语言大模型 / LLM - arXivDaily 专题

2505.22255 2025-12-23 cs.LG cs.CL 62%

Kronecker Factorization Improves Efficiency and Interpretability of Sparse Autoencoders

克罗内克分解提升稀疏自编码器的效率与可解释性

Vadim Kurochkin, Yaroslav Aksenov, Daniil Laptev, Daniil Gavrilov, Nikita Balagansky

机构 * T-Tech

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.CL、cs.LG

AI总结 KronSAE通过克罗内克分解减少计算开销，同时引入mAND提升稀疏自编码器的可解释性与性能。

2508.00969 2025-12-17 cs.LG cs.AI 62%

Masked Omics Modeling for Multimodal Representation Learning across Histopathology and Molecular Profiles

掩码组学建模用于病理学与分子特征的多模态表示学习

Lucas Robinet, Ahmad Berjaoui, Elizabeth Cohen-Jonathan Moyal

机构 * Oncopole（奥恩波尔）； IRT Saint Exupéry（国际研究与技术圣埃克苏佩里）； INSERM Cancer Research Center of Toulouse（里沃利癌症研究中心）； Toulouse（图卢兹）

专题命中知识编辑与模型理解：foundation model(abstract)；分类 cs.AI、cs.LG

AI总结 MORPHEUS通过整合病理学图像和多组学数据，提出了一种多模态预训练策略，以提升癌症研究中的跨模态表示学习能力。

2512.12842 2025-12-16 cs.RO cs.AI cs.LG 62%

SAGA: Open-World Mobile Manipulation via Structured Affordance Grounding

SAGA：通过结构化可及性 grounding 实现开放世界移动操作

Kuan Fang, Yuxin Chen, Xinghao Zhu, Farzad Niroui, Lingfeng Sun, Jiuguang Wang

机构 * RAI Institute（RAI研究院）

专题命中知识编辑与模型理解：foundation model(abstract)；分类 cs.AI、cs.LG

AI总结 SAGA通过结构化可及性接地实现开放世界移动操作，能有效处理多种任务形式并实现零样本执行。

Comments 9 pages, 7 figures

2510.03265 2025-11-25 cs.LG cs.AI 62%

MindCraft: How Concept Trees Take Shape In Deep Models

MindCraft: 深度模型中概念树如何形成

Bowei Tian, Yexiao He, Wanghao Ye, Ziyao Wang, Meng Liu, Ang Li

机构 * University of Maryland, College Park（马里兰大学 College Park 分校）

专题命中知识编辑与模型理解：foundation model(abstract)；分类 cs.AI、cs.LG

AI总结 MindCraft通过概念树框架揭示深度模型中概念的层次结构和分离过程，为可解释AI提供新的分析方法。

2511.13653 2025-11-18 cs.LG cs.AI 62%

Weight-sparse transformers have interpretable circuits

Leo Gao, Achyuta Rajaram, Jacob Coxon, Soham V. Govande, Bowen Baker, Dan Mossing

机构 * OpenAI

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.AI、cs.LG

2510.19172 2025-11-18 cs.CL cs.AI 62%

When Facts Change: Probing LLMs on Evolving Knowledge with evolveQA

Nishanth Sridhar Nakshatri, Shamik Roy, Manoj Ghuhan Arivazhagan, Hanhan Zhou, Vinayshekhar Bannihatti Kumar, Rashmi Gangadharaiah

机构 * Purdue University（普渡大学）； AWS AI Labs（AWS人工智能实验室）

专题命中知识编辑与模型理解：LLM(abstract)；分类 cs.CL、cs.AI

Comments Under submission

2511.12154 2025-11-18 cs.LG cs.AI 62%

Open Banking Foundational Model: Learning Language Representations from Few Financial Transactions

Gustavo Polleti, Marlesson Santana, Eduardo Fontes

机构 * Trustly

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.AI、cs.LG

2105.06421 2025-11-18 cs.CV cs.AI cs.LG 62%

Using Self-Supervised Auxiliary Tasks to Improve Fine-Grained Facial Representation

