# Research Papers — 2025-10 · bundle page 158 of 217 > Full-text markdown of papers #7851–7900 of 10843 published in 2025-10, newest first. AI-readable mirrors of the full 10843-paper 2025-10 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2025-10/page/158 [← Previous bundle](https://aipod.dev/papers/archive/2025-10/page/157.md) [Next bundle →](https://aipod.dev/papers/archive/2025-10/page/159.md) ### DM1: MeanFlow with Dispersive Regularization for 1-Step Robotic Manipulation Authors: Guowei Zou, Haitao Wang, Hejun Wu, Yukun Qian, Yuhang Wang, Weibing Li Published: 2025-10-09 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2510.07865v1 The ability to learn multi-modal action distributions is indispensable for robotic manipulation policies to perform precise and robust control. Flow-based generative models have recently emerged as a promising solution to learning distributions of actions, offering one-step action generation and thus achieving much higher sampling efficiency compared to diffusion-based methods. However, existing flow-based policies suffer from representation collapse, the inability to distinguish similar visual representations, leading to failures in precise manipulation tasks. We propose DM1 (MeanFlow with Dispersive Regularization for One-Step Robotic Manipulation), a novel flow matching framework that integrates dispersive regularization into MeanFlow to prevent collapse while maintaining one-step efficiency. DM1 employs multiple dispersive regularization variants across different intermediate embedding layers, encouraging diverse representations across training batches without introducing additional network modules or specialized training procedures. Experiments on RoboMimic benchmarks show that DM1 achieves 20-40 times faster inference (0.07s vs. 2-3.5s) and improves success rates by 10-20 percentage points, with the Lift task reaching 99% success over 85% of the baseline. Real-robot deployment on a Franka Panda further validates that DM1 transfers effectively from simulation to the physical world. To the best of our knowledge, this is the first work to leverage representation regularization to enable flow-based policies to achieve strong performance in robotic manipulation, establishing a simple yet powerful approach for efficient and robust manipulation. --- ### On the Optimality of Tracking Fisher Information in Adaptive Testing with Stochastic Binary Responses Authors: Sanghwa Kim, Dohyun Ahn, Seungki Min Published: 2025-10-09 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2510.07862v1 We study the problem of estimating a continuous ability parameter from sequential binary responses by actively asking questions with varying difficulties, a setting that arises naturally in adaptive testing and online preference learning. Our goal is to certify that the estimate lies within a desired margin of error, using as few queries as possible. We propose a simple algorithm that adaptively selects questions to maximize Fisher information and updates the estimate using a method-of-moments approach, paired with a novel test statistic to decide when the estimate is accurate enough. We prove that this Fisher-tracking strategy achieves optimal performance in both fixed-confidence and fixed-budget regimes, which are commonly invested in the best-arm identification literature. Our analysis overcomes a key technical challenge in the fixed-budget setting -- handling the dependence between the evolving estimate and the query distribution -- by exploiting a structural symmetry in the model and combining large deviation tools with Ville's inequality. Our results provide rigorous theoretical support for simple and efficient adaptive testing procedures. --- ### Understanding DeepResearch via Reports Authors: Tianyu Fan, Xinyao Niu, Yuxiang Zheng, Fengji Zhang, Chengen Huang, Bei Chen, Junyang Lin, Chao Huang Published: 2025-10-09 Categories: cs.AI arXiv: https://arxiv.org/abs/2510.07861v1 DeepResearch agents represent a transformative AI paradigm, conducting expert-level research through sophisticated reasoning and multi-tool integration. However, evaluating these systems remains critically challenging due to open-ended research scenarios and existing benchmarks that focus on isolated capabilities rather than holistic performance. Unlike traditional LLM tasks, DeepResearch systems must synthesize diverse sources, generate insights, and present coherent findings, which are capabilities that resist simple verification. To address this gap, we introduce DeepResearch-ReportEval, a comprehensive framework designed to assess DeepResearch systems through their most representative outputs: research reports. Our approach systematically measures three dimensions: quality, redundancy, and factuality, using an innovative LLM-as-a-Judge methodology achieving strong expert concordance. We contribute a standardized benchmark of 100 curated queries spanning 12 real-world categories, enabling systematic capability comparison. Our evaluation of four leading commercial systems reveals distinct design philosophies and performance trade-offs, establishing foundational insights as DeepResearch evolves from information assistants toward intelligent research partners. Source code and data are available at: https://github.com/HKUDS/DeepResearch-Eval. --- ### Augur: Modeling Covariate Causal Associations in Time Series via Large Language Models Authors: Zhiqing Cui, Binwu Wang, Qingxiang Liu, Yeqiang Wang, Zhengyang Zhou, Yuxuan Liang, Yang Wang Published: 2025-10-09 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2510.07858v2 Large language models (LLM) have emerged as a promising avenue for time series forecasting, offering the potential to integrate multimodal data. However, existing LLM-based approaches face notable limitations-such as marginalized role in model architectures, reliance on coarse statistical text prompts, and lack of interpretability. In this work, we introduce Augur, a fully LLM driven time series forecasting framework that exploits LLM causal reasoning to discover and use directed causal associations among covariates. Augur uses a two stage teacher student architecture where a powerful teacher LLM infers a directed causal graph from time series using heuristic search together with pairwise causality testing. A lightweight student agent then refines the graph and fine tune on high confidence causal associations that are encoded as rich textual prompts to perform forecasting. This design improves predictive accuracy while yielding transparent, traceable reasoning about variable interactions. Extensive experiments on real-world datasets with 26 baselines demonstrate that Augur achieves competitive performance and robust zero-shot generalization. --- ### XYZCylinder: Towards Compatible Feed-Forward 3D Gaussian Splatting for Driving Scenes via Unified Cylinder Lifting Method Authors: Haochen Yu, Qiankun Liu, Hongyuan Liu, Jianfei Jiang, Juntao Lyu, Jiansheng Chen, Huimin Ma Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07856v2 Feed-forward paradigms for 3D reconstruction have become a focus of recent research, which learn implicit, fixed view transformations to generate a single scene representation. However, their application to complex driving scenes reveals significant limitations. Two core challenges are responsible for this performance gap. First, the reliance on a fixed view transformation hinders compatibility to varying camera configurations. Second, the inherent difficulty of learning complex driving scenes from sparse 360° views with minimal overlap compromises the final reconstruction fidelity. To handle these difficulties, we introduce XYZCylinder, a novel method built upon a unified cylinder lifting method that integrates camera modeling and feature lifting. To tackle the compatibility problem, we design a Unified Cylinder Camera Modeling (UCCM) strategy. This strategy explicitly models projection parameters to unify diverse camera setups, thus bypassing the need for learning viewpoint-dependent correspondences. To improve the reconstruction accuracy, we propose a hybrid representation with several dedicated modules based on newly designed Cylinder Plane Feature Group (CPFG) to lift 2D image features to 3D space. Extensive evaluations confirm that XYZCylinder not only achieves state-of-the-art performance under different evaluation settings but also demonstrates remarkable compatibility in entirely new scenes with different camera settings in a zero-shot manner. Project page: \href{https://yuyuyu223.github.io/XYZCYlinder-projectpage/}{here} --- ### Self-Supervised Learning Strategies for a Platform to Test the Toxicity of New Chemicals and Materials Authors: Thomas Lautenschlager, Nils Friederich, Angelo Jovin Yamachui Sitcheu, Katja Nau, Gaëlle Hayot, Thomas Dickmeis, Ralf Mikut Published: 2025-10-09 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2510.07853v1 High-throughput toxicity testing offers a fast and cost-effective way to test large amounts of compounds. A key component for such systems is the automated evaluation via machine learning models. In this paper, we address critical challenges in this domain and demonstrate how representations learned via self-supervised learning can effectively identify toxicant-induced changes. We provide a proof-of-concept that utilizes the publicly available EmbryoNet dataset, which contains ten zebrafish embryo phenotypes elicited by various chemical compounds targeting different processes in early embryonic development. Our analysis shows that the learned representations using self-supervised learning are suitable for effectively distinguishing between the modes-of-action of different compounds. Finally, we discuss the integration of machine learning models in a physical toxicity testing device in the context of the TOXBOX project. --- ### FinMR: A Knowledge-Intensive Multimodal Benchmark for Advanced Financial Reasoning Authors: Shuangyan Deng, Haizhou Peng, Jiachen Xu, Rui Mao, Ciprian Doru Giurcăneanu, Jiamou Liu Published: 2025-10-09 Categories: cs.AI arXiv: https://arxiv.org/abs/2510.07852v2 Multimodal Large Language Models (MLLMs) have