime, Hirasawa, Tosho, Komachi, Mamoru

Abstract

Reducing the modality gap between image and text representations in CLIP is widely expected to improve cross-modal alignment and downstream performance. However, a smaller average image-text gap does not necessarily lead to consistent accuracy gains. We analyze this mismatch from the perspective of the decision structure in zero-shot classification, i.e. selecting the most similar class-text prototype for an input image. Zero-shot accuracy depends not only on average image--text alignment, but also on class-wise decision margins. Using Linear correction as an analytically tractable case, we show that modality gap correction can alter the relative decision structure among classes and cause predictions to concentrate on a small subset of classes. We refer to this output-space failure mode as prediction-level hubness. Furthermore, experiments across multiple datasets show that accuracy degradation under gap correction is consistently associated with increased prediction concentration, both for Linear correction and for learning-based correction methods. This provides a systematic explanation of why modality gap reduction does not consistently improve CLIP zero-shot accuracy from the perspective of downstream decision structure. Our results suggest that gap correction should be evaluated not only by average alignment, but also by its impact on downstream prediction structure.

Chinese Translation

在CLIP中，减少图像和文本表示之间的模态间隙被广泛期望能够改善跨模态对齐和下游性能。然而，较小的平均图像-文本间隙并不一定会导致一致的准确性提升。我们从零-shot 分类的决策结构角度分析这种不匹配，即为输入图像选择最相似的类-文本原型。零-shot 准确性不仅依赖于平均的图像-文本对齐，还依赖于类的决策边际。通过将线性校正作为一个可分析的案例，我们展示了模态间隙的校正可以改变类之间的相对决策结构，并导致预测集中在少数类上。我们将这种输出空间的失败模式称为预测级中心性。此外，在多个数据集上的实验表明，在间隙校正下准确性下降与预测集中度增加之间存在一致的关联，无论是对于线性校正还是基于学习的校正方法。这为模态间隙缩减为何未能一致改善CLIP零-shot准确性提供了系统性的解释，视角来自于下游决策结构。我们的结果表明，间隙校正的评估不仅应考虑平均对齐，还应考虑其对下游预测结构的影响。

View on arXiv Download PDF AI Translation

cs.CL / 90 / 2609.01111

ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues

ClinTraceBench：可源验证的基于电子健康记录衍生对话的纵向临床推理

Wang, Huimin, Zhao, Zhengyi, Zhao, Yutian

Abstract

Clinical LLM assistants must reason over multi-visit patient trajectories, yet whether the compact history representations used to scale them---retrieval, structured timelines, LLM summaries, agentic memory---preserve the longitudinal signal clinical reasoning needs has not been measured. We introduce ClinTraceBench: 385 MIMIC-IV-derived verified dialogues with event-ID provenance, a nine-task taxonomy (T1--T9), and L0--L4 deterministic + L5 human-audit validation (98.92\% agreement). We evaluate eight history representation strategies---a no-context floor, \textit{last-visit-only}, \textit{full-context}, BGE-M3 \textit{dense-retrieval}, two compression schemes, and two agentic-memory systems (\textit{Mem0}, \textit{A-Mem})---across four backbones (DeepSeek-V3, GPT-4o-mini, Haiku~4.5, Sonnet~4.6) on 6{,}271 questions: 32 cells, 200{,}672 predictions. Four findings: (SP4) a controlled T3 injection probe isolates compression-induced \textit{relation} loss---with the attribution sentence present \textit{before} construction, \textit{Mem0}, \textit{A-Mem} and \textit{llm-summary} still recover only 0--5.3\% of the injected positives; (SP1) compressed strategies pay an aggregation tax on multi-visit trends and cross-patient comparisons; (SP2) the blind-to-full gap spans $+29.8$~pp (GPT-4o-mini) to $+62.7$~pp (Haiku); (SP3) abstention scales non-monotonically with context length. On the Pareto frontier Haiku dominates Sonnet under \textit{full-context} (\$25.76 vs.\ \$106.21), inverting the ``biggest backbone wins'' heuristic.

Chinese Translation

临床大语言模型助手必须对多次就诊的患者轨迹进行推理，但用于扩展它们的紧凑历史表示——检索、结构化时间线、大语言模型摘要、代理记忆——是否保留了临床推理所需的纵向信号尚未被测量。我们引入了ClinTraceBench：385个基于MIMIC-IV的经过验证的对话，具有事件ID来源，包含九个任务分类（T1--T9），以及L0--L4的确定性验证和L5的人为审核验证（98.92%的一致性）。我们评估了八种历史表示策略——无上下文基线、仅最后一次就诊、完整上下文、BGE-M3密集检索、两种压缩方案和两种代理记忆系统（Mem0、A-Mem）——在四个基础模型（DeepSeek-V3、GPT-4o-mini、Haiku 4.5、Sonnet 4.6）上针对6,271个问题进行评估：32个单元，200,672个预测。四个发现：（SP4）受控的T3注入探针隔离了压缩引起的关系损失——在构建之前存在的归因句，Mem0、A-Mem和llm-summary仍然仅恢复了0-5.3%的注入正例；（SP1）压缩策略在多次就诊趋势和跨患者比较上支付了聚合税；（SP2）盲目与完整之间的差距跨度为+29.8个百分点（GPT-4o-mini）到+62.7个百分点（Haiku）；（SP3）弃权与上下文长度的关系呈非单调性。在帕累托前沿，Haiku在完整上下文下优于Sonnet（$25.76对$106.21），颠覆了“最大的基础模型胜出”的启发式规则。

View on arXiv Download PDF AI Translation

cs.CL / 91 / 2609.01113

EDRAC: Benchmarking Arabic Dialect Reading Comprehension

EDRAC：阿拉伯方言阅读理解基准测试

Mokh, Noor Abo, Chirkunov, Kirill, Lynn, Teresa, Habash, Nizar, Marzouk, Reham, Altakrori, Malik H., Samih, Younes, Odeh, Muhammed Abu, Rabih, Nour, Alshahrani, Rahaf, Alshehhi, Hamad, Al-Ali, Hamdan, Almahri, Muhra, Hassan, Besher, Anwar, Mohamed, Freihat, Abed Alhakim, Nakov, Preslav, Aji, Alham Fikri

Abstract

Dialectal Arabic (DA) remains under-resourced compared to Modern Standard Arabic (MSA), particularly for machine reading comprehension (MRC) and question answering (QA). Existing Arabic QA benchmarks primarily focus on formal written MSA or multiple-choice QA, with limited coverage of naturally spoken dialects. Here, we aim to bridge this gap. We introduce EDRAC, the first large-scale benchmark for dialectal Arabic machine reading comprehension (MRC) and generative QA, covering five major dialects: Egyptian, Moroccan, Emirati, Syrian, and Saudi Arabic. EDRAC contains 499 passages derived from naturally occurring spoken interactions and 4,977 corresponding QA pairs generated through a human--LLM collaborative pipeline combining iterative generation, LLM-as-a-judge evaluation, and human verification. We benchmark Arabic-centric and multilingual LLMs on EDRAC using lexical and semantic metrics. Our results reveal substantial gaps between semantic answer quality and dialectal fidelity, highlighting the limitations of existing evaluation metrics for dialectal Arabic generation. EDRAC provides a realistic and challenging MRC benchmark for future research on dialectal Arabic NLP.

Chinese Translation

与现代标准阿拉伯语（MSA）相比，方言阿拉伯语（DA）仍然资源匮乏，尤其是在机器阅读理解（MRC）和问答（QA）方面。现有的阿拉伯语问答基准主要集中在正式书面MSA或多项选择问答上，对自然口语方言的覆盖有限。在此，我们旨在填补这一空白。我们推出了EDRAC，这是第一个大规模的方言阿拉伯语机器阅读理解（MRC）和生成式问答（QA）基准，涵盖五种主要方言：埃及方言、摩洛哥方言、阿联酋方言、叙利亚方言和沙特阿拉伯方言。EDRAC包含499个段落，来源于自然发生的口语互动，以及4,977对通过人类与大语言模型（LLM）协作生成的问答对，这一过程结合了迭代生成、LLM作为评判者的评估和人工验证。我们使用词汇和语义指标对阿拉伯语中心和多语言LLM在EDRAC上的表现进行了基准测试。我们的结果揭示了语义答案质量与方言忠实度之间的显著差距，突显了现有评估指标在方言阿拉伯语生成中的局限性。EDRAC为未来方言阿拉伯语自然语言处理的研究提供了一个现实且具有挑战性的MRC基准。

View on arXiv Download PDF AI Translation

cs.CL / 92 / 2609.01135

Overfitting Mitigation via Singular Value Decomposition in Minimum Bayes Risk Decoding

通过奇异值分解减轻最小贝叶斯风险解码中的过拟合

Soetedjo, Riza Setiawan, Sakai, Yusuke, Kamigaito, Hidetaka, Hayashi, Katsuhiko, Watanabe, Taro

Abstract

Minimum Bayes Risk (MBR) decoding enables high-quality text generation by selecting the hypothesis that maximizes a utility metric over sampled pseudo-references. However, it is highly susceptible to metric overfitting: it can irregularly inflate the chosen utility metric at the direct expense of other unoptimized evaluation metrics. To mitigate this, we introduce SVD-MBR, which frames the pairwise utility matrix as a noisy information signal. By computing a low-rank approximation via Singular Value Decomposition (SVD) and retaining only the top-$k$ components, we effectively decouple true consensus from metric noise. Experiments demonstrate that SVD-MBR successfully regularizes decoding, yielding substantial gains across a range of generalized metrics. Furthermore, we reveal that this denoising is metric-dependent: neural metrics encode a robust low-rank consensus ideal for SVD, whereas surface-level metrics struggle to separate signal from metric noise.

Chinese Translation

最小贝叶斯风险（MBR）解码通过选择最大化采样伪参考的效用指标的假设，实现高质量的文本生成。然而，它对指标过拟合高度敏感：这可能会不规则地膨胀所选择的效用指标，从而直接影响其他未优化的评估指标。为此，我们提出了SVD-MBR，它将成对效用矩阵视为一种噪声信息信号。通过奇异值分解（SVD）计算低秩近似，并仅保留前$k$个分量，我们有效地将真实共识与指标噪声解耦。实验表明，SVD-MBR成功地对解码进行正则化，在一系列广义指标上获得显著提升。此外，我们揭示这种去噪是依赖于指标的：神经指标编码了适合SVD的稳健低秩共识，而表面指标则难以将信号与指标噪声分离。

View on arXiv Download PDF AI Translation

cs.CL / 93 / 2609.01139

Does task decomposition improve automatic NLG evaluation?

任务分解是否改善自动自然语言生成评估？

Steindl, Sebastian, Voskarides, Nikos, Gasparin, Alberto, Marcheggiani, Diego

Abstract

The LLM-as-a-judge (LLMaJ) framework has emerged as a promising solution for cheap, reproducible, reference-free Natural Language Generation (NLG) evaluation. Prior work seeks to improve LLMaJ by decomposing evaluation tasks into simpler sub-tasks. In this work, we systematically compare LLMaJ methods with and without decomposition on multiple NLG datasets. We find no evidence that LLMaJ with task decomposition leads to performance gains over a fair baseline that does not use decomposition. Instead, we find that previously reported performance gains in decomposition-based LLMaJ stem from using human labels as training data, and not task decomposition itself. Also, we find that, when human labels are available, LLMaJ without using task decomposition can perform comparably to human annotators.

Chinese Translation

LLM作为评审者（LLMaJ）框架已成为一种有前景的解决方案，用于廉价、可重复、无参考的自然语言生成（NLG）评估。之前的研究试图通过将评估任务分解为更简单的子任务来改进LLMaJ。在本研究中，我们系统地比较了使用和不使用任务分解的LLMaJ方法在多个NLG数据集上的表现。我们发现，使用任务分解的LLMaJ并未在不使用分解的公平基线之上带来性能提升的证据。相反，我们发现之前报告的基于分解的LLMaJ的性能提升源于使用人类标签作为训练数据，而不是任务分解本身。此外，我们发现，当人类标签可用时，不使用任务分解的LLMaJ可以与人类注释者的表现相当。

View on arXiv Download PDF AI Translation

cs.CL / 94 / 2609.01151

Subword Segmental BabyLMs: Learning to Tokenise for Sample-Efficient Pretraining

子词分段BabyLMs：为样本高效预训练学习分词

Meyer, Francois

Abstract

In the standard LM training pipeline, subword tokenisation is applied as a preprocessing step. Subword segmental language modelling is an alternative paradigm in which tokenisation is learned during training, allowing the model to discover subword units that optimise its training objective. In this paper, we present our submission to the 2026 BabyLM Challenge, for which we develop two new subword segmental LMs: SubSegGPT and SubSegDeBERTa. SubSegGPT is a decoder-only model that learns tokenisation during autoregressive pretraining. SubSegDeBERTa is an encoder-based model that jointly learns to generate and tokenise masked words. We train both for the Strict and Strict-small tracks. Our top submission to Strict is SubSegDeBERTa, which achieves notable gains in zero-shot evaluation. Our top submission to Strict-small is SubSegGPT, which outperforms tokenisation-based baselines. Our results show that learnable subword tokenisation can improve sample-efficiency for BabyLM pretraining. We analyse the subword learning dynamics of our models and find that tokenisation gradually converges on subword units that balance morphological alignment and fine-grained segmentation.