Mahdi Pourmirzaei, Gholam Ali Montazer, Farzaneh Esmaili

专题命中知识编辑与模型理解：pretraining(abstract)；分类 cs.AI、cs.LG

2511.07006 2025-11-11 cs.LG cs.AI 62%

S$^2$Drug: Bridging Protein Sequence and 3D Structure in Contrastive Representation Learning for Virtual Screening

Bowei He, Bowen Gao, Yankai Chen, Yanyan Lan, Chen Ma, Philip S. Yu, Ya-Qin Zhang, Wei-Ying Ma

专题命中知识编辑与模型理解：pretraining(abstract)；分类 cs.AI、cs.LG

Comments Accepted by AAAI 2026 Main Technical Track

2510.17959 2025-11-11 astro-ph.IM cs.AI cs.LG 62%

Universal Spectral Tokenization via Self-Supervised Panchromatic Representation Learning

Jeff Shen, Francois Lanusse, Liam Holden Parker, Ollie Liu, Tom Hehir, Leopoldo Sarra, Lucas Meyer, Micah Bowles, Sebastian Wagner-Carena, Sebastian Wagner-Carena, Helen Qu, Siavash Golkar, Alberto Bietti, Hatim Bourfoune, Nathan Cassereau, Pierre Cornette, Keiya Hirashima, Geraud Krawezik, Ruben Ohana, Nicholas Lourie, Michael McCabe, Rudy Morel, Payel Mukhopadhyay, Mariel Pettee, Bruno Régaldo-Saint Blancard, Kyunghyun Cho, Miles Cranmer, Shirley Ho

专题命中知识编辑与模型理解：foundation model(abstract)；分类 cs.AI、cs.LG

Comments Accepted at NeurIPS 2025 Machine Learning and the Physical Sciences Workshop; v2: added collaboration

2410.01334 2025-11-11 cs.CL cs.AI 62%

Skill Path: Unveiling Language Skills from Circuit Graphs

Hang Chen, Jiaying Zhu, Xinyu Yang, Wenya Wang

机构 * Hang Chen School of Computer Science and Technology Xi’an Jiaotong University（陈hang 王校计算机科学与技术学院西安交通大学）； Jiaying Zhu School of Computer Science and Engineering The Chinese University of Hong Kong（朱继燕王校计算机科学与工程学院香港中文大学）； Xinyu Yang School of Computer Science and Technology Xi’an Jiaotong University（杨新宇王校计算机科学与技术学院西安交通大学）； Wenya Wang School of Computer Science and Engineering Nanyang Technological University（王文雅王校计算机科学与工程学院新加坡国立大学）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.CL、cs.AI

Comments accepted by AAAI 2026 (oral)

2505.02850 2025-11-07 cs.CL cs.AI cs.CY cs.DB 62%

Harnessing Structured Knowledge: A Concept Map-Based Approach for High-Quality Multiple Choice Question Generation with Effective Distractors

Nicy Scaria, Silvester John Joseph Kennedy, Diksha Seth, Ananya Thakur, Deepak Subramani

机构 * Indian Institute of Science（印度科学研究院）

专题命中知识编辑与模型理解：LLM(abstract)；分类 cs.CL、cs.AI

Comments Accepted to ECAI 2025

Journal ref The European Conference on Artificial Intelligence. 413 (2025). pp 4089--4096. IOS Press

2510.23458 2025-10-29 cs.CL cs.AI 62%

BrowseConf: Confidence-Guided Test-Time Scaling for Web Agents

Litu Ou, Kuan Li, Huifeng Yin, Liwen Zhang, Zhongwang Zhang, Xixi Wu, Rui Ye, Zile Qiao, Pengjun Xie, Jingren Zhou, Yong Jiang

机构 * Tongyi Lab（通义实验室）； Alibaba Group（阿里巴巴集团）

专题命中知识编辑与模型理解：LLM(abstract)；分类 cs.CL、cs.AI

Comments 25 pages

2510.20229 2025-10-24 cs.CV cs.AI cs.CL 62%

Why LVLMs Are More Prone to Hallucinations in Longer Responses: The Role of Context