made substantial progress in recent years. However, their rigorous evaluation within specialized domains like finance is hindered by the absence of datasets characterized by professional-level knowledge intensity, detailed annotations, and advanced reasoning complexity. To address this critical gap, we introduce FinMR, a high-quality, knowledge-intensive multimodal dataset explicitly designed to evaluate expert-level financial reasoning capabilities at a professional analyst's standard. FinMR comprises over 3,200 meticulously curated and expertly annotated question-answer pairs across 15 diverse financial topics, ensuring broad domain diversity and integrating sophisticated mathematical reasoning, advanced financial knowledge, and nuanced visual interpretation tasks across multiple image types. Through comprehensive benchmarking with leading closed-source and open-source MLLMs, we highlight significant performance disparities between these models and professional financial analysts, uncovering key areas for model advancement, such as precise image analysis, accurate application of complex financial formulas, and deeper contextual financial understanding. By providing richly varied visual content and thorough explanatory annotations, FinMR establishes itself as an essential benchmark tool for assessing and advancing multimodal financial reasoning toward professional analyst-level competence. --- ### Meta-Learning Based Few-Shot Graph-Level Anomaly Detection Authors: Liting Li, Yumeng Wang, Yueheng Sun Published: 2025-10-09 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2510.07847v1 Graph-level anomaly detection aims to identify anomalous graphs or subgraphs within graph datasets, playing a vital role in various fields such as fraud detection, review classification, and biochemistry. While Graph Neural Networks (GNNs) have made significant progress in this domain, existing methods rely heavily on large amounts of labeled data, which is often unavailable in real-world scenarios. Additionally, few-shot anomaly detection methods based on GNNs are prone to noise interference, resulting in poor embedding quality and reduced model robustness. To address these challenges, we propose a novel meta-learning-based graph-level anomaly detection framework (MA-GAD), incorporating a graph compression module that reduces the graph size, mitigating noise interference while retaining essential node information. We also leverage meta-learning to extract meta-anomaly information from similar networks, enabling the learning of an initialization model that can rapidly adapt to new tasks with limited samples. This improves the anomaly detection performance on target graphs, and a bias network is used to enhance the distinction between anomalous and normal nodes. Our experimental results, based on four real-world biochemical datasets, demonstrate that MA-GAD outperforms existing state-of-the-art methods in graph-level anomaly detection under few-shot conditions. Experiments on both graph anomaly and subgraph anomaly detection tasks validate the framework's effectiveness on real-world datasets. --- ### AdaSwitch: Balancing Exploration and Guidance in Knowledge Distillation via Adaptive Switching Authors: Jingyu Peng, Maolin Wang, Hengyi Cai, Yuchen Li, Kai Zhang, Shuaiqiang Wang, Dawei Yin, Xiangyu Zhao Published: 2025-10-09 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2510.07842v3 Small language models (SLMs) are crucial for applications with strict latency and computational constraints, yet achieving high performance remains challenging. Knowledge distillation (KD) can transfer capabilities from large teacher models, but existing methods face a dilemma: off-policy distillation provides high-quality supervision but suffers from exposure bias (training inference mismatch), while on-policy approaches ensure consistency but are limited by the low quality of student-generated outputs. To address these issues, we propose AdaSwitch, a novel approach that dynamically combines on-policy and off-policy generation via an adaptive switching mechanism. AdaSwitch allows the student to explore its predictions within its capability and selectively integrates teacher guidance only when divergence exceeds a context-aware threshold. This paradigm preserves generation consistency while ensuring high-quality supervision. Experiments on three datasets demonstrate that AdaSwitch consistently improves accuracy and reasoning capability with moderate overhead. --- ### Self-Improving LLM Agents at Test-Time Authors: Emre Can Acikgoz, Cheng Qian, Heng Ji, Dilek Hakkani-Tür, Gokhan Tur Published: 2025-10-09 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2510.07841v1 One paradigm of language model (LM) fine-tuning relies on creating large training datasets, under the assumption that high quantity and diversity will enable models to generalize to novel tasks after post-training. In practice, gathering large sets of data is inefficient, and training on them is prohibitively expensive; worse, there is no guarantee that the resulting model will handle complex scenarios or generalize better. Moreover, existing techniques rarely assess whether a training sample provides novel information or is redundant with the knowledge already acquired by the model, resulting in unnecessary costs. In this work, we explore a new test-time self-improvement method to create more effective and generalizable agentic LMs on-the-fly. The proposed algorithm can be summarized in three steps: (i) first it identifies the samples that model struggles with (self-awareness), (ii) then generates similar examples from detected uncertain samples (self-data augmentation), and (iii) uses these newly generated samples at test-time fine-tuning (self-improvement). We study two variants of this approach: Test-Time Self-Improvement (TT-SI), where the same model generates additional training examples from its own uncertain cases and then learns from them, and contrast this approach with Test-Time Distillation (TT-D), where a stronger model generates similar examples for uncertain cases, enabling student to adapt using distilled supervision. Empirical evaluations across different agent benchmarks demonstrate that TT-SI improves the performance with +5.48% absolute accuracy gain on average across all benchmarks and surpasses other standard learning methods, yet using 68x less training samples. Our findings highlight the promise of TT-SI, demonstrating the potential of self-improvement algorithms at test-time as a new paradigm for building more capable agents toward self-evolution. --- ### AlignGS: Aligning Geometry and Semantics for Robust Indoor Reconstruction from Sparse Views Authors: Yijie Gao, Houqiang Zhong, Tianchi Zhu, Zhengxue Cheng, Qiang Hu, Li Song Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07839v1 The demand for semantically rich 3D models of indoor scenes is rapidly growing, driven by applications in augmented reality, virtual reality, and robotics. However, creating them from sparse views remains a challenge due to geometric ambiguity. Existing methods often treat semantics as a passive feature painted on an already-formed, and potentially flawed, geometry. We posit that for robust sparse-view reconstruction, semantic understanding instead be an active, guiding force. This paper introduces AlignGS, a novel framework that actualizes this vision by pioneering a synergistic, end-to-end optimization of geometry and semantics. Our method distills rich priors from 2D foundation models and uses them to directly regularize the 3D representation through a set of novel semantic-to-geometry guidance mechanisms, including depth consistency and multi-faceted normal regularization. Extensive evaluations on standard benchmarks demonstrate that our approach achieves state-of-the-art results in novel view synthesis and produces reconstructions with superior geometric accuracy. The results validate that leveraging semantic priors as a geometric regularizer leads to more coherent and complete 3D models from limited input views. Our code is avaliable at https://github.com/MediaX-SJTU/AlignGS . --- ### IsoSignVid2Aud: Sign Language Video to Audio Conversion without Text Intermediaries Authors: Harsh Kavediya, Vighnesh Nayak, Bheeshm Sharma, Balamurugan Palaniappan Published: 2025-10-09 Categories: cs.CV, cs.MM, cs.SD arXiv: https://arxiv.org/abs/2510.07837v1 Sign language to spoken language audio translation is important to connect the hearing- and speech-challenged humans with others. We consider sign language videos with isolated sign sequences rather than continuous grammatical signing. Such videos are useful in educational applications and sign prompt interfaces. Towards this, we propose IsoSignVid2Aud, a novel end-to-end framework that translates sign language videos with a sequence of possibly non-grammatic continuous signs to speech without requiring intermediate text representation, providing immediate communication benefits while avoiding the latency and cascading errors inherent in multi-stage translation systems. Our approach combines an I3D-based feature extraction module with a specialized feature transformation network and an audio generation pipeline, utilizing a novel Non-Maximal Suppression (NMS) algorithm for the temporal detection of signs in non-grammatic continuous sequences. Experimental results demonstrate competitive performance on ASL-Citizen-1500 and WLASL-100 datasets with Top-1 accuracies of 72.01\% and 78.67\%, respectively, and audio quality metrics (PESQ: 2.67, STOI: 0.73) indicating intelligible speech output. Code is available at: https://github.com/BheeshmSharma/IsoSignVid2Aud_AIMLsystems-2025. --- ### MetaDefense: Defending Finetuning-based Jailbreak Attack Before and During Generation Authors: Weisen Jiang, Sinno Jialin Pan Published: 2025-10-09 Categories: cs.LG, cs.AI, cs.CL, cs.CR arXiv: https://arxiv.org/abs/2510.07835v1 This paper introduces MetaDefense, a novel framework for defending against finetuning-based jailbreak attacks in large language models (LLMs). We observe that existing defense mechanisms fail to generalize to harmful queries disguised