Chinese Translation

在标准语言模型（LM）训练流程中，子词分词作为预处理步骤进行应用。子词分段语言建模是一种替代范式，在该范式中，分词是在训练过程中学习的，使模型能够发现优化其训练目标的子词单元。本文介绍了我们对2026年BabyLM挑战赛的提交，我们开发了两个新的子词分段语言模型：SubSegGPT和SubSegDeBERTa。SubSegGPT是一个仅解码器模型，在自回归预训练过程中学习分词。SubSegDeBERTa是一个基于编码器的模型，能够联合学习生成和分词被遮蔽的单词。我们为严格（Strict）和严格小型（Strict-small）赛道训练了这两个模型。我们在严格赛道的最佳提交是SubSegDeBERTa，它在零样本评估中取得了显著的提升。在严格小型赛道的最佳提交是SubSegGPT，它超越了基于分词的基线。我们的结果表明，可学习的子词分词可以提高BabyLM预训练的样本效率。我们分析了模型的子词学习动态，发现分词逐渐收敛于在形态对齐和细粒度分段之间取得平衡的子词单元。

View on arXiv Download PDF AI Translation

cs.CL / 95 / 2609.01182

LLMPEDIA: Browsing, Verifying, and Comparing the Parametric Encyclopedic Knowledge of LLMs

LLMPEDIA：浏览、验证和比较大型语言模型的参数化百科知识

Saeed, Muhammed, Razniewski, Simon

Abstract

Flagship language models appear saturated on benchmarks like MMLU (Hendrycks et al., 2021), scoring above 90% - yet benchmarks test only what the experimenter thought to ask, the availability bias of fixed question sets. LLMPEDIA makes this bias measurable and browsable. We recursively materialized ~1.3M articles from three model families' parametric memory (GPT-5-mini, DeepSeek-V3.2, Llama-3.3-70B) without retrieval, then audited a stratified sample of atomic claims against Wikipedia and a curated web stack, coloring every claim supported, refuted, or insufficient (Saeed and Razniewski, 2026). On a uniform random sample the true rate is 68.4% - more than 21 pp below MMLU - with 30.5% of claims insufficient: assertions no benchmark probes and the world's largest encyclopedia cannot adjudicate - long-tail knowledge or plausible hallucination, the evidence cannot tell - extending to free text the coverage gap GPTKB established for triples (Hu et al., 2025). The resulting live, open encyclopedia lets visitors inspect this frontier one claim at a time through five one-click views - link-traversal exploration, claim-level factuality, cross-model and political-persona comparison, and a guided topic drill-down - each page, claim, and verdict at a stable URL. LLMPEDIA is live at https://llmpedia.net

Chinese Translation

旗舰语言模型在诸如 MMLU（Hendrycks 等，2021）等基准测试中似乎已达到饱和，得分超过 90%——然而，基准测试仅测试实验者所想询问的内容，即固定问题集的可用性偏差。LLMPEDIA 使这种偏差可测量且可浏览。我们从三个模型家族的参数记忆中递归生成了约 130 万篇文章（GPT-5-mini、DeepSeek-V3.2、Llama-3.3-70B），无需检索，然后对与维基百科和经过策划的网络堆栈进行的原子声明的分层样本进行了审核，标记每个声明为支持、反驳或不足（Saeed 和 Razniewski，2026）。在一个均匀随机样本中，真实率为 68.4%——比 MMLU 低超过 21 个百分点——其中 30.5% 的声明被标记为不足：这些声明是没有基准探测且世界上最大的百科全书无法裁定的——长尾知识或合理的幻觉，证据无法判断——扩展到自由文本，填补了 GPTKB 为三元组（Hu 等，2025）建立的覆盖差距。最终生成的实时开放百科全书允许访问者通过五种一键视图逐条检查这一前沿——链接遍历探索、声明级事实性、跨模型和政治角色比较，以及引导主题深入——每个页面、声明和裁决都有一个稳定的 URL。LLMPEDIA 现已上线，网址为 https://llmpedia.net

View on arXiv Download PDF AI Translation

cs.CL / 96 / 2609.01188

PersuaRL: Reinforcement Learning-Driven Multi-Expert Selection for Persuasive Dialogue Generation in Insurance

PersuaRL：基于强化学习的多专家选择用于保险领域的劝说对话生成

Kirti, Rohan, Ghosh, Akash, Vats, Aryan, Ghosh, Niladri, Shriparn, Shipra, Ramnani, Roshni, Maitra, Anutosh, Saha, Sriparna

Abstract

Large Language Models (LLMs) are revolutionizing digital communication by powering conversational agents deployed across domains such as customer service, digital sales, and insurance. These agents, built on LLMs, can understand user input, retrieve relevant information, and generate coherent responses. However, while they excel at factual communication, they often lack the ability to engage in truly persuasive, context-sensitive dialogue, especially in domains like insurance, where trust and clarity are critical. Building on this need within the insurance domain, our work focuses on improving the persuasiveness of digital agents, aka LLMs. To support this, we introduce InsureDial, a Persuasive Insurance Dialogue dataset, designed to capture the nuances of persuasive communication specific to motor insurance interactions. We introduce PersuaRL, a reinforcement learning-based framework that equips LLM-driven dialogue agents with the ability to adaptively explore, select, and coordinate strategies across multiple expert modules, guided by the evolving dialogue context, to achieve more effective persuasion. We conduct extensive automatic human and qualitative evaluations on two benchmark persuasion dialogue datasets, including our InsureDial. Our evaluations consistently demonstrate that PersuaRL outperforms baseline, generating contextually appropriate and highly persuasive responses.

Chinese Translation

大型语言模型（LLMs）正在通过驱动跨多个领域（如客户服务、数字销售和保险）的对话代理，彻底改变数字通信。这些基于LLMs构建的代理能够理解用户输入、检索相关信息并生成连贯的响应。然而，尽管它们在事实交流方面表现出色，但在进行真正具有说服力的、上下文敏感的对话时，尤其是在信任和清晰度至关重要的保险领域，它们往往缺乏能力。基于保险领域的这一需求，我们的工作集中于提高数字代理（即LLMs）的说服力。为此，我们引入了InsureDial，一个劝说保险对话数据集，旨在捕捉与机动车保险互动相关的劝说交流的细微差别。我们提出了PersuaRL，一个基于强化学习的框架，使得基于LLM的对话代理能够根据不断变化的对话上下文，自适应地探索、选择和协调多个专家模块的策略，以实现更有效的劝说。我们在两个基准劝说对话数据集（包括我们的InsureDial）上进行了广泛的自动化人类和定性评估。我们的评估结果一致表明，PersuaRL在生成上下文适当且高度具有说服力的响应方面优于基线。

View on arXiv Download PDF AI Translation

cs.CL / 97 / 2609.01195

CaRL-EM: Cost-Aware Reinforcement Learning for Entity Matching with LLMs

CaRL-EM：基于成本感知的强化学习用于实体匹配与大型语言模型

Guo, Chaohui, Klein, Michel, Huang, Zhisheng

Abstract

Entity matching (EM) requires fine-grained contextual understanding and domain knowledge. Recent work shows that large language models (LLMs) can serve as strong matchers across domains, but most methods either make independent pairwise decisions or rely on manually designed composite pipelines, thus lacking flexibility in realistic multi-candidate settings. At the same time, they typically ignore inference cost at scale. We formulate LLM-based EM with candidates as a cost-aware sequential decision problem and propose CaRL-EM, a reinforcement learning controller that manages LLM operations. Given the state of an anchor record, its candidate set, and the cost, CaRL-EM adaptively chooses among different operators (Match/Compare/Select/Decide) and model capacities to maximize a quality-cost objective. The policy interacts with abstract operators, allowing the same controller to be reused with different underlying LLM backends at inference time without retraining. Experiments on 7 benchmarks show that CaRL-EM (i) learns to dynamically plan the usage of inexpensive and expensive operators based on task complexity, (ii) achieves robust zero-shot transfer across diverse datasets and domains, and (iii) consistently achieves a better quality-cost trade-off than strong LLM-based baselines and manually designed pipelines, yielding a lower inference cost at comparable or higher quality.

Chinese Translation

实体匹配（EM）需要细致的上下文理解和领域知识。近期研究表明，大型语言模型（LLMs）可以在多个领域中作为强大的匹配工具，但大多数方法要么做出独立的成对决策，要么依赖于手动设计的复合管道，从而在现实的多候选设置中缺乏灵活性。同时，它们通常忽视了大规模推理的成本。我们将基于LLM的实体匹配问题表述为一个成本感知的顺序决策问题，并提出了CaRL-EM，一个管理LLM操作的强化学习控制器。给定锚记录的状态、候选集和成本，CaRL-EM自适应地在不同的操作符（匹配/比较/选择/决策）和模型能力之间进行选择，以最大化质量-成本目标。该策略与抽象操作符交互，使得同一控制器可以在推理时与不同的底层LLM后端重用，而无需重新训练。在7个基准测试上的实验表明，CaRL-EM (i) 学会根据任务复杂性动态规划廉价和昂贵操作符的使用，(ii) 在不同数据集和领域中实现了稳健的零-shot迁移，(iii) 始终实现比强大的基于LLM的基线和手动设计管道更好的质量-成本权衡，在相当或更高的质量下降低了推理成本。

View on arXiv Download PDF AI Translation

cs.CL / 98 / 2609.01202

Towards AI-Assisted Clinical Trial Matching: Practical Considerations, Multicenter Evaluation, and Real-World Deployment

朝向AI辅助的临床试验匹配：实践考虑、多中心评估与现实世界部署

Fang, Yin, Jin, Qiao, Tian, Shubo, He, Lauren, Geer, Maya, Naffakh, Noor, Nguyen, Ryan Huu-Tuan, Wang, Zifeng, Sun, Jimeng, Floudas, Charalampos S., Gulley, James L., Moalem, Kamilia, Maia, Catarina Martins, Nottke, Amanda, Valle, Juan W., Bachini, Melinda, Rocha-Nussbaum, Lourdes, Ramage, Kari, Curry, Nikita, Barnes, Megan, Mansaray, Mandy, Gabeau, Darlene, Grossman, Craig E., Skinner, Heath, Burczynski, Michael, Consortium, NIH-TrialBench, Lu, Zhiyong

Abstract

Clinical trials are essential for advancing cancer care and drug development, but many fail because of insufficient patient enrollment. While there is growing interest in using AI to support patient recruitment, existing systems largely perform eligibility assessment alone and have rarely been evaluated in real-world oncology workflows. Here we present TrialGPT 2.0, an AI-assisted clinical trial recommendation system designed for real-world deployment. Rather than asking only whether a patient may qualify, the system also assesses which trials warrant further consideration given the patient's current clinical needs and local workflow priorities, and provides structured, inspectable explanations for expert review. Importantly, we evaluated TrialGPT 2.0 retrospectively and prospectively across multiple oncology-focused settings, spanning government, academic cancer-center, patient-advocacy, and NIH referral workflows. In retrospective multicenter cohorts comprising 288 cases, TrialGPT 2.0 retrieved at least one clinician-recommended trial in its top 10 recommendations for approximately 91% of cases while reducing clinician screening time by 55.0%. In a six-month prospective evaluation embedded in an active precision oncology tumor board, TrialGPT 2.0 contributed additional trial opportunities missed by the routine workflow, expanding patient access to clinical trial participation by 90.9%. To support scientific reproducibility, we also introduce NIH-TrialBench, a clinician-authored dataset comprising 126 diverse synthetic patient vignettes and matching scenarios from 11 NIH Institutes and Centers. Together, these results support the value of AI to assist clinical trial matching by improving clinician efficiency and identifying frequently overlooked trial opportunities, ultimately helping to expand and accelerate accrual to cancer trials.

Chinese Translation

临床试验对于推进癌症护理和药物开发至关重要，但许多试验因患者招募不足而失败。尽管对使用AI支持患者招募的兴趣日益增长，但现有系统主要仅进行资格评估，且在现实世界的肿瘤学工作流程中很少被评估。在此，我们介绍了TrialGPT 2.0，一种旨在现实世界部署的AI辅助临床试验推荐系统。该系统不仅评估患者是否符合资格，还考虑患者当前的临床需求和当地工作流程优先事项，评估哪些试验值得进一步考虑，并提供结构化、可检查的解释以供专家审查。重要的是，我们在多个以肿瘤学为重点的环境中对TrialGPT 2.0进行了回顾性和前瞻性的评估，涵盖了政府、学术癌症中心、患者倡导和NIH转诊工作流程。在包含288例的回顾性多中心队列中，TrialGPT 2.0在其前10个推荐中为约91%的案例检索到了至少一个临床医生推荐的试验，同时将临床医生筛查时间减少了55.0%。在一个嵌入活跃精准肿瘤委员会的六个月前瞻性评估中，TrialGPT 2.0提供了常规工作流程中遗漏的额外试验机会，使患者参与临床试验的机会增加了90.9%。为了支持科学可重复性，我们还介绍了NIH-TrialBench，这是一个由临床医生撰写的数据集，包含来自11个NIH研究所和中心的126个多样化的合成患者小插曲和匹配场景。这些结果共同支持了AI在临床试验匹配中的价值，通过提高临床医生的效率和识别常被忽视的试验机会，最终帮助扩大和加速癌症试验的招募。

View on arXiv Download PDF AI Translation

cs.CL / 99 / 2609.01246

Ready to Speak: Aligning LLMs for TTS-Friendly Text Generation

准备发声：为文本生成对TTS友好的大语言模型对齐

Thonet, Thibaut, Rozen, Jos, Besacier, Laurent

Abstract

Current Large Language Models (LLMs) are primarily optimized for written text, often producing outputs that are grammatically correct and helpful yet poorly suited for spoken delivery via Text-to-Speech (TTS). In this work, we study how to make LLMs natively generate TTS-friendly text, which we frame as a preference alignment problem: instead of relying on downstream rewriting modules, we directly align LLMs to generate text optimized for spoken delivery. We introduce two preference datasets spanning different target domains, CORA and Recipe, which contain paired TTS-friendly and TTS-unfriendly responses. We further propose an evaluation suite combining a pattern-based heuristic metric, a TTS$\to$ASR evaluation pipeline, and a MUSHRA listening study with human judges. Our experiments compare the recently proposed Feature-aware Sampling and Tuning (FaST) framework -- leveraging interpretable features instead of a black-box reward model -- against an array of alignment baselines on the TTS-friendly generation task. Notably, we found that FaST achieves the best overall tradeoff between TTS-friendliness and helpfulness across various settings. We also identified a strong correlation between our different metrics, highlighting the ability to reliably assess TTS-friendliness via an efficient heuristic.