Ge Zheng, Jiaye Qian, Jiajin Tang, Sibei Yang

机构 * School of Computer Science and Engineering, Sun Yat-sen University（中山大学计算机科学与工程学院）； ShanghaiTech University（上海理工大学）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.CL、cs.AI

Journal ref Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025, pp. 4101-4113

2502.09863 2025-10-20 cs.LG cs.CL stat.ML 62%

Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models

Dhruva Karkada, James B. Simon, Yasaman Bahri, Michael R. DeWeese

机构 * UC Berkeley（伯克利大学）； Imbue（Imbue公司）； Google DeepMind（谷歌DeepMind）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.CL、cs.LG

Comments 26 pages, 8 figures

2509.18111 2025-10-14 cs.LG cs.AI cs.CV 62%

Prompt Optimization Meets Subspace Representation Learning for Few-shot Out-of-Distribution Detection

Faizul Rakib Sayem, Shahana Ibrahim

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.AI、cs.LG

2507.03123 2025-10-14 cs.CV cs.CL cs.LG 62%

Investigating VLM Hallucination from a Cognitive Psychology Perspective: A First Step Toward Interpretation with Intriguing Observations

Xiangrui Liu, Man Luo, Agneet Chatterjee, Hua Wei, Chitta Baral, Yezhou Yang

机构 * Arizona State University（亚利桑那州立大学）； Intel Lab（英特尔实验室）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.CL、cs.LG

2510.06397 2025-10-09 cs.LG cs.AI 62%

Geometry-Aware Backdoor Attacks: Leveraging Curvature in Hyperbolic Embeddings

Ali Baheri

专题命中知识编辑与模型理解：foundation model(abstract)；分类 cs.AI、cs.LG

2505.24492 2025-10-08 cs.LG cs.AI 62%

Object Centric Concept Bottlenecks

David Steinmann, Wolfgang Stammer, Antonia Wüst, Kristian Kersting

机构 * Computer Science Department, TU Darmstadt（图宾根大学计算机科学系）； Hessian Center for AI (hessian.AI)（海德堡人工智能中心）； German Research Center for AI (DFKI)（德国人工智能研究中心）； Centre for Cognitive Science, TU Darmstadt（图宾根大学认知科学中心）

专题命中知识编辑与模型理解：foundation model(abstract)；分类 cs.AI、cs.LG

2503.11846 2025-10-08 eess.IV cs.AI cs.CV cs.LG q-bio.QM 62%

A Graph-Based Framework for Interpretable Whole Slide Image Analysis

Alexander Weers, Alexander H. Berger, Laurin Lux, Peter Schüffler, Daniel Rueckert, Johannes C. Paetzold

机构 * School of Computation, Information and Technology, Technical University of Munich（慕尼黑技术大学计算、信息与技术学院）； Department of Computing, Imperial College London（伦敦帝国学院计算机系）； Munich Center for Machine Learning（慕尼黑机器学习中心）； Munich Data Science Institute, Technical University of Munich（慕尼黑数据科学研究所）； Institute of Pathology, TUM School of Medicine and Health, Technical University of Munich（慕尼黑技术大学医学与健康学院病理学研究所）； Weill Cornell Medicine, Cornell University（韦尔医学院，康奈尔大学）

专题命中知识编辑与模型理解：foundation model(abstract)；分类 cs.AI、cs.LG

Comments 15 pages, 5 figures

2510.03442 2025-10-07 cs.LG cs.AI cs.MA 62%

The Argument is the Explanation: Structured Argumentation for Trust in Agents

Ege Cakar, Per Ola Kristensson

专题命中知识编辑与模型理解：LLM(abstract)；分类 cs.AI、cs.LG

Comments 8 pages, 4 figures, 6 tables, submitted to IAAI-26

2510.01494 2025-10-06 cs.LG cs.AI 62%

Understanding Adversarial Transfer: Why Representation-Space Attacks Fail Where Data-Space Attacks Succeed

Isha Gupta, Rylan Schaeffer, Joshua Kazdan, Ken Ziyu Liu, Sanmi Koyejo

机构 * ETH Zürich（苏黎世联邦理工学院）； Stanford CS（斯坦福大学计算机科学系）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.AI、cs.LG