by unseen attack templates, despite LLMs being capable of distinguishing disguised harmful queries in the embedding space. Based on these insights, we propose a two-stage defense approach: (i) pre-generation defense that detects harmful queries before response generation begins, and (ii) mid-generation defense that monitors partial responses during generation to prevent outputting more harmful content. Our MetaDefense trains the LLM to predict the harmfulness of both queries and partial responses using specialized prompts, enabling early termination of potentially harmful interactions. Extensive experiments across multiple LLM architectures (LLaMA-2-7B, Qwen-2.5-3B-Instruct, and LLaMA-3.2-3B-Instruct) demonstrate that MetaDefense significantly outperforms existing defense mechanisms, achieving robust defense against harmful queries with seen and unseen attack templates while maintaining competitive performance on benign tasks. Code is available at https://github.com/ws-jiang/MetaDefense. --- ### Surrogate Graph Partitioning for Spatial Prediction Authors: Yuta Shikuri, Hironori Fujisawa Published: 2025-10-09 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2510.07832v2 Spatial prediction refers to the estimation of unobserved values from spatially distributed observations. Although recent advances have improved the capacity to model diverse observation types, adoption in practice remains limited in industries that demand interpretability. To mitigate this gap, surrogate models that explain black-box predictors provide a promising path toward interpretable decision making. In this study, we propose a graph partitioning problem to construct spatial segments that minimize the sum of within-segment variances of individual predictions. The assignment of data points to segments can be formulated as a mixed-integer quadratic programming problem. While this formulation potentially enables the identification of exact segments, its computational complexity becomes prohibitive as the number of data points increases. Motivated by this challenge, we develop an approximation scheme that leverages the structural properties of graph partitioning. Experimental results demonstrate the computational efficiency of this approximation in identifying spatial segments. --- ### PrismGS: Physically-Grounded Anti-Aliasing for High-Fidelity Large-Scale 3D Gaussian Splatting Authors: Houqiang Zhong, Zhenglong Wu, Sihua Fu, Zihan Zheng, Xin Jin, Xiaoyun Zhang, Li Song, Qiang Hu Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07830v1 3D Gaussian Splatting (3DGS) has recently enabled real-time photorealistic rendering in compact scenes, but scaling to large urban environments introduces severe aliasing artifacts and optimization instability, especially under high-resolution (e.g., 4K) rendering. These artifacts, manifesting as flickering textures and jagged edges, arise from the mismatch between Gaussian primitives and the multi-scale nature of urban geometry. While existing ``divide-and-conquer'' pipelines address scalability, they fail to resolve this fidelity gap. In this paper, we propose PrismGS, a physically-grounded regularization framework that improves the intrinsic rendering behavior of 3D Gaussians. PrismGS integrates two synergistic regularizers. The first is pyramidal multi-scale supervision, which enforces consistency by supervising the rendering against a pre-filtered image pyramid. This compels the model to learn an inherently anti-aliased representation that remains coherent across different viewing scales, directly mitigating flickering textures. This is complemented by an explicit size regularization that imposes a physically-grounded lower bound on the dimensions of the 3D Gaussians. This prevents the formation of degenerate, view-dependent primitives, leading to more stable and plausible geometric surfaces and reducing jagged edges. Our method is plug-and-play and compatible with existing pipelines. Extensive experiments on MatrixCity, Mill-19, and UrbanScene3D demonstrate that PrismGS achieves state-of-the-art performance, yielding significant PSNR gains around 1.5 dB against CityGaussian, while maintaining its superior quality and robustness under demanding 4K rendering. --- ### The Rise of the Knowledge Sculptor: A New Archetype for Knowledge Work in the Age of Generative AI Authors: Cathal Doyle Published: 2025-10-09 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2510.07829v1 In the Generative Age, the nature of knowledge work is transforming. Traditional models that emphasise the organisation and retrieval of pre-existing information are increasingly inadequate in the face of generative AI (GenAI) systems capable of autonomous content creation. This paper introduces the Knowledge Sculptor (KS), a new professional archetype for Human-GenAI collaboration that transforms raw AI output into trustworthy, actionable knowledge. Grounded in a socio-technical perspective, the KS is conceptualised through a framework of competencies, including architecting a vision, iterative dialogue, information sculpting, and curiosity-driven synthesis. A practice-based vignette illustrates the KS role in action, and in a self-referential approach, the paper itself serves as an artefact of the sculpting process it describes. --- ### MMHOI: Modeling Complex 3D Multi-Human Multi-Object Interactions Authors: Kaen Kogashi, Anoop Cherian, Meng-Yu Jennifer Kuo Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07828v3 Real-world scenes often feature multiple humans interacting with multiple objects in ways that are causal, goal-oriented, or cooperative. Yet existing 3D human-object interaction (HOI) benchmarks consider only a fraction of these complex interactions. To close this gap, we present MMHOI -- a large-scale, Multi-human Multi-object Interaction dataset consisting of images from 12 everyday scenarios. MMHOI offers complete 3D shape and pose annotations for every person and object, along with labels for 78 action categories and 14 interaction-specific body parts, providing a comprehensive testbed for next-generation HOI research. Building on MMHOI, we present MMHOI-Net, an end-to-end transformer-based neural network for jointly estimating human-object 3D geometries, their interactions, and associated actions. A key innovation in our framework is a structured dual-patch representation for modeling objects and their interactions, combined with action recognition to enhance the interaction prediction. Experiments on MMHOI and the recently proposed CORE4D datasets demonstrate that our approach achieves state-of-the-art performance in multi-HOI modeling, excelling in both accuracy and reconstruction quality. The MMHOI dataset is publicly available at https://zenodo.org/records/17711786. --- ### An LLM-Powered Cooperative Framework for Large-Scale Multi-Vehicle Navigation Authors: Yuping Zhou, Siqi Lai, Jindong Han, Hao Liu Published: 2025-10-09 Categories: cs.AI arXiv: https://arxiv.org/abs/2510.07825v1 The rise of Internet of Vehicles (IoV) technologies is transforming traffic management from isolated control to a collective, multi-vehicle process. At the heart of this shift is multi-vehicle dynamic navigation, which requires simultaneously routing large fleets under evolving traffic conditions. Existing path search algorithms and reinforcement learning methods struggle to scale to city-wide networks, often failing to capture the nonlinear, stochastic, and coupled dynamics of urban traffic. To address these challenges, we propose CityNav, a hierarchical, LLM-powered framework for large-scale multi-vehicle navigation. CityNav integrates a global traffic allocation agent, which coordinates strategic traffic flow distribution across regions, with local navigation agents that generate locally adaptive routes aligned with global directives. To enable effective cooperation, we introduce a cooperative reasoning optimization mechanism, in which agents are jointly trained with a dual-reward structure: individual rewards promote per-vehicle efficiency, while shared rewards encourage network-wide coordination and congestion reduction. Extensive experiments on four real-world road networks of varying scales (up to 1.6 million roads and 430,000 intersections) and traffic datasets demonstrate that CityNav consistently outperforms nine classical path search and RL-based baselines in city-scale travel efficiency and congestion mitigation. Our results highlight the potential of LLMs to enable scalable, adaptive, and cooperative city-wide traffic navigation, providing a foundation for intelligent, large-scale vehicle routing in complex urban environments. Our project is available at https://github.com/usail-hkust/CityNav. --- ### Enhancing Visual Prompting through Expanded Transformation Space and Overfitting Mitigation Authors: Shohei Enomoto Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07823v2 Visual prompting (VP) has emerged as a promising parameter-efficient fine-tuning approach for adapting pre-trained vision models to downstream tasks without modifying model parameters. Despite offering advantages like negligible computational overhead and compatibility with black-box models, conventional VP methods typically achieve lower accuracy than other adaptation approaches. Our analysis reveals two critical limitations: the restricted expressivity of simple additive transformation and a tendency toward overfitting when the parameter count increases. To address these challenges, we propose ACAVP (Affine, Color, and Additive Visual Prompting), which enhances VP's expressive power by introducing complementary transformation operations: affine transformation for creating task-specific prompt regions while preserving original image information, and color transformation for emphasizing task-relevant visual features. Additionally, we identify that overfitting is a critical issue in VP training and introduce TrivialAugment as an effective data augmentation, which not only benefits our approach but also significantly improves existing VP methods, with performance gains of up to 12 percentage points on certain datasets. This demonstrates that appropriate