Chinese Translation

当前的大型语言模型（LLMs）主要针对书面文本进行优化，通常生成的输出在语法上是正确且有帮助的，但不适合通过文本转语音（TTS）进行口头表达。在本研究中，我们探讨如何使LLMs原生生成对TTS友好的文本，我们将其框架视为一个偏好对齐问题：我们不依赖下游重写模块，而是直接对LLMs进行对齐，以生成针对口头表达优化的文本。我们引入了两个跨不同目标领域的偏好数据集，CORA和Recipe，其中包含配对的TTS友好和TTS不友好的响应。我们进一步提出了一套评估工具，结合了基于模式的启发式指标、TTS到ASR的评估管道，以及与人类评审者进行的MUSHRA听觉研究。我们的实验比较了最近提出的特征感知采样和调优（FaST）框架——利用可解释特征而非黑箱奖励模型——与一系列对齐基线在TTS友好生成任务上的表现。值得注意的是，我们发现FaST在各种设置中实现了TTS友好性和有用性之间的最佳整体权衡。我们还发现不同指标之间存在强相关性，突显了通过有效的启发式方法可靠评估TTS友好性的能力。

View on arXiv Download PDF AI Translation

cs.CL / 100 / 2609.01274

From Base Rollouts to RL Reasoning: A Budgeted Search Perspective

从基本回滚到强化学习推理：预算搜索视角

Sun, Wenhe, Wang, Cunxiang, Yao, Zijun, Cao, Yixin

Abstract

Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning, but how these gains relate to inference-time decoding and search remains unclear. Does RL create reasoning the base model lacks, or shift the rollout distribution toward trajectories it can already reach but rarely samples? We study this behaviorally with a Unified Decoding Framework (UDF), which expresses token-level sampling, beam-like search, tree search, and sequence-level resampling as executable policies over a shared budgeted operating space, scored post hoc with pass@$k$, self-consistency, best-of-$N$, and first-finish success. Using paired Base/RL checkpoints from SimpleRL-Zoo, we ask whether an RL default-policy curve can be approximated by a structured path of Base operating points. On Math500, AIME, GPQA, and IFEval, the pass@$k$ recovery path follows a Budgeted Operating-Point Transition Rule (BOPTR), $N_{\mathrm{Base}} \approx \alpha N_{\mathrm{RL}}^{\beta}$, with benchmark-conditioned exponents. On Qwen2.5-7B, BOPTR gives the lowest transfer error among the non-oracle rules we test, 3.41 pp (95% CI [2.32, 5.53]); a three-seed replication gives 3.07 $\pm$ 0.39 pp. The rule extends to ten models across four families (3.28 to 4.87 pp on checkpoints added after fitting), to four benchmarks it was never fitted on (5.03 pp vs. 4.44 pp in fit), and holds without an RL checkpoint for the target model (4.19 pp) or without RL supervision of any kind (5.08 pp). These results support a qualified internalized-search reading: under the recipe we test, much of the measured RL gain corresponds to a change in sampling efficiency toward operating points the base model can already reach under search. We treat the scaling patterns as descriptive of this recipe and cohort, report where they break down, and use UDF and BOPTR as behavioral diagnostics rather than evidence of parameter-level equivalence.

Chinese Translation

具有可验证奖励的强化学习（RLVR）提升了语言模型的推理能力，但这些提升与推理时解码和搜索之间的关系仍不明确。强化学习是否创造了基本模型所缺乏的推理能力，或者是否将回滚分布转向其已经可以达到但很少采样的轨迹？我们通过统一解码框架（Unified Decoding Framework, UDF）从行为上研究这一现象，该框架将令牌级采样、类似束搜索、树搜索和序列级重采样表示为在共享预算操作空间上的可执行策略，并通过 pass@$k$、自一致性、最佳 $N$ 和首次完成成功进行事后评分。利用来自 SimpleRL-Zoo 的成对基本/强化学习检查点，我们探讨 RL 默认策略曲线是否可以通过一条结构化的基本操作点路径进行近似。在 Math500、AIME、GPQA 和 IFEval 数据集上，pass@$k$ 恢复路径遵循预算操作点转移规则（Budgeted Operating-Point Transition Rule, BOPTR），$N_{ ext{Base}} ext{ 近似 } eta N_{ ext{RL}}^{eta}$，并具有基准条件的指数。在 Qwen2.5-7B 上，BOPTR 在我们测试的非神谕规则中给出了最低的转移误差，3.41 个百分点（95% 置信区间 [2.32, 5.53]）；三次种子复制给出 3.07 $ ext{±}$ 0.39 个百分点。该规则扩展到四个家族中的十个模型（在拟合后添加的检查点上为 3.28 到 4.87 个百分点），扩展到四个从未拟合的基准（在拟合中为 5.03 个百分点对比 4.44 个百分点），并且在目标模型没有 RL 检查点（4.19 个百分点）或没有任何形式的 RL 监督下（5.08 个百分点）仍然成立。这些结果支持了一种合格的内化搜索解读：在我们测试的配方下，测得的 RL 增益大部分对应于采样效率向基本模型在搜索中已经可以达到的操作点的变化。我们将这些缩放模式视为对该配方和群体的描述，报告它们的失效情况，并将 UDF 和 BOPTR 作为行为诊断，而不是参数级等价性的证据。

View on arXiv Download PDF AI Translation

cs.CL / 101 / 2609.01279

Some Emotions Run Deeper: Layer-wise Probing and Causal Intervention in Large Language Models

某些情感更为深刻：大语言模型中的层级探测与因果干预

Fang, Tian, Guibon, Gaël, Buscaldi, Davide

Abstract

Emotion is expressed in text along a wide spectrum, from surface lexical cues to inferences entangled with content. Most layer-wise analyses of emotion in LLMs use a single corpus, leaving open whether the depth at which emotion becomes accessible is a property of the model or also of the text source. We investigate this across three datasets spanning different degrees of explicitness and contextualization in emotion expression (Twitter posts, Reddit comments, and autobiographical narratives) and eight 1B--9B open-weight LLMs from the Llama, Qwen, and Granite families. We combine layer-wise probing with offline feature scaling and online forward interventions, transfer analyses, and an early-exit classifier. We find that (i) the best probing layer shifts systematically across corpora, from input-adjacent layers to over half model depth, and this ordering persists after matching label-by-length-bin distributions; (ii) across the evaluated settings, forward-pass interventions on probe-selected bands reduce test accuracy by 5--6 points more than same-width random bands ($q < 0.01$); (iii) selected bands transfer across datasets and emotion categories, suggesting partially shared affective information rather than strictly per-emotion substrates; and (iv) probe-selected early-exit representations outperform full-depth exits by $6.9$ percentage points on average.

Chinese Translation

情感在文本中以广泛的谱系表达，从表层的词汇线索到与内容交织的推断。大多数对大语言模型（LLMs）中情感的层级分析使用单一语料库，因此尚不清楚情感可被访问的深度是模型的特性还是文本来源的特性。我们在三个数据集上进行研究，这些数据集涵盖了不同程度的情感表达的显性和情境化（Twitter 帖子、Reddit 评论和自传叙述），以及来自 Llama、Qwen 和 Granite 家族的八个 1B-9B 开放权重的 LLMs。我们将层级探测与离线特征缩放、在线前向干预、转移分析和早期退出分类器相结合。我们发现：（i）最佳探测层在不同语料库之间系统性地变化，从输入相邻层到模型深度的一半以上，并且这种顺序在按标签-长度-箱分布匹配后仍然存在；（ii）在评估的设置中，针对探测选择的频带进行的前向干预比同宽随机频带减少测试准确率多 5-6 个百分点（$q < 0.01$）；（iii）选择的频带在数据集和情感类别之间转移，表明部分共享的情感信息，而非严格的情感基础；（iv）探测选择的早期退出表示在平均上比全深度退出高出 6.9 个百分点。

View on arXiv Download PDF AI Translation

cs.CL / 102 / 2609.01294

Explore Before Committing: Hypothesis-Guided Search for Deep Research Agents

探索后再决策：基于假设的深度研究代理搜索

Zhou, Ruochen, Chen, Zhengyu, Zhang, Luan, Gao, Siyang, Teh, Yee Whye, Chen, Shiqi

Abstract

Deep-research agents answer complex questions by interacting with search and browsing tools, yet they often search along a single evolving trajectory. Our trajectory-level analysis reveals a common failure mode in which the agent may encounter an early search state with several plausible directions, but follow one direction before collecting enough comparative evidence. Once this happens, subsequent tool calls tend to reinforce the same path, increasing the chance of failure when the initial direction is misleading. We further find that successful trajectories reduce this risk through two behaviors: grounding vague exploration in concrete candidates and shifting directions when the current path is weak or incomplete. Based on these findings, we propose HypoSearch, which generates lightweight hypotheses as soft search hints, explores them through bounded independent branches, and compares branch-level evidence before commitment. Across four deep-research benchmarks and three backbone models, HypoSearch consistently outperforms single-trajectory search and standard parallel baselines, improving Qwen3.5-122B from 46.7 to 60.0 on BC-small while using fewer tool calls than five independent trajectories. A pilot supervised fine-tuning study further shows that these behavioral signals can curate compact training trajectories and reduce degradation from unfiltered data.

Chinese Translation

深度研究代理通过与搜索和浏览工具的交互来回答复杂问题，但它们通常沿着单一的演变轨迹进行搜索。我们的轨迹级分析揭示了一种常见的失败模式，即代理可能在早期搜索状态中遇到多个合理的方向，但在收集足够的比较证据之前就选择了一个方向。一旦发生这种情况，后续的工具调用往往会强化同一路径，从而在初始方向误导时增加失败的可能性。我们进一步发现，成功的轨迹通过两种行为降低了这种风险：将模糊的探索基于具体候选者进行扎根，以及在当前路径薄弱或不完整时改变方向。基于这些发现，我们提出了HypoSearch，它生成轻量级的假设作为软搜索提示，通过有限的独立分支进行探索，并在决策前比较分支级证据。在四个深度研究基准和三个基础模型中，HypoSearch始终优于单轨迹搜索和标准并行基线，在BC-small上将Qwen3.5-122B的性能从46.7提升至60.0，同时使用的工具调用次数少于五个独立轨迹。一项初步的监督微调研究进一步表明，这些行为信号可以策划紧凑的训练轨迹，并减少来自未过滤数据的退化。

View on arXiv Download PDF AI Translation

cs.CL / 103 / 2609.01322

Exploring Sparse Autoencoders in Text-Based Causal Confounding Adjustment

探索文本基础因果混淆调整中的稀疏自编码器

Zhong, Mian, Keith, Katherine A., Field, Anjalie

Abstract

In many settings, studying causal questions based on text data requires adjusting for confounding information within texts. Yet there is a tradeoff in constructing text representations for adjustment: they must be sufficiently large and/or dense to preserve the confounding variables necessary for unbiased effect estimation, but sufficiently small and/or sparse to satisfy finite-sample overlap and yield low-variance estimates. To address this tradeoff, we turn to sparse autoencoders (SAEs), and propose a novel causal adjustment pipeline that iteratively selects a minimal set of SAE features via conditional independence tests. We find that SAE representations achieve better adjustments (lower bias and and higher coverage) than alternative representations in standard semi-synthetic evaluations with binary confounders, and their interpretability offers opportunities for falsification. We also introduce a more realistic semi-synthetic evaluation that uses multi-label data as the unobserved confounders and find off-the-shelf adjustment methods require increased investigation for these more complex settings. Code: https://github.com/mianzg/sae-text-confounder

Chinese Translation

在许多情况下，基于文本数据研究因果问题需要对文本中的混淆信息进行调整。然而，在构建用于调整的文本表示时存在权衡：这些表示必须足够大和/或密集，以保留必要的混淆变量，从而实现无偏效应估计，但同时又必须足够小和/或稀疏，以满足有限样本重叠并产生低方差估计。为了解决这一权衡，我们转向稀疏自编码器（Sparse Autoencoders, SAEs），并提出了一种新的因果调整流程，通过条件独立性测试迭代选择最小的SAE特征集。我们发现，在标准的半合成评估中，SAE表示在具有二元混淆变量的情况下实现了比替代表示更好的调整（更低的偏差和更高的覆盖率），且其可解释性为虚假检验提供了机会。我们还引入了一种更现实的半合成评估，使用多标签数据作为未观察到的混淆变量，并发现现成的调整方法在这些更复杂的环境中需要更多的研究。代码链接： https://github.com/mianzg/sae-text-confounder

View on arXiv Download PDF AI Translation

cs.CL / 104 / 2609.01325

VerTox: Verifiable Reward-Guided Corpus Poisoning Against Neural Ranking Models

VerTox：针对神经排名模型的可验证奖励引导语料中毒

Huang, Zhiqi, Datla, Vivek, Xu, Zhichao, Yu, Puxuan, Srikumar, Vivek, Samuel, Alfy

Abstract

Neural ranking models have become core components of modern information retrieval systems and important building blocks of AI systems such as retrieval-augmented generation (RAG) pipelines. However, their robustness remains insufficiently understood in the presence of large language models (LLMs), which can generate fluent and deceptive content at scale. This work investigates the vulnerability of neural ranking models to corpus poisoning attacks, in which an adversary injects a small number of maliciously crafted documents into the corpus to distort ranking behavior. We propose VerTox, the first framework to formulate corpus poisoning as a verifiable reward-guided reinforcement learning (RLVR) problem. By explicitly coupling ranking distortion with factual corruption through specialized reward shaping, we fine-tune compact LLMs into adversarial generators. Experiments demonstrate that our method achieves near-perfect attack success rates, producing adversarial documents that frequently rank higher than target documents across major neural ranking architectures, as well as a proprietary commercial embedding model. The generated adversarial documents are fluent and exhibit low perplexity, making them difficult to detect. Furthermore, by explicitly encouraging factual corruption, our adversarial documents significantly degrade the performance of a downstream RAG application.