data augmentation is universally beneficial for VP training. Extensive experiments across twelve diverse image classification datasets with two different model architectures demonstrate that ACAVP achieves state-of-the-art accuracy among VP methods, surpasses linear probing in average accuracy, and exhibits superior robustness to distribution shifts, all while maintaining minimal computational overhead during inference. Our code is available at https://github.com/s-enmt/ACAVP. --- ### SIMU: Selective Influence Machine Unlearning Authors: Anu Agarwal, Mihir Pamnani, Dilek Hakkani-Tur Published: 2025-10-09 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2510.07822v1 The undesired memorization of sensitive information by Large Language Models (LLMs) has emphasized the need for safety mechanisms that can regulate model behavior. This has led to the development of machine unlearning techniques that enable models to precisely forget sensitive and unwanted information. For machine unlearning, first-order and second-order optimizer-based methods have shown significant progress in enabling LLMs to forget targeted information. However, in doing so, these approaches often compromise the model's original capabilities, resulting in unlearned models that struggle to retain their prior knowledge and overall utility. To address this, we propose Selective Influence Machine Unlearning (SIMU), a two-step framework that enhances second-order optimizer-based unlearning by selectively updating only the critical neurons responsible for encoding the forget-set. By constraining updates to these targeted neurons, SIMU achieves comparable unlearning efficacy while substantially outperforming current methods in retaining the model's original knowledge. --- ### From Keywords to Clusters: AI-Driven Analysis of YouTube Comments to Reveal Election Issue Salience in 2024 Authors: Raisa M. Simoes, Timoteo Kelly, Eduardo J. Simoes, Praveen Rao Published: 2025-10-09 Categories: cs.SI, cs.CL arXiv: https://arxiv.org/abs/2510.07821v1 This paper aims to explore two competing data science methodologies to attempt answering the question, "Which issues contributed most to voters' choice in the 2024 presidential election?" The methodologies involve novel empirical evidence driven by artificial intelligence (AI) techniques. By using two distinct methods based on natural language processing and clustering analysis to mine over eight thousand user comments on election-related YouTube videos from one right leaning journal, Wall Street Journal, and one left leaning journal, New York Times, during pre-election week, we quantify the frequency of selected issue areas among user comments to infer which issues were most salient to potential voters in the seven days preceding the November 5th election. Empirically, we primarily demonstrate that immigration and democracy were the most frequently and consistently invoked issues in user comments on the analyzed YouTube videos, followed by the issue of identity politics, while inflation was significantly less frequently referenced. These results corroborate certain findings of post-election surveys but also refute the supposed importance of inflation as an election issue. This indicates that variations on opinion mining, with their analysis of raw user data online, can be more revealing than polling and surveys for analyzing election outcomes. --- ### PAGCNet: A Pose-Aware and Geometry Constrained Framework for Panoramic Depth Estimation Authors: Kanglin Ning, Ruzhao Chen, Penghong Wang, Xingtao Wang, Ruiqin Xiong, Xiaopeng Fan Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07817v2 Explicitly modeling room background depth as a geometric constraint has proven effective for panoramic depth estimation. However, reconstructing this background depth for regular enclosed regions in a complex indoor scene without external measurements remains an open challenge. To address this, we propose a pose-aware and geometry-constrained framework for panoramic depth estimation. Our framework first employs multiple task-specific decoders to jointly estimate room layout, camera pose, depth, and region segmentation from a input panoramic image. A pose-aware background depth resolving (PA-BDR) component uses tasks decoder's prediction to resolve the camera pose. Subsequently, the proposed PA-BDR component uses the camera pose to compute the background depth of regular enclosed regions and uses this background depth as a strong geometric prior. Based on the output of the region segmentation decoder, a fusion mask generation (FMG) component produces a fusion weight map to guide where and to what extent the geometry-constrained background depth should correct the depth decoder's prediction. Finally, an adaptive fusion component integrates this refined background depth with the initial depth prediction, guided by the fusion weight. Extensive experiments on Matterport3D, Structured3D, and Replica datasets demonstrate that our method achieves significantly superior performance compared to current open-source methods. Code is available at https://github.com/emiyaning/PAGCNet. --- ### Strategic Communication under Threat: Learning Information Trade-offs in Pursuit-Evasion Games Authors: Valerio La Gatta, Dolev Mutzari, Sarit Kraus, VS Subrahmanian Published: 2025-10-09 Categories: cs.AI arXiv: https://arxiv.org/abs/2510.07813v1 Adversarial environments require agents to navigate a key strategic trade-off: acquiring information enhances situational awareness, but may simultaneously expose them to threats. To investigate this tension, we formulate a PursuitEvasion-Exposure-Concealment Game (PEEC) in which a pursuer agent must decide when to communicate in order to obtain the evader's position. Each communication reveals the pursuer's location, increasing the risk of being targeted. Both agents learn their movement policies via reinforcement learning, while the pursuer additionally learns a communication policy that balances observability and risk. We propose SHADOW (Strategic-communication Hybrid Action Decision-making under partial Observation for Warfare), a multi-headed sequential reinforcement learning framework that integrates continuous navigation control, discrete communication actions, and opponent modeling for behavior prediction. Empirical evaluations show that SHADOW pursuers achieve higher success rates than six competitive baselines. Our ablation study confirms that temporal sequence modeling and opponent modeling are critical for effective decision-making. Finally, our sensitivity analysis reveals that the learned policies generalize well across varying communication risks and physical asymmetries between agents. --- ### Multilingual Generative Retrieval via Cross-lingual Semantic Compression Authors: Yuxin Huang, Simeng Wu, Ran Song, Yan Xiang, Yantuan Xian, Shengxiang Gao, Zhengtao Yu Published: 2025-10-09 Categories: cs.CL arXiv: https://arxiv.org/abs/2510.07812v1 Generative Information Retrieval is an emerging retrieval paradigm that exhibits remarkable performance in monolingual scenarios.However, applying these methods to multilingual retrieval still encounters two primary challenges, cross-lingual identifier misalignment and identifier inflation. To address these limitations, we propose Multilingual Generative Retrieval via Cross-lingual Semantic Compression (MGR-CSC), a novel framework that unifies semantically equivalent multilingual keywords into shared atoms to align semantics and compresses the identifier space, and we propose a dynamic multi-step constrained decoding strategy during retrieval. MGR-CSC improves cross-lingual alignment by assigning consistent identifiers and enhances decoding efficiency by reducing redundancy. Experiments demonstrate that MGR-CSC achieves outstanding retrieval accuracy, improving by 6.83% on mMarco100k and 4.77% on mNQ320k, while reducing document identifiers length by 74.51% and 78.2%, respectively. --- ### Adaptive Execution Scheduler for DataDios SmartDiff Authors: Aryan Poduri Published: 2025-10-09 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2510.07811v1 We present an adaptive scheduler for a single differencing engine (SmartDiff) with two execution modes: (i) in-memory threads and (ii) Dask based parallelism. The scheduler continuously tunes batch size and worker/thread count within fixed CPU and memory budgets to minimize p95 latency. A lightweight preflight profiler estimates bytes/row and I/O rate; an online cost/memory model prunes unsafe actions; and a guarded hill-climb policy favors lower latency with backpressure and straggler mitigation. Backend selection is gated by a conservative working-set estimate so that in-memory execution is chosen when safe, otherwise Dask is used. Across synthetic and public tabular benchmarks, the scheduler reduces p95 latency by 23 to 28 percent versus a tuned warm-up heuristic (and by 35 to 40 percent versus fixed grid baselines), while lowering peak memory by 16 to 22 percent (25 to 32 percent vs. fixed) with zero OOMs and comparable throughput. --- ### FMANet: A Novel Dual-Phase Optical Flow Approach with Fusion Motion Attention Network for Robust Micro-expression Recognition Authors: Luu Tu Nguyen, Vu Tram Anh Khuong, Thi Bich Phuong Man, Thi Duyen Ngo, Thanh Ha Le Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07810v3 Facial micro-expressions, characterized by their subtle and brief nature, are valuable indicators of genuine emotions. Despite their significance in psychology, security, and behavioral analysis, micro-expression recognition remains challenging due to the difficulty of capturing subtle facial movements. Optical flow has been widely employed as an input modality for this task due to its effectiveness. However, most existing methods compute optical flow only between the onset and apex frames, thereby overlooking essential motion information in the apex-to-offset phase. To address this limitation, we first introduce a comprehensive motion representation, termed Magnitude-Modulated Combined Optical Flow (MM-COF), which integrates motion dynamics from both micro-expression phases into a unified descriptor suitable for direct use in recognition networks. Building upon this principle, we then propose FMANet, a novel end-to-end neural network architecture that internalizes the dual-phase analysis and magnitude modulation into learnable modules. This allows the network to adaptively fuse motion cues and focus on salient facial regions for classification. Experimental evaluations on the MMEW, SMIC, CASME-II, and SAMM datasets, widely recognized as standard benchmarks, demonstrate that our proposed MM-COF representation and FMANet outperforms existing methods, underscoring the potential of a learnable, dual-phase framework in advancing micro-expression recognition. --- ### Invisible to Humans, Triggered by Agents: Stealthy Jailbreak Attacks on Mobile Vision-Language Agents Authors: Renhua Ding, Xiao Yang, Zhengwei Fang, Jun Luo, Kun He, Jun Zhu Published: 2025-10-09 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2510.07809v4 Large Vision-Language Models (LVLMs) empower autonomous mobile agents, yet their security under realistic mobile deployment constraints remains underexplored. While agents are vulnerable to visual prompt injections, stealthily executing such attacks without requiring system-level privileges remains challenging, as existing methods rely on persistent visual manipulations that are noticeable to users. We uncover a consistent discrepancy between human and agent interactions: automated agents generate near-zero contact touch signals. Building on this insight, we propose a new attack paradigm, agent-only perceptual injection, where malicious content is exposed only during agent interactions, while remaining not readily perceived by human users. To accommodate mobile UI constraints and one-shot interaction settings, we introduce HG-IDA*, an efficient one-shot optimization method for constructing jailbreak prompts that evade LVLM safety filters. Experiments demonstrate that our approach induces unauthorized cross-app actions, achieving 82.5% planning and 75.0% execution hijack rates on GPT-4o. Our findings highlight a previously underexplored attack surface in mobile agent systems and underscore the need for defenses that incorporate interaction-level signals. --- ### Dynamic Generation of Multi-LLM Agents Communication Topologies with Graph Diffusion Models Authors: Eric Hanchen Jiang, Mengting Li, Guancheng Wan, Sophia Yin, Yuchen Wu, Xiao Liang, Xinfeng Li, Yizhou Sun, Wei Wang, Kai-Wei Chang, Ying Nian Wu Published: 2025-10-09 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2510.07799v2 The efficiency of multi-agent systems driven by large language models (LLMs) largely hinges on their communication topology. However, designing an optimal topology is a non-trivial challenge, as it requires balancing competing objectives such as task performance, communication cost, and robustness. Existing frameworks often rely on static or hand-crafted topologies, which inherently fail to adapt to diverse task requirements, leading to either excessive token consumption for simple problems or performance bottlenecks for complex ones. To address this challenge, we introduce a novel generative framework called \textit{Guided Topology Diffusion (GTD)}. Inspired by conditional discrete graph diffusion models, GTD formulates topology synthesis as an iterative construction process. At each step, the generation is steered by a lightweight proxy model that predicts multi-objective rewards (e.g., accuracy, utility, cost), enabling real-time, gradient-free optimization towards task-adaptive topologies. This iterative, guided synthesis process distinguishes GTD from single-step generative frameworks, enabling it to better navigate complex design trade-offs. We validated GTD across multiple benchmarks, and experiments show that this framework can generate highly task-adaptive, sparse, and efficient communication topologies, significantly outperforming existing methods in LLM agent collaboration. --- ### HySim-LLM: Embedding-Weighted Fine-Tuning Bounds and Manifold Denoising for Domain-Adapted LLMs Authors: Majid Jaberi-Douraki, Hossein Sholehrasa, Xuan Xu, Remya Ampadi Ramachandran Published: 2025-10-09 Categories: cs.LG, cs.IR arXiv: https://arxiv.org/abs/2510.07796v1 The extraction and standardization of pharmacokinetic (PK) information from scientific literature remain significant challenges in computational pharmacology, which limits the reliability of data-driven models in drug development. Large language models (LLMs) have achieved remarkable progress in text understanding and reasoning, yet their adaptation to structured biomedical data, such as PK tables, remains constrained by heterogeneity, noise, and domain shift. To address these limitations, we propose HySim-LLM, a unified mathematical and computational framework that integrates embedding-weighted fine-tuning and manifold-aware denoising to enhance the robustness and interpretability of LLMs. We establish two theoretical results: (1) a similarity-weighted generalization bound that quantifies adaptation performance under embedding divergence, and (2) a manifold-based denoising guarantee that bounds loss contributions from noisy or off-manifold samples. These theorems provide a principled foundation for fine-tuning LLMs in structured biomedical settings. The framework offers a mathematically grounded pathway toward reliable and interpretable LLM adaptation for biomedical and data-intensive scientific domains. --- ### HiPRAG: Hierarchical Process Rewards for Efficient Agentic Retrieval Augmented Generation Authors: Peilin Wu, Mian Zhang, Kun Wan, Wentian Zhao, Kaiyu He, Xinya Du, Zhiyu Chen Published: 2025-10-09 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2510.07794v2 Agentic RAG is a powerful technique for incorporating external information that LLMs lack, enabling better problem solving and question answering. However, suboptimal search behaviors exist widely, such as over-search (retrieving information already known) and under-search (failing to search when necessary), which leads to unnecessary overhead and unreliable outputs. Current training methods, which typically rely on outcome-based rewards in a RL framework, lack the fine-grained control needed to address these inefficiencies. To overcome this, we introduce Hierarchical Process Rewards for Efficient agentic RAG (HiPRAG), a training methodology that incorporates a fine-grained, knowledge-grounded process reward into the RL training. Our approach evaluates the necessity of each search decision on-the-fly by decomposing the agent's reasoning trajectory into discrete, parsable steps. We then apply a hierarchical reward function that provides an additional bonus based on the proportion of optimal search and non-search steps, on top of commonly used outcome and format rewards. Experiments on the Qwen2.5 and Llama-3.2 models across seven diverse QA benchmarks show that our method achieves average accuracies of 65.4% (3B) and 67.2% (7B). This is accomplished while improving search efficiency, reducing the over-search rate to just 2.3% and concurrently lowering the under-search rate. These results demonstrate the efficacy of optimizing the reasoning process itself, not just the final outcome. Further experiments and analysis demonstrate that HiPRAG shows good generalizability across a wide range of RL algorithms, model families, sizes, and types. This work demonstrates the importance and potential of fine-grained control through RL, for improving the efficiency and optimality of reasoning for search agents. --- ### LLM4Cell: A Survey of Large Language and Agentic Models for Single-Cell Biology Authors: Sajib Acharjee Dip, Adrika Zafor, Bikash Kumar Paul, Uddip Acharjee Shuvo, Muhit Islam Emon, Xuan Wang, Liqing Zhang Published: 2025-10-09 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2510.07793v3 Large language models (LLMs) and emerging agentic frameworks are beginning to transform single-cell biology by enabling natural-language reasoning, generative annotation, and multimodal data integration. However, progress remains fragmented across data modalities, architectures, and evaluation standards. LLM4Cell presents the first unified survey of 58 foundation and agentic models developed for single-cell research, spanning RNA, ATAC, multi-omic, and spatial modalities. We categorize these methods into five families-foundation, text-bridge, spatial, multimodal, epigenomic, and agentic-and map them to eight key analytical tasks including annotation, trajectory and perturbation modeling, and drug-response prediction. Drawing on over 40 public datasets, we analyze benchmark suitability, data diversity, and ethical or scalability constraints, and evaluate models across 10 domain dimensions covering biological grounding, multi-omics alignment, fairness, privacy, and explainability. By linking datasets, models, and evaluation domains, LLM4Cell provides the first integrated view of language-driven single-cell intelligence and outlines open challenges in interpretability, standardization, and trustworthy model development. --- ### GTR-Bench: Evaluating Geo-Temporal Reasoning in Vision-Language Models Authors: Qinghongbing Xie, Zhaoyuan Xia, Feng Zhu, Lijun Gong, Ziyue Li, Rui Zhao, Long Zeng Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07791v3 Recently spatial-temporal intelligence of Visual-Language Models (VLMs) has attracted much attention due to its importance for autonomous driving, embodied AI and general AI. Existing spatial-temporal benchmarks mainly focus on egocentric (first-person) perspective reasoning using images/video contexts, or geographic reasoning with graphical context (e.g., maps), thus fail to assess VLMs' geographic spatial-temporal intelligence that requires integrating both images/video and graphical context, which is crucial for real-world scenarios such as traffic management and emergency response. To address the gaps, we introduce Geo-Temporal Reasoning benchmark (GTR-Bench), a novel challenge for geographic temporal reasoning of moving targets in a large-scale camera network. GTR-Bench is more challenging as it requires multiple perspective switches between maps and videos, joint