Chinese Translation

神经排名模型已成为现代信息检索系统的核心组成部分，也是检索增强生成（RAG）管道等人工智能系统的重要构建模块。然而，在大型语言模型（LLMs）存在的情况下，它们的鲁棒性仍然不够清楚，这些模型能够大规模生成流畅且具误导性的内容。本研究探讨了神经排名模型在语料中毒攻击下的脆弱性，其中对手向语料中注入少量恶意构造的文档，以扭曲排名行为。我们提出了VerTox，这是第一个将语料中毒形式化为可验证的奖励引导强化学习（RLVR）问题的框架。通过通过专门的奖励塑造将排名扭曲与事实腐败明确耦合，我们将紧凑的LLM微调为对抗生成器。实验表明，我们的方法实现了近乎完美的攻击成功率，生成的对抗文档在主要神经排名架构以及一个专有商业嵌入模型中，常常排名高于目标文档。生成的对抗文档流畅且展现出低困惑度，使其难以被检测。此外，通过明确鼓励事实腐败，我们的对抗文档显著降低了下游RAG应用的性能。

View on arXiv Download PDF AI Translation

cs.CL / 105 / 2609.01341

Probing Factual Knowledge Transfer with Training Data Interventions

通过训练数据干预探测事实知识转移

Oji, Romina, Braun, Marc, Bollmann, Marcel, Kuhlmann, Marco, Kunz, Jenny

Abstract

Do multilingual language models transfer factual knowledge across languages during continued pretraining, or do they mostly recall facts learned directly from the target-language data? To answer this question more reliably, we propose an intervention-based framework: starting from an English-pretrained model, we continue pretraining on Persian data from which specific facts have been systematically removed at varying levels of granularity. We construct SIFT, a resource of 500 triples across 20 relations, stratified by the cultural origin of each fact's subject into general (globally prominent) and Persian-related entities, designed for both systematic fact removal from training data and evaluation, with natively written Persian cloze templates. Our results show that fact transfer is very limited: under the strictest removal condition, a large majority of English-acquired facts fail to transfer into Persian. We further show that sentence-level co-occurrence removal is insufficient to eliminate fact signal, and that easier (randomly selected) negative candidate sets substantially inflate apparent transfer by rewarding shallow associative heuristics, while performance on a harder candidate set that allows for less reliance on heuristics is much lower. Finally, we show that source-language entity frequency has a large influence, with Persian-related facts, which are orders of magnitude rarer in the English corpus, hardly transferring.

Chinese Translation

多语言语言模型在继续预训练过程中是否能够跨语言转移事实知识，还是主要依赖于从目标语言数据中直接学习的事实？为了更可靠地回答这个问题，我们提出了一种基于干预的框架：从一个经过英语预训练的模型开始，我们在波斯语数据上继续预训练，这些数据中系统性地移除了特定事实，移除的粒度各不相同。我们构建了SIFT，这是一个包含500个三元组和20种关系的资源，按照每个事实主题的文化来源分层为一般（全球显著）和波斯相关实体，旨在用于系统性地从训练数据中移除事实和评估，并配有原生编写的波斯语填空模板。我们的结果表明，事实转移非常有限：在最严格的移除条件下，大多数从英语获得的事实未能转移到波斯语中。我们进一步表明，句子级的共现移除不足以消除事实信号，而更简单（随机选择的）负候选集显著夸大了表面转移，因为它奖励了浅层的联想启发式，而在一个允许较少依赖启发式的更难候选集上的表现则低得多。最后，我们表明源语言实体的频率有很大影响，波斯相关的事实在英语语料库中稀有程度是数量级的差异，几乎无法转移。

View on arXiv Download PDF AI Translation

cs.CL / 106 / 2609.01352

CHARM: Character Hallucination for Multicultural Role Play Benchmark

CHARM：多文化角色扮演基准中的角色幻觉

Han, Sunkyung, Park, Nahyeon, Seo, Gaeun, Yoon, Seunghyun, Bak, JinYeong

Abstract

Role-playing large language models (LLMs) are expected to adopt a character's style while also respecting that character's knowledge boundaries. Prior evaluations detect character hallucination but rarely distinguish whether errors arise from failure to recognize a boundary or from failure to comply despite recognition. We introduce CHARM, a multicultural benchmark of 40 real and fictional characters drawn from five cultural-linguistic regions, and validated by native reviewers. It probes two boundary types, Temporal (historical vs. modern) and Cross-Universe (entities outside a character's narrative or historical universe), using abstention-enabled multiple-choice questions. We propose a two-stage evaluation that separates Boundary-Awareness (explicit recognition that a query is out of scope) from Boundary-Compliance (abstention when answering concrete questions). Evaluations across six LLMs show that hallucination is driven predominantly by compliance failures. Models frequently acknowledge that a query lies outside the character's knowledge yet still provide factual, out-of-character answers. By re-posing the same questions to the target character, we confirm that a large fraction of these cases are verified parametric overrides; the model stores the relevant fact but fails to suppress it. We also observe systematic cultural variation in these failures, consistent with imbalances in how characters from different regions are represented in model knowledge.

Chinese Translation

角色扮演的大型语言模型（LLMs）被期望能够采用角色的风格，同时尊重该角色的知识边界。之前的评估检测到了角色幻觉，但很少区分错误是由于未能识别边界还是尽管识别却未能遵守。我们引入了CHARM，这是一个涵盖来自五个文化语言区域的40个真实和虚构角色的多文化基准，并由母语评审员进行验证。该基准探讨了两种边界类型：时间性（历史与现代）和跨宇宙（角色叙事或历史宇宙之外的实体），使用允许弃权的多项选择题进行评估。我们提出了一个两阶段的评估方法，将边界意识（明确识别查询超出范围）与边界遵守（在回答具体问题时选择弃权）分开。对六个LLMs的评估显示，幻觉主要是由于遵守失败所驱动。模型经常承认查询超出了角色的知识范围，但仍然提供事实性的、超出角色的答案。通过将相同的问题重新提问给目标角色，我们确认这些案例中有很大一部分是经过验证的参数覆盖；模型存储了相关事实但未能抑制它。我们还观察到这些失败中存在系统性的文化差异，这与不同区域角色在模型知识中的表现不平衡是一致的。

View on arXiv Download PDF AI Translation

cs.CL / 107 / 2609.01354

Where the Verifier Fails: A Category-Level Audit of Reward Signals in RLVR

验证者失败的地方：强化学习可验证奖励中的奖励信号的类别级审计

Xin, Esther

Abstract

Reinforcement learning with verifiable rewards (RLVR) and standard benchmark evaluation both rely on an automatic verifier that turns a free text answer into a binary reward. Prior work reports that one evaluation harness accepts only about 94% of its own ground truth answers, blaming LaTeX parsing. That is an aggregate: it does not say which answer forms consume the error budget. We supply the decomposition. We apply metamorphic testing to the verifier rather than the model, generating certified equivalent answer variants, that is, rewrites that preserve mathematical meaning by construction, so that any rejection is a provable false negative needing no human adjudication. We then measure rejection per answer category across four widely used verifiers over 307,420 verdicts. We find three things. (1) Self validation ranges from 53.8% to 95.2% on identical inputs, a spread of 41.3 points. The published figure describes one implementation, not the task; two configurations of the same library disagree on 49.9% of pairs. (2) The residual is not spread across parsing categories but concentrated in whitespace and punctuation, which account for 93.0% of in contract failures for the default LaTeX configuration. A trailing period or newline dominates the budget. (3) Separating rejection from execution failure shows that verifiers with similar aggregate error fail for opposite reasons, and that a reference numeric cascade accepts off by one wrong answers as a step function of magnitude, from 0% below 10^4 to 100% at or above, because its relative tolerance is scale invariant.

Chinese Translation

可验证奖励的强化学习（RLVR）和标准基准评估都依赖于一个自动验证器，该验证器将自由文本答案转换为二元奖励。先前的研究报告称，一个评估工具仅接受其自身约94%的真实答案，原因归咎于LaTeX解析。这是一个总的统计数据，并未说明哪些答案形式消耗了错误预算。我们提供了这一分解。我们将变形测试应用于验证器而非模型，生成经过认证的等效答案变体，即通过构造保留数学意义的重写，因此任何拒绝都是可证明的假阴性，无需人工裁决。然后，我们在307,420个裁决中测量了四个广泛使用的验证器中每个答案类别的拒绝情况。我们发现三点：(1) 自我验证在相同输入下的范围为53.8%到95.2%，差异为41.3个百分点。已发布的数字描述的是一种实现，而非任务；同一库的两种配置在49.9%的对中存在不一致。(2) 残差并未分布在解析类别中，而是集中在空白和标点符号上，这占据了默认LaTeX配置下93.0%的合同失败。一个尾随的句号或换行符主导了预算。(3) 将拒绝与执行失败分开显示，具有相似总错误的验证器因相反原因失败，而参考数字级联接受偏差为1的错误答案作为幅度的阶跃函数，从低于10^4的0%到在或以上的100%，因为其相对容忍度是尺度不变的。

View on arXiv Download PDF AI Translation

cs.CL / 108 / 2609.01356

Separating Syntax from Language: A Mechanistic Account of Translation in Multilingual LLMs

将语法与语言分离：多语言大型语言模型中翻译的机制性解释

Sonkin, Mikhail, Baeumel, Tanja, Gurgurov, Daniil, van Genabith, Josef, Ostermann, Simon

Abstract

Multilingual large language models (mLLMs) achieve strong performance in machine translation, yet our understanding of the mechanisms by which they transform representations from one language to another remains incomplete. Prior work suggests that translation decomposes into separable processes within an mLLM, where conceptual content is first represented independently, followed by a production into language-specific form. In this work, we show that translation is even more modular than previously assumed and that the output language production in translation processes is actually further separable into a syntax and a surface language process. We construct controlled multilingual datasets that isolate cross-linguistic differences in word-order and use causal interventions and probing to track how representations are transformed during translation. We find that models first construct target-side word-order before realizing the target language surface form. We identify individual attention heads that are selectively sensitive to syntactic transformations while remaining largely invariant to language identity. These results establish the commitment to a syntactic structure as an independent stage in translation, extending prior decompositions and showing how translation is implemented by functionally different components within mLLMs.

Chinese Translation

多语言大型语言模型（mLLMs）在机器翻译中表现出色，但我们对它们如何将一种语言的表征转化为另一种语言的机制的理解仍然不完整。先前的研究表明，翻译在mLLM中分解为可分离的过程，其中概念内容首先独立表示，然后转化为特定语言的形式。在本研究中，我们展示了翻译的模块化程度超出了先前的假设，翻译过程中的输出语言生成实际上可以进一步分为语法和表层语言过程。我们构建了受控的多语言数据集，以隔离词序的跨语言差异，并使用因果干预和探测技术跟踪翻译过程中表征的转化。我们发现模型首先构建目标语言的词序，然后再实现目标语言的表层形式。我们识别出对语法转化具有选择性敏感性的单个注意力头，同时对语言身份保持相对不变。这些结果确立了语法结构作为翻译中的独立阶段的承诺，扩展了先前的分解，并展示了翻译如何通过mLLMs中功能不同的组件实现。

View on arXiv Download PDF AI Translation

cs.CL / 109 / 2609.01361

Investigating Linear Probe Robustness to Linguistic Register, Medical Specialty, and Corpus Shifts in Medical QA

探讨线性探针对语言风格、医学专业和语料库变化的鲁棒性在医学问答中的应用

Mishra, Nishant, Abu-Hanna, Ameen, Calixto, Iacer

Abstract

Linear classifiers trained on hidden states of a large language model (LLM), linear probes, can flag factual errors from a single forward pass. Geometrically, that implies that true and false statements separate along a stable direction in hidden state space, i.e., the truth direction. Prior work disagrees on whether this generalises across input shifts, but the disagreement is hard to interpret because cross-dataset probe transfer experiments confound several kinds of input change at once. We isolate three such variables in medical question-answering (QA): writing style (register), domain (medical specialty), and corpus (dataset). We build a benchmark using 500 MedQA entries, each rewritten into four styles (textbook, patient, clinical note, colloquial), annotated with clinical specialty, and grouped with two other exam corpora, MedMCQA and MMLU-medical, for cross-dataset evaluation. Probing four open-weight LLMs (2--8B), we find that the truth direction is largely robust to writing style (mean $\Delta_\text{register} \approx 0.10$ AUROC on held-out facts) and to medical specialty ($\Delta_\text{specialty} \approx 0.03$), but degrades unevenly across corpora: by $0.12$ AUROC on MMLU-medical and by $0.21$ on MedMCQA, roughly twice the register gap. The register result replicates with a second generator and carries over to human-written patient questions. The truth direction is therefore largely stable within the medical domain but breaks under some corpus shifts, and question format does not explain the break, which suggests that the signal a linear probe recovers is partly bound to dataset structure rather than to medical knowledge alone.

Chinese Translation

基于大型语言模型（LLM）隐藏状态训练的线性分类器（线性探针）能够通过一次前向传播标记事实错误。从几何角度来看，这意味着真实和虚假的陈述在隐藏状态空间中沿着一个稳定的方向分离，即真理方向。先前的研究对这一现象是否能够推广到输入变化上存在分歧，但由于跨数据集探针迁移实验同时混合了多种输入变化，这种分歧难以解释。我们在医学问答（QA）中隔离了三个变量：写作风格（语言风格）、领域（医学专业）和语料库（数据集）。我们使用500个MedQA条目构建了一个基准，每个条目被重写为四种风格（教科书、患者、临床记录、口语），并标注了临床专业，同时与另外两个考试语料库MedMCQA和MMLU-medical分组，以进行跨数据集评估。在对四个开放权重的LLM（2-8B）进行探测时，我们发现真理方向在写作风格上大体上是鲁棒的（在保留事实上的平均$ ext{Δ}_{ ext{register}} ext{约} 0.10$ AUROC）以及在医学专业上（$ ext{Δ}_{ ext{specialty}} ext{约} 0.03$），但在不同语料库间的表现不均匀下降：在MMLU-medical上下降了$0.12$ AUROC，在MedMCQA上下降了$0.21$，大约是语言风格差距的两倍。使用第二个生成器的结果重复，并且延续到人类撰写的患者问题上。因此，真理方向在医学领域内大体上是稳定的，但在某些语料库变化下会出现破裂，而问题格式并不能解释这一破裂，这表明线性探针恢复的信号部分与数据集结构相关，而不仅仅是医学知识。

View on arXiv Download PDF AI Translation

cs.CL / 110 / 2609.01369

How Correct Is Your Answer? A Semantic Correctness Framework for Open QA Evaluation

你的答案有多正确？开放式问答评估的语义正确性框架

Yotkova, Elitsa, Kastreva, Violeta, Velkov, Petar, Boyanov, Hristo, Dimitrov, Dimitar, Koychev, Ivan, Nakov, Preslav

Abstract

Reliable evaluation of open-ended question answering remains a bottleneck for measuring answer correctness of modern LLMs. Unlike multiple-choice tasks, free-form answers may be correct in many surface forms and may fail in qualitatively different ways, including incompleteness, contradiction, overgeneration, and endorsement of false premises. Existing judgment-based and similarity-based metrics often collapse these distinctions. We address this gap with three reusable contributions. First, we introduce a semantic correctness taxonomy that assigns open-ended answers to eight ordered classes, separating verbose-but-correct answers from those contaminated by hallucinated content. Second, we release CAP-Correctness, an 8.8k-example benchmark spanning widely used QA datasets, and CAP-Statements, an 11k-example dataset for converting question-answer pairs into declarative statements for natural language inference (NLI) training and statement-based evaluation. Third, we introduce CAP (Context-Aware Precision), a reference-based metric that scores question-conditioned statements using bidirectional NLI. Under a monotonicity protocol testing whether metrics respect the taxonomy's intended ordering, CAP outperforms established baselines.

Chinese Translation

开放式问答的可靠评估仍然是衡量现代大语言模型（LLMs）答案正确性的瓶颈。与多项选择任务不同，自由形式的答案可能以多种表面形式正确，并可能在质上以不同方式失败，包括不完整性、矛盾、过度生成和支持错误前提。现有的基于判断和相似性的度量往往模糊了这些区别。我们通过三个可重复使用的贡献来填补这一空白。首先，我们引入了一种语义正确性分类法，将开放式答案分为八个有序类别，区分冗长但正确的答案与被虚假内容污染的答案。其次，我们发布了CAP-Correctness，一个包含8,800个示例的基准，涵盖广泛使用的问答数据集，以及CAP-Statements，一个包含11,000个示例的数据集，用于将问答对转换为用于自然语言推理（NLI）训练和基于声明的评估的陈述。第三，我们引入了CAP（上下文感知精度），一种基于参考的度量，使用双向NLI对问题条件的陈述进行评分。在测试度量是否遵循分类法预期排序的单调性协议下，CAP的表现优于现有基线。

View on arXiv Download PDF AI Translation

cs.CL / 111 / 2609.01374

Behaviorally Effective LoRA Writes Are Sparse and Structured

行为有效的 LoRA 写入是稀疏且结构化的

Sato, Haruto, Tanaka, Yuki, Nakamura, Ren, Kobayashi, Aoi, Ito, Mei

Abstract

Low-rank adaptation fixes the rank of the update, but it does not identify which parts of a trained write actually carry behavior. We study that question directly and show that behaviorally effective LoRA writes are sparse, structured, and far more concentrated than the raw low-rank parameterization suggests. We use Learned-Basis LoRA, a learned-basis continuation recipe, to expose that structure. The recipe warms up an unconstrained adapter, converts its learned write columns into a module-wise orthonormal basis, freezes that basis, and continues training inside the constrained parameterization. Across 14 exact switches from unconstrained to constrained form, held-out accuracy is unchanged at the conversion step and reconstructed write matrices differ by at most 0.25% relative Frobenius error. Same-state continuation then shows that the same trained checkpoint develops differently under different write subspaces, establishing write geometry as a causal state variable. A no-retraining projection test shows that useful write signal stays inside the learned write space and largely disappears from random or frozen-activation PCA controls. The concentration pattern is strong at both local and global scales. Across GSM8K, MathQA, and AQuA, per-module top-k continuation reaches its optimum at k in {2, 4} in all twelve seed-level cases we test. A stricter global ranking test shows that learned top-16 and top-32 subsets outperform matched random subsets, especially on GSM8K/Qwen and MathQA/Qwen. Single-direction ablations further reveal a sparse set of late q_proj, o_proj, and down_proj components with outsized behavioral impact.

Chinese Translation

低秩适应固定了更新的秩，但并未识别出训练写入中哪些部分实际承载了行为。我们直接研究了这个问题，并表明行为有效的 LoRA 写入是稀疏的、结构化的，并且比原始低秩参数化所暗示的要集中得多。我们使用学习基础 LoRA（Learned-Basis LoRA），一种学习基础的延续方法，来揭示这种结构。该方法首先对一个不受限制的适配器进行预热，将其学习到的写入列转换为模块级的正交基，随后冻结该基，并在受限参数化内继续训练。在14个从不受限制到受限制形式的精确转换中，转换步骤的保留准确率保持不变，重构的写入矩阵的相对 Frobenius 误差最多为 0.25%。相同状态的延续进一步表明，同一训练检查点在不同的写入子空间下发展不同，确立了写入几何作为因果状态变量。无重训练投影测试表明，有用的写入信号保持在学习到的写入空间内，并在随机或冻结激活的 PCA 控制中大幅消失。集中模式在局部和全局尺度上都很强。在 GSM8K、MathQA 和 AQuA 数据集上，每模块的 top-k 延续在我们测试的所有十二个种子级案例中，在 k 为 {2, 4} 时达到最佳。更严格的全局排名测试显示，学习到的 top-16 和 top-32 子集的表现优于匹配的随机子集，尤其是在 GSM8K/Qwen 和 MathQA/Qwen 上。单向消融进一步揭示了一组稀疏的晚期 q_proj、o_proj 和 down_proj 组件，其对行为的影响显著。

View on arXiv Download PDF AI Translation

cs.CL / 112 / 2609.01379

Polish ModernBERT: The Long and Short of Polish Language Understanding

波兰现代BERT：波兰语言理解的长短期考量

Perełkiewicz, Michał, Dadas, Sławomir, Poświata, Rafał, Grębowiec, Małgorzata

Abstract

Encoder-only Transformers remain effective for discriminative and representation-learning tasks, yet Polish encoders still largely rely on BERT/RoBERTa-style architectures. We introduce \textbf{Polish ModernBERT}, a family of four Polish encoders available at Base and Large scales, each with 512-token and 8K context variants. We adapt the ModernBERT pretraining recipe through staged selection experiments and release a long-context benchmark covering legal topic classification, ideological decision-direction prediction, factual-consistency assessment over literary plot summaries, and human-rights violation assessment. Across 30 tasks, Polish ModernBERT achieves the best overall performance among the evaluated Polish encoders, reaching 83.99 and 85.11 for the Base-8K and Large-8K models, respectively. On long-context tasks, the 8K variants improve over matched Polish RoBERTa-8K baselines from 67.47 to 77.15 and from 75.88 to 78.49 at the Base and Large scales, respectively. The Base-8K model achieves this gain with 22\% fewer parameters (149M vs.\ 190M). Efficiency measurements in representative inference setups show lower peak memory usage and latency than matched Polish RoBERTa baselines in both 512-token and 8K settings. Polish ModernBERT-8K-Base additionally achieves the best result on a Polish retrieval benchmark among the evaluated encoders below 300M parameters.

Chinese Translation

仅编码器的变换器在区分性和表示学习任务中仍然有效，但波兰编码器仍在很大程度上依赖于BERT/RoBERTa风格的架构。我们介绍了 extbf{波兰现代BERT}，这是一个包含四个波兰编码器的系列，提供基础（Base）和大型（Large）两种规模，每种规模均有512-token和8K上下文的变体。我们通过分阶段选择实验调整了ModernBERT的预训练方案，并发布了一个涵盖法律主题分类、意识形态决策方向预测、文学情节摘要的事实一致性评估以及人权侵犯评估的长上下文基准。在30个任务中，波兰现代BERT在评估的波兰编码器中实现了最佳整体性能，基础-8K和大型-8K模型分别达到了83.99和85.11。在长上下文任务中，8K变体在基础和大型规模上分别将匹配的波兰RoBERTa-8K基线从67.47提高到77.15，从75.88提高到78.49。基础-8K模型在参数数量上减少了22 ext{%}（149M对比190M）而实现了这一增益。在代表性推理设置中的效率测量显示，在512-token和8K设置下，其峰值内存使用和延迟均低于匹配的波兰RoBERTa基线。波兰现代BERT-8K-Base在评估的参数少于300M的编码器中，还在波兰检索基准上取得了最佳结果。

View on arXiv Download PDF AI Translation

cs.CL / 113 / 2609.01383

InSight: A Benchmark for Agentic Claim Verification in Interactive Visualizations

InSight：交互式可视化中代理性主张验证的基准

Hutchinson, Maeve, Huq, Syed Mahbubul, Albinhassan, Mohammad, Jianu, Radu, Slingsby, Aidan, Madhyastha, Pranava

Abstract

Vision Language Models have demonstrated remarkable proficiency in interpreting static visual artifacts, but modern data analysis is inherently dynamic, requiring the active interrogation of interactive environments. Existing benchmarks are predominantly constrained to static imagery and one-shot question answering and fail to capture the epistemic demands of this domain, where evidence is frequently occluded, distributed across linked views, or conditionally revealed through user agency. In this paper, we introduce InSight, a benchmark for agentic claim verification over interactive visualizations. The dataset consists of 21,349 claims derived from human-authored analytical narratives and grounded in fully interactive web-based environments. Agents must navigate these environments to determine whether a natural language claim is supported, refuted or not verifiable given the available evidence. Unlike traditional evaluations, InSight treats interaction traces as intrinsic proxies for reasoning, enabling a rigorous audit of how models seek and synthesize visual evidence. We evaluate state-of-the-art models, revealing that interactive verification remains a non-trivial challenge. We release InSight at https://github.com/maevehutch/insight.

Chinese Translation

视觉语言模型在解释静态视觉文物方面表现出显著的能力，但现代数据分析本质上是动态的，要求对交互环境进行主动探究。现有的基准主要局限于静态图像和一次性问答，未能捕捉该领域的认识论需求，在该领域中，证据常常被遮蔽、分布在关联视图中，或通过用户的主动性有条件地揭示。本文介绍了InSight，一个用于交互式可视化中代理性主张验证的基准。该数据集由21,349个主张组成，这些主张源自人类撰写的分析叙述，并基于完全交互的网络环境。代理必须在这些环境中导航，以确定自然语言主张是否得到支持、被反驳或在可用证据下不可验证。与传统评估不同，InSight将交互轨迹视为推理的内在代理，从而能够严格审计模型如何寻求和综合视觉证据。我们评估了最先进的模型，揭示了交互验证仍然是一个非平凡的挑战。我们在https://github.com/maevehutch/insight发布了InSight。

View on arXiv Download PDF AI Translation

cs.CL / 114 / 2609.01386

When Tokenization is Secretly Output Supervision

当标记化秘密地输出监督

Baeumel, Tanja, van Genabith, Josef, Ostermann, Simon

Abstract

Tokenization in language models is treated by default as an input preprocessing decision. We argue that this framing is incomplete: in autoregressive models, tokenizer granularity determines what the model must resolve in a single forward pass, and therefore the supervision signal it receives. This affects both the difficulty of the learning problem and the representations that emerge inside the model. We test this in a controlled experiment on numeric reasoning with a novel decoupling of input and output tokenization. As the output supervision view predicts, differences in task performance, training dynamics, and model internals are induced by output tokenization and largely invariant to input tokenization. This may matter in practice, because models with different tokenization strategies differ not only in input representation but in the task they were trained on. Comparisons between models may thus partly reflect task definition rather than ability. A survey of 120 recent *CL papers on numeric reasoning confirms that this is rarely acknowledged: only about 10% report the numeric tokenization of the models they evaluate, while 69% compare across tokenization, and thus supervision, regimes without reporting it. While prior work documents that tokenization consistently affects model performance, there is no principled account of why. We argue that framing tokenization as output supervision provides that account.