reasoning across multiple videos with non-overlapping fields of view, and inference over spatial-temporal regions that are unobserved by any video context. Evaluations of more than 10 popular VLMs on GTR-Bench show that even the best proprietary model, Gemini-2.5-Pro (34.9\%), significantly lags behind human performance (78.61\%) on geo-temporal reasoning. Moreover, our comprehensive analysis on GTR-Bench reveals three major deficiencies of current models for geo-temporal reasoning. (1) VLMs exhibit imbalanced utilization of spatial and temporal context during reasoning. (2) they show weak temporal forecasting ability, leading to poorer performance on temporally focused tasks. (3) they lack the capability to effectively align and integrate map data with multi-view video inputs. We believe GTR-Bench offers valuable insights and opens up new opportunities for research and applications in spatial-temporal intelligence. Benchmark and code will be released at https://github.com/X-Luffy/GTR-Bench. --- ### GCPO: When Contrast Fails, Go Gold Authors: Hao Wu, Wei Liu Published: 2025-10-09 Categories: cs.AI arXiv: https://arxiv.org/abs/2510.07790v1 Reinforcement learning has been widely applied to enhance the reasoning capabilities of large language models. Extending the inference limits of smaller models has become a prominent research focus. However, algorithms such as Group Relative Policy Optimization (GRPO) suffer from a clear drawback: the upper bound of a model's rollout responses is entirely determined by the model itself, preventing the acquisition of knowledge from samples that are either all incorrect or all correct. In this paper, we introduce Group Contrastive Policy Optimization (GCPO), a method that incorporates external standard reference answers. When the model cannot solve a problem, the reference answer supplies the correct response, steering the model toward an unequivocally accurate update direction. This approach offers two main advantages: (1) it improves training efficiency by fully utilizing every sample; (2) it enables the model to emulate the problem solving strategy of the reference answer during training, thereby enhancing generalization in reasoning. GCPO achieves outstanding results across multiple benchmark datasets, yielding substantial improvements over the baseline model. Our code is available at: https://github.com/AchoWu/GCPO. --- ### Weak Form Learning for Mean-Field Partial Differential Equations: an Application to Insect Movement Authors: Seth Minor, Bret D. Elderd, Benjamin Van Allen, David M. Bortz, Vanja Dukic Published: 2025-10-09 Categories: cs.LG, math.DS, q-bio.PE arXiv: https://arxiv.org/abs/2510.07786v1 Insect species subject to infection, predation, and anisotropic environmental conditions may exhibit preferential movement patterns. Given the innate stochasticity of exogenous factors driving these patterns over short timescales, individual insect trajectories typically obey overdamped stochastic dynamics. In practice, data-driven modeling approaches designed to learn the underlying Fokker-Planck equations from observed insect distributions serve as ideal tools for understanding and predicting such behavior. Understanding dispersal dynamics of crop and silvicultural pests can lead to a better forecasting of outbreak intensity and location, which can result in better pest management. In this work, we extend weak-form equation learning techniques, coupled with kernel density estimation, to learn effective models for lepidopteran larval population movement from highly sparse experimental data. Galerkin methods such as the Weak form Sparse Identification of Nonlinear Dynamics (WSINDy) algorithm have recently proven useful for learning governing equations in several scientific contexts. We demonstrate the utility of the method on a sparse dataset of position measurements of fall armyworms (Spodoptera frugiperda) obtained in simulated agricultural conditions with varied plant resources and infection status. --- ### Demystifying Deep Learning-based Brain Tumor Segmentation with 3D UNets and Explainable AI (XAI): A Comparative Analysis Authors: Ming Jie Ong, Sze Yinn Ung, Sim Kuan Goh, Jimmy Y. Zhong Published: 2025-10-09 Categories: cs.CV arXiv: https://arxiv.org/abs/2510.07785v1 The current study investigated the use of Explainable Artificial Intelligence (XAI) to improve the accuracy of brain tumor segmentation in MRI images, with the goal of assisting physicians in clinical decision-making. The study focused on applying UNet models for brain tumor segmentation and using the XAI techniques of Gradient-weighted Class Activation Mapping (Grad-CAM) and attention-based visualization to enhance the understanding of these models. Three deep learning models - UNet, Residual UNet (ResUNet), and Attention UNet (AttUNet) - were evaluated to identify the best-performing model. XAI was employed with the aims of clarifying model decisions and increasing physicians' trust in these models. We compared the performance of two UNet variants (ResUNet and AttUNet) with the conventional UNet in segmenting brain tumors from the BraTS2020 public dataset and analyzed model predictions with Grad-CAM and attention-based visualization. Using the latest computer hardware, we trained and validated each model using the Adam optimizer and assessed their performance with respect to: (i) training, validation, and inference times, (ii) segmentation similarity coefficients and loss functions, and (iii) classification performance. Notably, during the final testing phase, ResUNet outperformed the other models with respect to Dice and Jaccard similarity scores, as well as accuracy, recall, and F1 scores. Grad-CAM provided visuospatial insights into the tumor subregions each UNet model focused on while attention-based visualization provided valuable insights into the working mechanisms of AttUNet's attention modules. These results demonstrated ResUNet as the best-performing model and we conclude by recommending its use for automated brain tumor segmentation in future clinical assessments. Our source code and checkpoint are available at https://github.com/ethanong98/MultiModel-XAI-Brats2020 --- ### PLUM: Adapting Pre-trained Language Models for Industrial-scale Generative Recommendations Authors: Ruining He, Lukasz Heldt, Lichan Hong, Raghunandan Keshavan, Shifan Mao, Nikhil Mehta, Zhengyang Su, Alicia Tsai, Yueqi Wang, Shao-Chuan Wang, Xinyang Yi, Lexi Baugher et al. Published: 2025-10-09 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2510.07784v1 Large Language Models (LLMs) pose a new paradigm of modeling and computation for information tasks. Recommendation systems are a critical application domain poised to benefit significantly from the sequence modeling capabilities and world knowledge inherent in these large models. In this paper, we introduce PLUM, a framework designed to adapt pre-trained LLMs for industry-scale recommendation tasks. PLUM consists of item tokenization using Semantic IDs, continued pre-training (CPT) on domain-specific data, and task-specific fine-tuning for recommendation objectives. For fine-tuning, we focus particularly on generative retrieval, where the model is directly trained to generate Semantic IDs of recommended items based on user context. We conduct comprehensive experiments on large-scale internal video recommendation datasets. Our results demonstrate that PLUM achieves substantial improvements for retrieval compared to a heavily-optimized production model built with large embedding tables. We also present a scaling study for the model's retrieval performance, our learnings about CPT, a few enhancements to Semantic IDs, along with an overview of the training and inference methods that enable launching this framework to billions of users in YouTube. --- ### RCPU: Rotation-Constrained Error Compensation for Structured Pruning of Large Language Models Authors: Shuichiro Haruta, Kazunori Matsumoto, Zhi Li, Yanan Wang, Mori Kurokawa Published: 2025-10-09 Categories: cs.CL arXiv: https://arxiv.org/abs/2510.07782v2 In this paper, we propose a rotation-constrained compensation method to address the errors introduced by structured pruning of large language models (LLMs). LLMs are trained on massive datasets and accumulate rich semantic knowledge in their representation space. In contrast, pruning is typically carried out with only a small amount of calibration data, which makes output mismatches unavoidable. Although direct least-squares fitting can reduce such errors, it tends to overfit to the limited calibration set, destructively modifying pretrained weights. To overcome this difficulty, we update the pruned parameters under a rotation constraint. This constrained update preserves the geometry of output representations (i.e., norms and inner products) and simultaneously re-aligns the pruned subspace with the original outputs. Furthermore, in rotation-constrained compensation, removing components that strongly contribute to the principal directions of the output makes error recovery difficult. Since input dimensions with large variance strongly affect these principal directions, we design a variance-aware importance score that ensures such dimensions are preferentially kept in the pruned model. By combining this scoring rule with rotation-constrained updates, the proposed method effectively compensates errors while retaining the components likely to be more important in a geometry-preserving manner. In the experiments, we apply the proposed method to Llama-7B and Llama-2-13B, and evaluate it on WikiText2 and multiple language understanding benchmarks. The results demonstrate consistently better perplexity and task accuracy compared with existing baselines. --- ### IntentionVLA: Generalizable and Efficient Embodied Intention Reasoning for Human-Robot Interaction Authors: Yandu Chen, Kefan Gu, Yuqing Wen, Yucheng Zhao, Tiancai Wang, Liqiang Nie Published: 2025-10-09 Categories: cs.RO, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2510.07778v1 Vision-Language-Action (VLA) models leverage pretrained vision-language models (VLMs) to couple perception with robotic control, offering a