Chinese Translation

在语言模型中，标记化通常被视为输入预处理的决策。我们认为这种框架是不完整的：在自回归模型中，标记器的粒度决定了模型在单次前向传播中必须解决的问题，因此也决定了它所接收的监督信号。这影响了学习问题的难度以及模型内部所产生的表示。我们通过对数字推理进行的受控实验来测试这一点，采用了一种输入和输出标记化的新解耦方式。正如输出监督视角所预测的，任务表现、训练动态和模型内部的差异是由输出标记化引起的，并且在很大程度上与输入标记化无关。这在实践中可能是重要的，因为采用不同标记化策略的模型不仅在输入表示上有所不同，而且在其训练的任务上也有所不同。因此，模型之间的比较可能部分反映了任务定义而非能力。一项对120篇近期关于数字推理的*CL论文的调查确认，这一点很少被承认：只有约10%的论文报告了他们评估的模型的数字标记化，而69%的论文在没有报告的情况下比较了不同的标记化及其监督机制。尽管先前的研究记录了标记化始终影响模型性能，但尚无原则性的解释。我们认为，将标记化视为输出监督提供了这一解释。

View on arXiv Download PDF AI Translation

cs.CL / 115 / 2609.01422

From Rollouts to Recipes: Self-Contained Post-Training for LLMs

从回滚到配方：大语言模型的自包含后训练

Li, Yifei, Zhang, Lingling, Huang, Muye, Ma, Zihan, Liu, Jiashuai, Liu, Jun

Abstract

Post-training large language models usually applies a single training recipe to all samples, even though the model's own rollouts reveal different sample-level learning states. We propose Self-Routing, a behavior-conditioned post-training framework that uses rollout correctness and confidence to decide how each sample should be optimized. Depending on its behavior state, a sample is routed to GRPO, on-policy self-distillation, regularization, or skipping, allowing training to adapt without external teachers, extra annotations, or additional sampling. Experiments on mathematical reasoning across Qwen3 and Qwen3.5 backbones show that Self-Routing consistently improves over uniform GRPO, uniform OPSD, fixed mixtures, and simpler routing baselines. Further analyses show that the routing distribution changes over training and reduces unnecessary updates on low-signal or already stable samples.

Chinese Translation

后训练大语言模型通常对所有样本应用单一的训练配方，尽管模型自身的回滚揭示了不同样本级别的学习状态。我们提出了自路由（Self-Routing），一种基于行为的后训练框架，利用回滚的正确性和置信度来决定每个样本应如何优化。根据其行为状态，样本被路由到GRPO（基于策略的自蒸馏）、在线自蒸馏、正则化或跳过，从而允许训练在没有外部教师、额外注释或额外采样的情况下进行自适应。针对Qwen3和Qwen3.5骨干网络的数学推理实验表明，自路由在均匀GRPO、均匀在线自蒸馏、固定混合和更简单的路由基线之上始终表现出改进。进一步分析表明，路由分布在训练过程中发生变化，并减少了对低信号或已稳定样本的不必要更新。

View on arXiv Download PDF AI Translation

cs.CL / 116 / 2609.01491

GlossoGen: Emergent Language in Complex Multi-Agent LLM Interactions

GlossoGen：复杂多智能体大语言模型交互中的自发语言

Stengel-Eskin, Elias, Sander, Newton, Bonetti, Carlos, Boguraev, Sasha, Bowler, James, Sirin, Hale, Kirby, Simon

Abstract

The growing rate at which LLM agents interact with one another raises key questions about language evolution in multi-LLM-agent settings, with implications for safety and monitorability as well as for linguistic accounts of LLMs. To address these questions, we introduce GlossoGen, a novel platform for studying multi-agent language evolution in complex scenarios. Within GlossoGen, we build the SaveVeyru scenario, which requires agents with partial information to communicate under pressure. We find that language evolution does occur between LLM agents, that the resulting languages are compositional and morphologically productive, and that they deviate from the LLMs' English prior in ways that render them incomprehensible to humans. Moreover, we identify several qualities essential to this evolution: pressure towards efficiency; the strength of the models backing the agents; and access to a "postmortem" stage in which agents can agree on linguistic conventions. Importantly, we observe that different conditions govern the transmission of language to new agents. Specifically, we find that agents learn new languages from usage alone, take an active role in this learning, and that while stronger models are required for novel language emergence, weaker models can learn an existing language once it has emerged. Taken together, our results indicate that current LLMs have the potential for cumulative cultural evolution -- previously attested only in humans -- with mixed populations of agents developing capacities that go beyond their lowest common denominator.

Chinese Translation

大语言模型（LLM）智能体之间互动的增长速度引发了关于多LLM智能体环境中语言演化的关键问题，这些问题对安全性、可监控性以及LLM的语言学解释都有重要影响。为了解决这些问题，我们引入了GlossoGen，一个用于研究复杂场景中多智能体语言演化的新平台。在GlossoGen中，我们构建了SaveVeyru场景，该场景要求拥有部分信息的智能体在压力下进行沟通。我们的研究发现，LLM智能体之间确实发生了语言演化，所产生的语言具有组合性和形态生产性，并且在某些方面偏离了LLM的英语先验，使其对人类而言难以理解。此外，我们确定了几种对这种演化至关重要的特质：对效率的压力；支持智能体的模型的强度；以及在“事后”阶段，智能体可以就语言习惯达成一致。重要的是，我们观察到不同的条件支配着语言向新智能体的传递。具体而言，我们发现智能体仅通过使用就能学习新语言，并在这一学习过程中发挥积极作用；同时，虽然新语言的出现需要更强的模型，但一旦新语言出现，较弱的模型也能够学习已有语言。综合来看，我们的结果表明，当前的LLM具备累积文化演化的潜力——这一现象此前仅在人类中得到证实——混合智能体群体发展出超越其最低共同分母的能力。

View on arXiv Download PDF AI Translation

cs.CL / 117 / 2609.01532

Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall

中期训练中的知识蒸馏有利于推理而非事实回忆

He, Jacqueline, Yen, Howard, Li, Shuyue Stella, Li, Margaret, Zeng, Hanqing, Xia, Yinglong, Zhang, Benyu, Zhao, Zhuokai, Zhang, Qiang, Koh, Pang Wei, Zettlemoyer, Luke, Yih, Wen-tau

Abstract

Logit-based knowledge distillation (KD) is used to train smaller language models (LMs) via supervision from stronger teachers, but whether its benefits are consistent across training stages remains unclear. Through controlled experiments, we find that forward Kullback-Leibler (KL) distillation--the standard KD formulation--with post-trained teachers behaves fundamentally differently during mid-training, an intermediate phase of self-supervised learning on curated corpora. Surprisingly, while forward KD simultaneously improves reasoning and factual recall during pre-training relative to standard next-token prediction (NTP), it instead slows factual recall acquisition during mid-training despite continued reasoning gains. We trace this stage dependence to an asymmetry in teacher confidence across data domains and the student's evolving knowledge state: teachers are more confident on procedural than knowledge-intensive data, while students acquire low-entropy factual knowledge earlier in training. To mitigate this imbalance, we propose Switch Distillation, a simple mid-training objective that distills on tokens where the teacher is confident, using teacher predictive entropy as a lightweight routing signal, and otherwise falls back to cross-entropy. Switch Distillation consistently outperforms existing distillation objectives across teacher sizes. Relative to standard NTP, it achieves 1.61-1.71x the reasoning performance and 1.13-1.19x the knowledge and commonsense performance while preserving 96.7-96.8% of factual recall. Crucially, these benefits persist after post-training: Switch Distillation closes the factual recall gap while maintaining 1.25-1.32x and 1.13-1.20x gains in reasoning and knowledge and commonsense, respectively.

Chinese Translation

基于Logit的知识蒸馏（KD）用于通过强教师的监督训练较小的语言模型（LM），但其在训练阶段的效益是否一致仍然不清楚。通过控制实验，我们发现前向Kullback-Leibler（KL）蒸馏——标准的KD公式——在中期训练阶段表现出根本不同的行为，这是自监督学习在精心策划的语料库上的一个中间阶段。令人惊讶的是，尽管前向KD在预训练阶段相对于标准的下一个标记预测（NTP）同时提高了推理和事实回忆，但在中期训练阶段却减缓了事实回忆的获取，尽管推理的提升仍在继续。我们将这种阶段依赖性追溯到教师在数据领域中的信心不对称以及学生不断发展的知识状态：教师在程序性数据上的信心高于知识密集型数据，而学生在训练早期获得低熵的事实知识。为了缓解这种不平衡，我们提出了Switch Distillation，这是一种简单的中期训练目标，专注于教师有信心的标记进行蒸馏，使用教师预测熵作为轻量级路由信号，否则退回到交叉熵。Switch Distillation在不同教师规模下始终优于现有的蒸馏目标。相对于标准NTP，它在推理性能上达到了1.61-1.71倍，在知识和常识性能上达到了1.13-1.19倍，同时保持了96.7-96.8%的事实回忆。关键是，这些好处在后训练后依然存在：Switch Distillation缩小了事实回忆的差距，同时在推理和知识与常识上分别保持了1.25-1.32倍和1.13-1.20倍的提升。

View on arXiv Download PDF AI Translation

cs.CL / 118 / 2609.01548

SDARE-Bench: Evaluating Large Language Models on Conversational Stigma Detection and Response in Dyadic and Group Dialogue

SDARE-Bench：在双人和群体对话中评估大型语言模型的对话污名检测与响应

Fong, Stephanie, Jiang, Yiwen, Wang, Zimu, Yang, Hongxi, Shen, Yaling, Chow, Hiu Weh Naomi, Lai, Heung Ying, Zhao, Xiangyu, Xu, Qingyang, Xu, Zhongxing, Liu, Jiahe, Oliveira, Guilherme C., Lee, Vincent, Ge, Zongyuan, Dwyer, Dominic

Abstract

Large Language Models (LLMs) are increasingly used in advice seeking and decision making that may affect social judgements. Despite stigma's profound effects on people and communities, benchmarks remain scarce. Existing general-domain evaluations typically rely on static prompts and fixed-format tasks, overlooking conversational contexts and audience effects in everyday communication. To address these gaps, we introduce SDARE-Bench, the first scenario-based benchmark evaluating both stigma detection and open-ended response generation in LLMs, comprising 1,138 dyadic queries and 1,388 group dialogue. Empirical results across 8 LLMs consistently demonstrate poor identification of stigma components, especially in group dialogues. In open-ended response generation, stigma expression was substantially higher in group settings than in dyadic, with weaker resistance to stigma and more unrealistic advice. Responses were evaluated using a classifier trained on 1,392 human annotated responses. In constructed group pressure settings, stigma expression rates further increased to a striking average of 97.5%. Our findings identify stigma response as a recurring LLM safety vulnerability, especially in socially complex conversational contexts.

Chinese Translation

大型语言模型（LLMs）在寻求建议和决策中越来越多地被使用，这可能影响社会判断。尽管污名对个人和社区有深远的影响，但相关基准仍然稀缺。现有的一般领域评估通常依赖于静态提示和固定格式任务，忽视了日常交流中的对话上下文和受众效应。为了解决这些问题，我们引入了SDARE-Bench，这是第一个基于场景的基准，评估LLMs中的污名检测和开放式响应生成，包含1,138个双人查询和1,388个群体对话。对8个LLMs的实证结果一致表明，污名成分的识别能力较差，尤其是在群体对话中。在开放式响应生成中，群体环境中的污名表达显著高于双人环境，且对污名的抵抗力较弱，提供了更多不切实际的建议。响应使用一个在1,392个人工标注响应上训练的分类器进行评估。在构建的群体压力环境中，污名表达率进一步上升至惊人的平均值97.5%。我们的研究发现，污名响应是LLMs在社会复杂对话上下文中反复出现的安全脆弱性。

View on arXiv Download PDF AI Translation

cs.CL / 119 / 2609.01563

A systematic Approach to constructing a Chance-and-Risk Matrix for Semiconductor Supply Chains

构建半导体供应链机会与风险矩阵的系统方法

Salkić, Ema, Fichtl, Alexander, Ulrich, Philipp, Ehm, Hans, Bonik, Marta, Groh, Georg

Abstract

Semiconductor supply chains face escalating risks from geopolitical tensions, geographic concentration, and rapid technological shifts, yet no scalable system continuously extracts, structures, and prioritizes risk intelligence from public corporate disclosures. We present an end-to-end pipeline that retrieves corporate documents for semiconductor companies and uses large language models (LLMs) to extract the risks and opportunities they describe. It organizes these into a knowledge graph linking each item to its category, sources, and related events, then merges duplicates and ranks them with a three-layer mechanism combining an algorithmic formula, an LLM relevance adjustment, and expert validation. Applied to five companies across the value chain, the pipeline produces 76,207 scored items, of which an independent check finds 92.6% valid. The automated rankings match expert judgment at an average Spearman correlation of 0.55 for risks and 0.72 for opportunities, and the resulting matrices identify trade restrictions as the dominant cross-company risk.

Chinese Translation

半导体供应链面临来自地缘政治紧张局势、地理集中度和快速技术变革的日益增加的风险，但尚无可扩展的系统能够持续提取、结构化和优先排序来自公共企业披露的风险情报。我们提出了一种端到端的流程，检索半导体公司的企业文件，并利用大型语言模型（LLMs）提取其中描述的风险和机会。该流程将这些信息组织成一个知识图谱，将每个项目链接到其类别、来源和相关事件，然后合并重复项，并通过结合算法公式、LLM相关性调整和专家验证的三层机制对其进行排名。应用于价值链上的五家公司，该流程生成了76,207个评分项，独立检查发现其中92.6%有效。自动排名与专家判断的平均斯皮尔曼相关系数为0.55（风险）和0.72（机会），最终生成的矩阵识别出贸易限制作为跨公司风险的主要因素。

View on arXiv Download PDF AI Translation

cs.CL / 120 / 2609.01564

From Confusion to Clarity: Confusion-Aware Retrieval and Knowledge Injection for Text Classification

从混淆到清晰：基于混淆感知的检索与知识注入用于文本分类

Gupta, Manish, Giri, Chaitanya, Talur, Jayasimha

Abstract

Large language models (LLMs) struggle to classify text into taxonomies with many semantically similar labels, as the distinctions are domain-specific and not captured by pre-training. To handle large label spaces, a common approach retrieves top-$K$ candidate labels by embedding similarity and prompt the LLM to choose among them. However, top-$K$ retrieval reduces the number of candidates but does not help the model tell similar ones apart. When two similar labels both appear as candidates, the model lacks the signal to choose correctly between them. We propose a framework that (1) identifies which label pairs the model struggles to distinguish, (2) expands the candidate set to include confusable labels, and (3) generates targeted rules to differentiate between similar candidates. The framework requires no fine-tuning, and the generated rules transfer to smaller, cheaper models. On three benchmarks (WOS, Flipkart, LEDGAR), our approach improves Macro F1 by up to 10.0pp over retrieval baselines, with smaller models (2B--20B) gaining up to 11.5pp via cross-model transfer.