promising path toward general-purpose embodied intelligence. However, current SOTA VLAs are primarily pretrained on multimodal tasks with limited relevance to embodied scenarios, and then finetuned to map explicit instructions to actions. Consequently, due to the lack of reasoning-intensive pretraining and reasoning-guided manipulation, these models are unable to perform implicit human intention reasoning required for complex, real-world interactions. To overcome these limitations, we propose \textbf{IntentionVLA}, a VLA framework with a curriculum training paradigm and an efficient inference mechanism. Our proposed method first leverages carefully designed reasoning data that combine intention inference, spatial grounding, and compact embodied reasoning, endowing the model with both reasoning and perception capabilities. In the following finetuning stage, IntentionVLA employs the compact reasoning outputs as contextual guidance for action generation, enabling fast inference under indirect instructions. Experimental results show that IntentionVLA substantially outperforms $π_0$, achieving 18\% higher success rates with direct instructions and 28\% higher than ECoT under intention instructions. On out-of-distribution intention tasks, IntentionVLA achieves over twice the success rate of all baselines, and further enables zero-shot human-robot interaction with 40\% success rate. These results highlight IntentionVLA as a promising paradigm for next-generation human-robot interaction (HRI) systems. --- ### Drift No More? Context Equilibria in Multi-Turn LLM Interactions Authors: Vardhan Dongre, Ryan A. Rossi, Viet Dac Lai, David Seunghyun Yoon, Dilek Hakkani-Tür, Trung Bui Published: 2025-10-09 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2510.07777v2 Large Language Models (LLMs) excel at single-turn tasks such as instruction following and summarization, yet real-world deployments require sustained multi-turn interactions where user goals and conversational context persist and evolve. A recurring challenge in this setting is context drift: the gradual divergence of a model's outputs from goal-consistent behavior across turns. Unlike single-turn errors, drift unfolds temporally and is poorly captured by static evaluation metrics. In this work, we present a study of context drift in multi-turn interactions and propose a simple dynamical framework to interpret its behavior. We formalize drift as the turn-wise KL divergence between the token-level predictive distributions of the test model and a goal-consistent reference model, and propose a recurrence model that interprets its evolution as a bounded stochastic process with restoring forces and controllable interventions. We instantiate this framework in both synthetic long-horizon rewriting tasks and realistic user-agent simulations such as in $τ$-Bench, measuring drift for several open-weight LLMs that are used as user simulators. Our experiments consistently reveal stable, noise-limited equilibria rather than runaway degradation, and demonstrate that simple reminder interventions reliably reduce divergence in line with theoretical predictions. Together, these results suggest that multi-turn drift can be understood as a controllable equilibrium phenomenon rather than as inevitable decay, providing a foundation for studying and mitigating context drift in extended interactions. --- ### Instance Relation Learning Network with Label Knowledge Propagation for Few-shot Multi-label Intent Detection Authors: Shiman Zhao, Shangyuan Li, Wei Chen, Tengjiao Wang, Jiahui Yao, Jiabin Zheng, Kam Fai Wong Published: 2025-10-09 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2510.07776v1 Few-shot Multi-label Intent Detection (MID) is crucial for dialogue systems, aiming to detect multiple intents of utterances in low-resource dialogue domains. Previous studies focus on a two-stage pipeline. They first learn representations of utterances with multiple labels and then use a threshold-based strategy to identify multi-label results. However, these methods rely on representation classification and ignore instance relations, leading to error propagation. To solve the above issues, we propose a multi-label joint learning method for few-shot MID in an end-to-end manner, which constructs an instance relation learning network with label knowledge propagation to eliminate error propagation. Concretely, we learn the interaction relations between instances with class information to propagate label knowledge between a few labeled (support set) and unlabeled (query set) instances. With label knowledge propagation, the relation strength between instances directly indicates whether two utterances belong to the same intent for multi-label prediction. Besides, a dual relation-enhanced loss is developed to optimize support- and query-level relation strength to improve performance. Experiments show that we outperform strong baselines by an average of 9.54% AUC and 11.19% Macro-F1 in 1-shot scenarios. --- ### The Unintended Trade-off of AI Alignment:Balancing Hallucination Mitigation and Safety in LLMs Authors: Omar Mahmoud, Ali Khalil, Buddhika Laknath Semage, Thommen George Karimpanal, Santu Rana Published: 2025-10-09 Categories: cs.CL arXiv: https://arxiv.org/abs/2510.07775v2 Hallucination in large language models (LLMs) has been widely studied in recent years, with progress in both detection and mitigation aimed at improving truthfulness. Yet, a critical side effect remains largely overlooked: enhancing truthfulness can negatively impact safety alignment. In this paper, we investigate this trade-off and show that increasing factual accuracy often comes at the cost of weakened refusal behavior. Our analysis reveals that this arises from overlapping components in the model that simultaneously encode hallucination and refusal information, leading alignment methods to suppress factual knowledge unintentionally. We further examine how fine-tuning on benign datasets, even when curated for safety, can degrade alignment for the same reason. To address this, we propose a method that disentangles refusal-related features from hallucination features using sparse autoencoders, and preserves refusal behavior during fine-tuning through subspace orthogonalization. This approach prevents hallucinations from increasing while maintaining safety alignment.We evaluate our method on commonsense reasoning tasks and harmful benchmarks (AdvBench and StrongReject). Results demonstrate that our approach preserves refusal behavior and task utility, mitigating the trade-off between truthfulness and safety. --- ### Curing Miracle Steps in LLM Mathematical Reasoning with Rubric Rewards Authors: Youliang Yuan, Qiuyang Mang, Jingbang Chen, Hong Wan, Xiaoyuan Liu, Junjielong Xu, Jen-tse Huang, Wenxuan Wang, Wenxiang Jiao, Pinjia He Published: 2025-10-09 Categories: cs.CL arXiv: https://arxiv.org/abs/2510.07774v3 In this paper, we observe that current models are susceptible to reward hacking, leading to a substantial overestimation of a model's reasoning ability. This is evidenced by a high incidence of false positives-solutions that reach the correct answer through an unsound process. Through a systematic analysis with human verification, we establish a taxonomy of these failure modes, identifying patterns like Miracle Steps-abrupt jumps to a correct output without a valid preceding derivation. Probing experiments suggest that these Miracle Steps are linked to answer-recall shortcuts, including memorization from pretraining, where the model accesses the correct answer independently of its reasoning chain. To mitigate this systemic issue, we introduce the Rubric Reward Model (RRM), a process-oriented reward function that evaluates the entire reasoning trajectory against problem-specific rubrics. The RRM explicitly penalizes logical flaws and encourages rigorous deduction. When integrated into an RL pipeline, RRM-based training consistently outperforms outcome-only supervision across four math benchmarks. Notably, it boosts Verified Pass@1024 on AIME2024 from 26.7% to 62.6% and reduces the incidence of Miracle Steps by 71%. Our work demonstrates that rewarding the solution process is crucial for building accurate and reliable models. --- ### Trajectory Conditioned Cross-embodiment Skill Transfer Authors: YuHang Tang, Yixuan Lou, Pengfei Han, Haoming Song, Xinyi Ye, Dong Wang, Bin Zhao Published: 2025-10-09 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2510.07773v1 Learning manipulation skills from human demonstration videos presents a promising yet challenging problem, primarily due to the significant embodiment gap between human body and robot manipulators. Existing methods rely on paired datasets or hand-crafted rewards, which limit scalability and generalization. We propose TrajSkill, a framework for Trajectory Conditioned Cross-embodiment Skill Transfer, enabling robots to acquire manipulation skills directly from human demonstration videos. Our key insight is to represent human motions as sparse optical flow trajectories, which serve as embodiment-agnostic motion cues by removing morphological variations while preserving essential dynamics. Conditioned on these trajectories together with visual and textual inputs, TrajSkill jointly synthesizes temporally consistent robot manipulation videos and translates them into executable actions, thereby achieving cross-embodiment skill transfer. Extensive experiments are conducted, and the results on simulation data (MetaWorld) show that TrajSkill reduces FVD by 39.6\% and KVD by 36.6\% compared with the state-of-the-art, and improves cross-embodiment success rate by up to 16.7\%. Real-robot experiments in kitchen manipulation tasks further validate the effectiveness of our approach, demonstrating practical human-to-robot skill transfer across embodiments. --- ### An approach for systematic decomposition of complex llm tasks Authors: Tianle Zhou, Jiakai Xu, Guanhong Liu, Jiaxiang Liu, Haonan Wang, Eugene Wu Published: 2025-10-09 Categories: cs.AI arXiv: https://arxiv.org/abs/2510.07772v3 Large Language Models (LLMs) suffer from reliability issues on complex tasks, as existing decomposition methods are heuristic and rely on agent or manual decomposition. This work introduces a novel, systematic decomposition framework that we call Analysis of CONstraint-Induced Complexity (ACONIC), which models the task as a constraint problem and leverages formal complexity measures to guide decomposition. On combinatorial (SAT-Bench) and LLM database querying tasks (Spider), we find that by decomposing the tasks following the measure of complexity, agent can perform considerably better. --- ### ToolLibGen: Scalable Automatic Tool Creation and Aggregation for LLM Reasoning Authors: Murong Yue, Zhiwei Liu, Liangwei Yang, Jianguo Zhang, Zuxin Liu, Haolin Chen, Ziyu Yao, Silvio Savarese, Caiming Xiong, Shelby Heinecke, Huan Wang Published: 2025-10-09 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2510.07768v1 Large Language Models (LLMs) equipped with external tools have demonstrated enhanced performance on complex reasoning tasks. The widespread adoption of this tool-augmented reasoning is hindered by the scarcity of domain-specific tools. For instance, in domains such as physics question answering, suitable and specialized tools are often missing. Recent work has explored automating tool creation by extracting reusable functions from Chain-of-Thought (CoT) reasoning traces; however, these approaches face a critical scalability bottleneck. As the number of generated tools grows, storing them in an unstructured collection leads to significant retrieval challenges, including an expanding search space and ambiguity between function-related tools. To address this, we propose a systematic approach to automatically refactor an unstructured collection of tools into a structured tool library. Our system first generates discrete, task-specific tools and clusters them into semantically coherent topics. Within each cluster, we introduce a multi-agent framework to consolidate scattered functionalities: a code agent refactors code to extract shared logic and creates versatile, aggregated tools, while a reviewing agent ensures that these aggregated tools maintain the complete functional capabilities of the original set. This process transforms numerous question-specific tools into a smaller set of powerful, aggregated tools without loss of functionality. Experimental results demonstrate that our approach significantly improves tool retrieval accuracy and overall reasoning performance across multiple reasoning tasks. Furthermore, our method shows enhanced scalability compared with baselines as the number of question-specific increases. --- ### FedLAM: Low-latency Wireless Federated Learning via Layer-wise Adaptive Modulation Authors: Linping Qu, Shenghui Song, Chi-Ying Tsui Published: 2025-10-09 Categories: cs.LG arXiv: https://arxiv.org/abs/2510.07766v1 In wireless federated learning (FL), the clients need to transmit the high-dimensional deep neural network (DNN) parameters through bandwidth-limited channels, which causes the communication latency issue. In this paper, we propose a layer-wise adaptive modulation scheme to save the communication latency. Unlike existing works which assign the same modulation level for all DNN layers, we consider the layers' importance which provides more freedom to save the latency. The proposed scheme can automatically decide the optimal modulation levels for different DNN layers. Experimental results show that the proposed scheme can save up to 73.9% of communication latency compared with the existing schemes. --- ### From Noisy to Native: LLM-driven Graph Restoration for Test-Time Graph Domain Adaptation Authors: Xiangwei Lv, JinLuan Yang, Wang Lin, Jingyuan Chen, Beishui Liao Published: 2025-10-09 Categories: cs.AI arXiv: https://arxiv.org/abs/2510.07762v1 Graph domain adaptation (GDA) has achieved great attention due to its effectiveness in addressing the domain shift between train and test data. A significant bottleneck in existing graph domain adaptation methods is their reliance on source-domain data, which is often unavailable due to privacy or security concerns. This limitation has driven the development of Test-Time Graph Domain Adaptation (TT-GDA), which aims to transfer knowledge without accessing the source examples. Inspired by the generative power of large language models (LLMs), we introduce a novel framework that reframes TT-GDA as a generative graph restoration problem, "restoring the target graph to its pristine, source-domain-like state". There are two key challenges: (1) We need to construct a reasonable graph restoration process and design an effective encoding scheme that an LLM can understand, bridging the modality gap. (2) We need to devise a mechanism to ensure the restored graph acquires the intrinsic features of the source domain, even without access to the source data. To ensure the effectiveness of graph restoration, we propose GRAIL, that restores the target graph into a state that is well-aligned with the source domain. Specifically, we first compress the node representations into compact latent features and then use a graph diffusion process to model the graph restoration process. Then a quantization module encodes the restored features into discrete tokens. Building on this, an LLM is fine-tuned as a generative restorer to transform a "noisy" target graph into a "native" one. To further improve restoration quality, we introduce a reinforcement learning process guided by specialized alignment and confidence rewards. Extensive experiments demonstrate the effectiveness of our approach across various datasets. --- ### Test-Time Reasoners Are Strategic Multiple-Choice Test-Takers Authors: Nishant Balepur, Atrey Desai, Rachel Rudinger Published: 2025-10-09 Categories: cs.CL arXiv: https://arxiv.org/abs/2510.07761v2 Large language models (LLMs) now give reasoning before answering, excelling in tasks like multiple-choice question answering (MCQA). Yet, a concern is that LLMs do not solve MCQs as intended, as work finds LLMs sans reasoning succeed in MCQA without using the question, i.e., choices-only. Such partial-input success is often linked to trivial shortcuts, but reasoning traces could reveal if choices-only strategies are truly shallow. To examine these strategies, we have reasoning LLMs solve MCQs in full and choices-only inputs; test-time reasoning often boosts accuracy in full and in choices-only, half the time. While possibly due to shallow shortcuts, choices-only success is barely affected by the length of reasoning traces, and after finding traces pass faithfulness tests, we show they use less problematic strategies like inferring missing questions. In all, we challenge claims that partial-input success is always a flaw, so we propose how reasoning traces could separate problematic data from less problematic reasoning. --- ### VAO: Validation-Aligned Optimization for Cross-Task Generative Auto-Bidding Authors: Yiqin Lv, Zhiyu Mou, Miao Xu, Jinghao Chen, Qi Wang, Yixiu Mao, Yun Qu, Rongquan Bai, Chuan Yu, Jian Xu, Bo Zheng, Xiangyang Ji Published: 2025-10-09 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2510.07760v3 Generative auto-bidding has demonstrated strong performance in online advertising, yet it often suffers from data scarcity in small-scale settings with limited advertiser participation. While cross-task data sharing is a natural remedy to mitigate this issue, naive approaches often introduce gradient bias due to distribution shifts across different tasks, and existing methods are not readily applicable to generative auto-bidding. In this paper, we propose Validation-Aligned Optimization (VAO), a principled data-sharing method that adaptively reweights cross-task data contributions based on validation performance feedback. Notably, VAO aligns training dynamics to prioritize updates that improve generalization on the target task, effectively leveraging auxiliary data and mitigating gradient bias. Building on VAO, we introduce a unified generative autobidding framework that generalizes across multiple tasks using a single model and all available task data. Extensive experiments on standard auto-bidding benchmarks validate the effectiveness of our approach. --- ### Rényi Sharpness: A Novel Sharpness that Strongly Correlates with Generalization Authors: Qiaozhe Zhang, Jun Sun, Ruijie Zhang, Yingzhuang Liu Published: 2025-10-09 Categories: cs.LG arXiv: https://arxiv.org/abs/2510.07758v4 Sharpness (of the loss minima) is widely believed to be a good indicator of generalization of neural networks. Unfortunately, the correlation between existing sharpness measures and generalization is not as strong as expected, and sometimes even contradiction occurs. To address this problem, a key observation in this paper is: what really matters for generalization is the average spread (or unevenness) of the spectrum of loss Hessian $\mathbf{H}$. For this reason, conventional sharpness measures, such as trace sharpness $\operatorname{tr}(\mathbf{H})$, which cares about the average value of the spectrum, or max-eigenvalue sharpness $λ_{\max}(\mathbf{H})$, which concerns the maximum spread of the spectrum, are not sufficient to well predict generalization. To characterize the average spread of the Hessian spectrum, we leverage the notion of Rényi entropy in information theory, which captures the unevenness of a probability vector and can thus be extended to a general non-negative vector, such as the Hessian spectrum at loss minima. Specifically, we propose Rényi sharpness, defined as the negative of the Rényi entropy of loss Hessian $\mathbf{H}$. Extensive experiments demonstrate that Rényi sharpness exhibits strong and consistent correlation with generalization in various scenarios. Moreover, two generalization bounds with respect to Rényi sharpness are established by exploiting its desirable reparametrization invariance property. Finally, as an initial attempt to exploit Rényi sharpness for regularization, Rényi Sharpness Aware Minimization (RSAM) is proposed, where a variant of Rényi sharpness is used as the regularizer. RSAM is competitive with state-of-the-art SAM algorithms and far better than conventional SAM based on max-eigenvalue sharpness. ---