Chinese Translation

大型语言模型（LLMs）在将文本分类到具有许多语义相似标签的分类体系中时面临困难，因为这些区分是特定于领域的，并未在预训练中得到体现。为了处理大型标签空间，一种常见的方法是通过嵌入相似性检索前$K$个候选标签，并提示LLM在其中进行选择。然而，前$K$检索虽然减少了候选数量，却并未帮助模型区分相似标签。当两个相似标签均作为候选出现时，模型缺乏信号来正确选择其中一个。我们提出了一个框架，该框架（1）识别模型难以区分的标签对，（2）扩展候选集以包括混淆标签，以及（3）生成有针对性的规则以区分相似候选。该框架不需要微调，生成的规则可以迁移到更小、更便宜的模型。在三个基准测试（WOS、Flipkart、LEDGAR）上，我们的方法在检索基线的基础上将宏观F1提高了最多10.0个百分点，而较小的模型（2B--20B）通过跨模型迁移获得了最多11.5个百分点的提升。

View on arXiv Download PDF AI Translation

cs.CL / 121 / 2609.01572

From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix

从生产流量到后训练：构建一个覆盖企业请求组合的自托管大型语言模型

Tsymboi, Olga, Stoianov, Dmitrii, Latypov, Ramil, Taranets, Danil, Dryabin, Daniil, Gashkov, Mikhail, Zelenkovskiy, Viktor, Fida, Aleksandr, Alektorov, Gleb, Gulyakov, Nikita, Babkin, Arthur, Medvedev, Aleksandr, Gein, Pavel, Potapov, Anatolii

Abstract

Data-residency constraints force enterprises to self-host LLMs, but continuous adoption of newer models without decommissioning their predecessors expands the serving fleet, fragmenting a finite GPU pool. We consolidate traffic from over 200 internal applications onto a single model by closing quality gaps identified through production error analysis along three axes: instruction following, function-calling, and internal task distribution. Quality is tracked by offline benchmarks stratified to production traffic and scored by deterministic verifiers or calibrated LLM judges. Rather than optimising all objectives jointly, which introduces cross-domain reward interference, we train a separate GRPO expert per axis and merge them via two-stage SLERP. Each expert's reward exposes a distinct failure mode, namely semantic collapse, over-calling, and verbosity hacking, each requiring a domain-specific fix. In non-reasoning mode the recipe surpasses a ${\sim}7\times$ larger by total parameters baseline on the in-house Arena with 69.6 to 65.8, instruction following with 0.85 to 0.83, and function-calling with 0.79 to 0.77, while lifting general dialogue benchmarks. The model absorbs 50% of platform traffic, 116M requests per month, at a fraction of the serving cost.

Chinese Translation

数据驻留限制迫使企业自托管大型语言模型（LLMs），但在不退役前代模型的情况下持续采用新模型会扩大服务队伍，导致有限的GPU资源池碎片化。我们通过关闭在生产错误分析中识别的质量差距，将来自200多个内部应用的流量整合到单一模型上，分析的三个维度包括：指令遵循、函数调用和内部任务分配。质量通过离线基准测试进行跟踪，这些基准测试根据生产流量进行分层，并由确定性验证器或经过校准的LLM评审员进行评分。我们并不联合优化所有目标，因为这会引入跨领域奖励干扰，而是为每个维度训练一个独立的GRPO专家，并通过两阶段的SLERP进行合并。每个专家的奖励揭示了不同的失败模式，即语义崩溃、过度调用和冗长黑客，每种模式都需要特定领域的修复。在非推理模式下，该方案在内部Arena上超越了总参数量约为7倍的基线，得分从69.6提升至65.8，指令遵循从0.85提升至0.83，函数调用从0.79提升至0.77，同时提升了通用对话基准。该模型吸收了50%的平台流量，每月处理116M请求，且服务成本仅为一小部分。

View on arXiv Download PDF AI Translation

cs.CL / 122 / 2609.01573

Scaling Near-Optimal SFT-RL Annotation Budget Allocation from Small to Large LLMs

从小型到大型语言模型的近似最优SFT-RL注释预算分配的扩展

Wang, Jingtan, Verma, Arun, Lin, Xiaoqiang, Liu, Zhengyuan, Chen, Nancy F., Rus, Daniela, Low, Bryan Kian Hsiang

Abstract

How to divide a fixed annotation budget between supervised fine-tuning (SFT) and reinforcement learning (RL) during LLM post-training remains an open problem. Existing work characterizes only broad trends (e.g., SFT dominates in low-data regimes), lacks a principled allocation framework, and does not examine whether the optimal ratio transfers across model sizes. We frame this problem in terms of near-optimality: rather than seeking a single optimal SFT-RL ratio, we characterize the near-optimal region, the set of allocations within a specified tolerance of peak performance. Empirically, this region is wide even for small tolerances (2-10%), widens with model scale, and transfers reliably from small proxy models to large target models. This yields a practical strategy: small proxy-model experiments suffice to identify a transferable near-optimal region, eliminating the need for exhaustive large-scale search. Our results hold consistently across tasks, model families, and both preference-based off-policy and reward-supervision on-policy RL methods. We further analyze how the asymmetry in annotation costs between SFT and RL data shifts the near-optimal region.

Chinese Translation

在大型语言模型（LLM）后训练过程中，如何在监督微调（SFT）和强化学习（RL）之间分配固定的注释预算仍然是一个未解决的问题。现有研究仅描述了广泛的趋势（例如，在低数据环境中SFT占主导地位），缺乏一个有原则的分配框架，并且没有考察最佳比例是否能够在不同模型规模之间转移。我们将此问题框定为近似最优性：我们不是寻求单一的最优SFT-RL比例，而是描述近似最优区域，即在特定性能峰值容忍度内的分配集合。从经验上看，即使在小的容忍度（2-10%）下，这个区域也很宽广，并且随着模型规模的增加而扩大，并且能够可靠地从小型代理模型转移到大型目标模型。这提供了一种实用的策略：小型代理模型实验足以识别可转移的近似最优区域，从而消除对大规模搜索的需求。我们的结果在不同任务、模型家族以及基于偏好的离线策略和奖励监督的在线RL方法中都保持一致。我们进一步分析了SFT和RL数据之间注释成本的不对称性如何影响近似最优区域的变化。

View on arXiv Download PDF AI Translation

cs.CL / 123 / 2609.01575

Closing Cost-Quality Gap in Document VLMs: Difficulty-Aware Data Curation and Quality-Adjusted Deployment Economics

缩小文档VLM中的成本-质量差距：基于难度感知的数据管理与质量调整的部署经济学

Evdokimov, Maksim, Ivanov, Matvey, Tsiupin, Dmitrii, Tsymboi, Olga, Potapov, Anatolii, Ivanov, Aleksandr

Abstract

Extracting structured fields from hundreds of millions of documents annually remains costly in regulated industries: bespoke OCR cascades cover only a fraction of workflows, privacy rules preclude external models, and existing open-source VLMs that clear quality thresholds cost more to serve than human annotation. We present a deployed document-understanding system built on a Mixture-of-Experts VLM (35B total, 3B active), fine-tuned on in-house production data mixed with open-domain documents curated by a Difficulty-Aware pipeline for layout diversity, fact-extractability, and cross-model consistency. Fitting on a single H100 and serving heterogeneous workflows via prompting, the model leads all deployable (non-reasoning) baselines up to an order of magnitude larger. A quality-adjusted cost analysis, with confirmation and correction costs calibrated from production telemetry, shows it reduces expected costs by over 80% against the human baseline and by more than 50% against the best competing open-source model, while larger baselines remain economically unviable.

Chinese Translation

在受监管行业中，每年从数亿份文档中提取结构化字段仍然成本高昂：定制的OCR级联仅覆盖了一小部分工作流程，隐私规则禁止使用外部模型，而现有的满足质量标准的开源VLM的服务成本高于人工标注。我们提出了一种基于混合专家模型（Mixture-of-Experts VLM，35B总参数，3B活跃参数）的文档理解系统，该系统经过在内部生产数据与通过难度感知管道策划的开放域文档的混合微调，以实现布局多样性、事实可提取性和跨模型一致性。该模型在单个H100上进行训练，并通过提示服务异构工作流程，超越了所有可部署（非推理）基线，规模大一个数量级。经过质量调整的成本分析，结合从生产遥测中校准的确认和纠正成本，显示其预期成本比人工基线降低超过80%，比最佳竞争开源模型降低超过50%，而更大的基线在经济上仍然不可行。

View on arXiv Download PDF AI Translation

cs.CL / 124 / 2609.01591

StudentSim: Training LLM-based Student Simulators

StudentSim：基于大型语言模型的学生模拟器训练

Yang, Ke, Wang, Chenglong, Galley, Michel, Singh, Chandan, Inala, Jeevana Priya, Zhai, ChengXiang, Gao, Jianfeng

Abstract

AI tutors are most useful when they adapt to each student's strengths, weaknesses, and preferred guidance, but evidence about which guidance works for which student is sparse, slow, and costly to collect from real learners. Student simulators can provide this signal as a proxy, yet existing approaches are limited: state-tracking models fit student behavior but struggle to process explanations or corrections, while LLM role-play follows guidance fluently but does not reliably match the competence of the student being imitated. We present StudentSim, a training framework that turns sparse per-student data into individualized simulators through pooled training followed by per-student specialization. The resulting simulators both mirror a student's own responses and update them under tutor guidance. We also introduce StudentSimEval, a standardized protocol covering 60 students across chess, second-language English writing, and mathematics, using public learner datasets with de-identified records shared for research. StudentSimEval measures behavioral fidelity (F), or how well a simulator matches a student's responses, and guidance responsiveness (R), or how readily it updates under tutor guidance, with all methods fit and evaluated on the same records. Across all three domains, StudentSim outperforms GPT-5.4 on both metrics. In chess, StudentSim reaches F=0.51 and R=0.91, compared with 0.23 and 0.72 for GPT-5.4 and 0.45 and 0.27 for Maia2. As a proof of concept, using StudentSim as a reward model for tutor reinforcement learning produces a chess tutor that expert humans rate as more accurate, better-guided, and more personalized than a no-RL baseline and a tutor trained against a GPT-5.4 simulator reward. Code is available at https://github.com/microsoft/StudentSim.

Chinese Translation

人工智能辅导员在适应每个学生的优势、劣势和偏好指导时最为有效，但关于哪种指导适合哪个学生的证据稀缺、收集缓慢且成本高昂。学生模拟器可以作为代理提供这种信号，但现有的方法存在局限性：状态跟踪模型能够适应学生行为，但在处理解释或纠正方面表现不佳，而大型语言模型（LLM）角色扮演能够流畅地遵循指导，但无法可靠地匹配被模仿学生的能力。我们提出了StudentSim，一个训练框架，通过汇总训练将稀疏的每位学生数据转化为个性化的模拟器，随后进行每位学生的专业化。生成的模拟器不仅能够反映学生自身的反应，还能在辅导员的指导下进行更新。我们还引入了StudentSimEval，一个标准化协议，涵盖了60名学生在国际象棋、第二语言英语写作和数学等领域的表现，使用公开的学习者数据集，包含为研究共享的去标识化记录。StudentSimEval测量行为保真度（F），即模拟器与学生反应的匹配程度，以及指导响应性（R），即在辅导员指导下的更新能力，所有方法均在相同记录上进行拟合和评估。在所有三个领域中，StudentSim在这两个指标上均优于GPT-5.4。在国际象棋中，StudentSim达到F=0.51和R=0.91，而GPT-5.4分别为0.23和0.72，Maia2为0.45和0.27。作为概念验证，使用StudentSim作为辅导员强化学习的奖励模型，生成的国际象棋辅导员被专家人类评估为比无强化学习基线和针对GPT-5.4模拟器奖励训练的辅导员更准确、更具指导性和更个性化。代码可在https://github.com/microsoft/StudentSim获取。

View on arXiv Download PDF AI Translation

cs.CL / 125 / 2609.01597

The Rise of Verbal Reinforcement Learning

语言强化学习的崛起

Tayal, Kshitij, Sharma, Arun, Winata, Genta Indra, Das, Anirban, Sahu, Sambit

Abstract

Natural language is emerging as a primary feedback channel for improving language agents, capable of conveying intent, preferences, and causal structure in forms interpretable by both humans and modern language models. We call this paradigm Verbal Reinforcement Learning (VRL) and offer the first unified account of it. We organize the field around a single axis, \textit{when} verbal feedback takes effect in an agent's lifecycle and \textit{what} it modifies, yielding three pillars: (1) \textbf{Language as Grounding Signal}, where language defines the task itself by specifying goals, states, and reward structures; (2) \textbf{Language as Deliberative Feedback}, where natural language guides reasoning at test time without the need to update model parameters; (3) \textbf{Language as Learning Signal}, where language-based feedback shapes model parameters through training. Within each pillar, we synthesize representative work, distinguish key subcategories of approaches, and outline the distinct role language plays in shaping agent behavior. Together, this taxonomy shows how verbal reinforcement is reshaping agent development, while also defining the challenges and opportunities for building more capable and aligned agents.

Chinese Translation

自然语言正逐渐成为改善语言代理的主要反馈渠道，能够以人类和现代语言模型均可理解的形式传达意图、偏好和因果结构。我们称这种范式为语言强化学习（Verbal Reinforcement Learning, VRL），并提供了首个统一的描述。我们围绕一个单一的轴线组织该领域，即 extit{何时}语言反馈在代理生命周期中生效以及 extit{什么}被修改，从而形成三个支柱：(1) extbf{语言作为基础信号}，在此，语言通过指定目标、状态和奖励结构来定义任务本身；(2) extbf{语言作为深思反馈}，在此，自然语言在测试时引导推理，而无需更新模型参数；(3) extbf{语言作为学习信号}，在此，基于语言的反馈通过训练塑造模型参数。在每个支柱中，我们综合了代表性工作，区分了方法的关键子类别，并概述了语言在塑造代理行为中所扮演的独特角色。总体而言，这一分类法展示了语言强化如何重塑代理开发，同时定义了构建更强大和更一致的代理所面临的挑战与机遇。

View on arXiv Download PDF AI Translation

cs.CL / 126 / 2609.01600

CordisBench: Can Language Models Reason About Component Lifecycles in Dynamic Agent Harnesses?

CordisBench：语言模型能否推理动态代理工具中的组件生命周期？

Sileo, Damien, Kachler, Dimitri

Abstract

Dynamic agent harnesses let language models change the software that shapes their own execution. This flexibility brings a new reasoning burden: a local plugin change can propagate through dependencies and cleanup. We introduce CordisBench, a 1,200-question benchmark of this lifecycle reasoning. It combines a controlled formal setting with programs executed against Cordis, a runtime that manages component dependencies and cleanup, and asks models to identify affected components, predict state after a specified teardown order, determine which conditions hold under all or some orders, and choose reconfigurations that succeed when executed. Across these tasks, we evaluate three efficiency-oriented models at low reasoning effort with 2, 4, 8, 16, 24, or 32 relevant interactions, using deterministic task-specific scoring. Models usually handle small systems well but grow less reliable as more interactions become relevant, especially when predicting final state and when reasoning across teardown orders. Additional inference effort recovers marked gains for some models. The cost is nontrivial: on our 16-interaction subset, GPT-5.6 Luna uses nearly 3,000 reasoning tokens per question at medium effort. For these controlled instances, that cost is avoidable: an independent finite reference semantics agrees with Cordis execution on every observation and action outcome used for scoring across all 528 executable questions.

Chinese Translation

动态代理工具允许语言模型更改影响其自身执行的软件。这种灵活性带来了新的推理负担：本地插件的更改可能通过依赖关系和清理过程传播。我们引入了CordisBench，这是一个包含1,200个问题的生命周期推理基准。它结合了一个受控的形式化环境与在Cordis上执行的程序，Cordis是一个管理组件依赖关系和清理的运行时，并要求模型识别受影响的组件，预测在指定的拆解顺序后的状态，确定在所有或某些顺序下哪些条件成立，以及选择在执行时成功的重配置。在这些任务中，我们评估了三种以效率为导向的模型，使用确定性的任务特定评分，在2、4、8、16、24或32个相关交互的低推理努力下。模型通常能够很好地处理小型系统，但随着更多交互变得相关，其可靠性降低，尤其是在预测最终状态和跨拆解顺序推理时。额外的推理努力为某些模型带来了显著的收益。其成本并不低：在我们的16交互子集上，GPT-5.6 Luna在中等努力下每个问题使用近3,000个推理令牌。对于这些受控实例，这一成本是可以避免的：一个独立的有限参考语义在所有528个可执行问题中，对每个观察和动作结果的评分均与Cordis执行一致。

View on arXiv Download PDF AI Translation

cs.CL / 127 / 2609.01604

Beyond Scores: Understanding LLM-as-a-Judge Mechanisms in Summarization Evaluation

超越分数：理解摘要评估中 LLM 作为评判者的机制

Vasava, Himil, Jiang, Ming

Abstract

LLM-based evaluators of natural language generation (NLG) quality are widely deployed as scoring tools and as automated training signals, yet the internal procedure by which they assign a rating remains poorly understood. We investigate this procedure mechanistically through an eight-attack perturbation taxonomy across the Readability and Adequacy dimensions of NLG quality, a generation pipeline that produces paired clean and corrupt summaries with controlled error intensity and explicit token-level modification maps, and a four-experiment battery of causal tracing, logit-lens vocabulary projection, and attention-head knockout applied to Themis (Llama-3-8B) and Prometheus (Mistral-7B). Both evaluators implement a structured, coherent evaluation pipeline operating in two stages: below layer 15, attention performs local error comparison and routes the result to the final input position; above it, the MLP cascade integrates the signal and writes the rating, with the decision crystallizing in the residual stream at a sharp late layer (L = 26 on Themis, L = 25 on Prometheus). Furthermore, a base-model control at the same scale (Llama-3-8B) reproduces the routing architecture and crystallization but not the stage separation, isolating the two mechanisms that fine-tuning specifically installs, suppression of below-L15 MLP contribution at the last position and a two-layer advance of the crystallization depth, indicating that fine-tuning sculpts an existing substrate rather than building the pipeline from scratch. We release the source code and data at https://github.com/himil-v/judge-mech

Chinese Translation

基于 LLM 的自然语言生成（NLG）质量评估工具被广泛应用于评分工具和自动化训练信号，然而它们分配评分的内部过程仍然不甚明了。我们通过对 NLG 质量的可读性和充分性维度进行八种攻击扰动分类法的机制性研究，探讨这一过程。我们构建了一个生成管道，产生成对的干净和损坏摘要，控制错误强度并提供明确的标记级别修改映射，并进行了四项实验，应用因果追踪、logit-lens 词汇投影和注意力头剔除，针对 Themis (Llama-3-8B) 和 Prometheus (Mistral-7B) 进行分析。两个评估器都实现了一个结构化、连贯的评估管道，分为两个阶段：在第 15 层以下，注意力执行局部错误比较，并将结果路由到最终输入位置；在此之上，多层感知器（MLP）级联整合信号并生成评分，决策在残差流中于一个较晚的层次（Themis 的 L = 26，Prometheus 的 L = 25）中形成。此外，规模相同的基础模型控制（Llama-3-8B）再现了路由架构和结晶过程，但未能实现阶段分离，隔离了微调特别安装的两个机制，即在最后位置抑制低于 L15 的 MLP 贡献和结晶深度的两层提前，表明微调是在现有基础上进行雕刻，而非从零开始构建管道。我们在 https://github.com/himil-v/judge-mech 发布了源代码和数据。

View on arXiv Download PDF AI Translation

arXiv Papers

Inverse kinematic solution for generic 3R positional robots using Conformal Geometric Algebra

Geometric analysis of generic 3R robots, and necessary and sufficient conditions for a class of orthogonal robots to have four IKS

Scene Graph-based Driving Scenario Extraction for Automotive Egocentric Datasets

Risk-Aware Decision-Making for Autonomous Overtaking: A World Model-Based Mixture-of-Experts Framework

Beyond Object Selection:Markerless Gaze-based Robot Placement at Arbitrary Position

Exploring Nonlinear Body Oscillations for Natural Quadruped Gaits

Potential-Guided Particle Steering for Negation-Constrained Dexterous Grasping

A Wearable Pneumatic Device for Continuous, Closed-Loop, Bidirectional Tactile Interaction

DSG: Dynamic 3D Scene Graph Construction for Embodied Agents in Changing Indoor Environments

AM-Bench: A Modular Simulation Suite and Benchmark for Aerial Manipulation Policy Learning

Human-robot conversation with multiple participants in noisy public spaces

Fleets Need a Context Plane: Rethinking Cooperative Perception for Autonomous Drones

Behavior--Realization Separation for Constrained Physical Human--Robot Interaction

ADAPT: Agile Diffusion Action Priors for Robust and Steerable Online Text-Driven Humanoid Control

Context-Aware Intelligent Vehicles

One Print, Many Moves: Monolithic Origami-inspired Folding Actuator for Composable Soft Multi-DoF Systems

A Compact Robotic Finger with 2-DoF MCP Joint Embedding DoF-Selective Passive Continuously Variable Transmission for Wide Force-Speed Operating Range

Non-Prehensile Throwing: A Reinforcement Learning Perspective

Connectivity-Aware Graph Extension for Decentralized Multi-Robot Exploration

Peg-in-Bench: A Modular Benchmark for High-Precision Robotic Insertion

Knowing When to Stop: Adaptive Action Chunking via Internal Cross-Attention Dynamics in VLAs

VerNav: Verifier-First Low-Latency Vision-and-Language Navigation

ProxPI: Proximal Prior Injection for Sampling-Based MPC under Learned-Prior Mismatch

HitMem: Hierarchical Temporal 3D Memory with Multi-Modal Context-Aware Retrieval for Dynamic Environments

Accelerating Reinforcement Learning via MPC Solver-Gradient Guidance for Weights-varying MPC

Adaptive Depth-Map-Guided Bundle Adjustment for Correspondence-Free Multi-View Point Cloud Registration

DNC-IMM: Early Lane-Change Intention Recognition via Neural Calibration Based on Driving Context Information

On Global Regulatability of Robot Manipulators by Classical PID

EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents

Integrating Traffic Noise Emission Modelling into Variable Speed Limit Control

Scalable Rao-Blackwellized Online Planning for High-Dimensional POMDPs

Obstacle-Aware Autonomous Coverage and Navigation for Outdoor Robots

Autonomous robotic bridging using distributed swarm control without inter-agent communication

Evaluating Multimodal LLMs as Generalist Vision-Language-Action Agents for Drone Control: Commanding, Approaching, Tracking and Searching

Vision-Based Leader-Follower Formation Control for Cooperative UAVs in GPS-Degraded Environments

Does Imitation Learning Preserve Temporal Robustness in Dexterous Manipulation? An Expert-Learner Comparison Across Task Execution Speeds

A System for Fast, Resilient, and Adaptable Loco-Manipulation Behaviors on Humanoid Robots

SG-AMP: Scene-Graph-Guided Active Perception and Semantics-Aware Motion Planning for Pepper Plants

Facet-0: A Robotic Foundation Model for Contact-Rich Precise Manipulation

A Cone-Constrained Bilinear Decomposition for Total Scaled-Gradient Variation Models

Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving

ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training

A Lagrangian View of Flow Matching

Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM-Based Video Moderation

Beyond Language Priors: Diagnosing and Fixing Visual-Origin Hallucinations in Multimodal LLM

Beyond Blind Compliance: Benchmarking Task Verification in OCR Reasoning

CoLT-Drive: Counterfactual Long-Tail Benchmarking and Knowledge-Preserving Adaptation for Driving Affordance Prediction

CrossFeat: Bridging Imaging Modalities in Feature Descriptor Space

StreamScout: Learning When to Look Deeper for Streaming Video Understanding

TRUST: Threshold-Recalibrated Uncertainty-Safe Training for Certified Dismissal in Breast Cancer Screening

Where Should Experience Live? Hierarchical Hebbian Memory for Continual Vision Transformers

Puppeteer: Object-Grounded Posture-Aware Co-Speech Gesture Generation

FoldingAgent: Inferring Parametric Origami Procedures from Demonstration Videos

SlideMix: Enhancing Whole Slide Image Analysis via Multimodal Shuffling

Unmasking Face Embeddings: Reading, Rendering and Naming with Foundation Models

Instance-Guided Report Anchoring for Text-Free 3D Abnormality Segmentation in Chest CT

SAM3-LoRA: Parameter-Efficient Adaptation of a Concept-Promptable Foundation Model for Multi-Class Structural Defect Segmentation

Less Is More: Balancing Positive and Negative Space in Visual Concept Blending

ViTAL-X: Video-Text Alignment with Cross-Modal Temporal Edits

Soft-Argmax for the Projective Plane via the Veronese Embedding

GenScale: A Benchmark for Relative Object Scale in Image Generation and Editing

A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss

BrainDiff: Longitudinal Report Generation for Multimodal Brain MRI

Streaming4D: Accelerate 4D World Models via Block-wise Video Generation and Incremental Reconstruction

Restrict, Don't Retrain: Inference-Time VLM Guidance for Zero-Shot Aerial Segmentation

Beyond Landmark Extraction: A Framework for Robust Geometric Feature Construction in Structured Image Classification

You Cannot Photograph the Same Street Twice: Reliability Limits in Vision-Language Measurement of Urban Change

Physically Plausible Video Generation via Visual-Semantic Chain-of-Events Conditioning

Teaching Vision-Language Models to Use the Scale They Are Given: Label-Free Equivariance Training for Metric Physical Reasoning

Do Satellites See Commuters? A Critical Benchmark of Vision Foundation Models

Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures

DGNet: Dual-knowledge Guided Network for Infrared Small Target Detection

FTU-Seek: Foundation Model-Guided Hard-Negative Learning for Sparse Functional Tissue Unit Segmentation

Controllable Image Captioning with Prompt-Conditioned Scene Rewards

EarthLD: Towards Unified Open-World Landslide Understanding via Vision-Language Guided Diffusion Models

Efficient and Robust Absolute Pose Estimation via Gravity-Prior-Driven Transformation Decoupling and Pose Refinement

Design and Implementation of a Kalman Filter-Infused Algorithm for Tilt Estimation

Mind the Rift: Cross-Scale Coupling Mismatch for AI-Generated Video Detection

Feed-Forward Multi-view Multi-person Reconstruction with Contrastive Human-Aware 3D Representation