2505.24858 2025-10-03 cs.CL cs.LG 62%

MetaFaith: Faithful Natural Language Uncertainty Expression in LLMs

Gabrielle Kaili-May Liu, Gal Yona, Avi Caciularu, Idan Szpektor, Tim G. J. Rudner, Arman Cohan

机构 * Yale University（耶鲁大学）； Google Research（谷歌研究）； University of Toronto（多伦多大学）

专题命中知识编辑与模型理解：prompting(abstract)；分类 cs.CL、cs.LG

Comments EMNLP 2025

2510.01004 2025-10-02 cs.CV cs.AI cs.LG 62%

TextCAM: Explaining Class Activation Map with Text

Qiming Zhao, Xingjian Li, Xiaoyu Cao, Xiaolong Wu, Min Xu

机构 * Carnegie Mellon University（卡内基梅隆大学）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.AI、cs.LG

2402.19411 2025-10-02 cs.IR cs.CL cs.LG 62%

PaECTER: Patent-level Representation Learning using Citation-informed Transformers

Mainak Ghosh, Michael E. Rose, Sebastian Erhardt, Erik Buunk, Dietmar Harhoff

机构 * Max Planck Institute for Innovation and Competition（马克斯·普朗克创新与竞争研究所）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.CL、cs.LG

Comments 8 pages, 3 figures, 4 tables

2505.16002 2025-10-01 cs.CL cs.AI 62%

Causal Interventions Reveal Shared Structure Across English Filler-Gap Constructions

Sasha Boguraev, Christopher Potts, Kyle Mahowald

机构 * The University of Texas at Austin（德克萨斯大学奥斯汀分校）； Stanford University（斯坦福大学）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.CL、cs.AI

Comments 22 pages, 21 figures, 10 tables; EMNLP (Main) 2025

2509.11625 2025-10-01 cs.LG cs.AI cs.CR 62%

Inducing Uncertainty on Open-Weight Models for Test-Time Privacy in Image Recognition

Muhammad H. Ashiq, Peter Triantafillou, Hung Yun Tseng, Grigoris G. Chrysos

机构 * University of Wisconsin-Madison（威斯康星大学麦迪逊分校）； University of Warwick（沃里克大学）

专题命中知识编辑与模型理解：pretraining(abstract)；分类 cs.AI、cs.LG

2506.08885 2025-09-30 cs.CL cs.LG 62%

AdversariaL attacK sAfety aLIgnment(ALKALI): Safeguarding LLMs through GRACE: Geometric Representation-Aware Contrastive Enhancement- Introducing Adversarial Vulnerability Quality Index (AVQI)

Danush Khanna, Gurucharan Marthi Krishna Kumar, Basab Ghosh, Yaswanth Narsupalli, Vinija Jain, Vasu Sharma, Aman Chadha, Amitava Das

专题命中知识编辑与模型理解：preference optimization(abstract)；分类 cs.CL、cs.LG

2406.01651 2025-09-30 q-bio.QM cs.AI cs.LG q-bio.BM 62%

FusionDTI: Fine-grained Binding Discovery with Token-level Fusion for Drug-Target Interaction

Zhaohan Meng, Zaiqiao Meng, Ke Yuan, Iadh Ounis

机构 * School of Computing Science（计算科学学院）； School of Cancer Sciences（癌症科学学院）； Cancer Research UK Scotland Institute（英国癌症研究苏格兰研究所）； University of Glasgow（格拉斯哥大学）

专题命中知识编辑与模型理解：language model(abstract)；分类 cs.AI、cs.LG

Comments Findings of EMNLP 2025

2509.20088 2025-09-25 cs.CL cs.AI 62%

Causal Understanding by LLMs: The Role of Uncertainty

Oscar Lithgow-Serrano, Vani Kanjirangat, Alessandro Antonucci

机构 * SUPSI, IDSIA（瑞士SUPSI和IDSIA）

专题命中知识编辑与模型理解：pretraining(abstract)；分类 cs.CL、cs.AI

Comments Accepted in second UncertaiNLP workshop at EMNLP 2025

1. 知识编辑与模型理解 7565 篇

Kronecker Factorization Improves Efficiency and Interpretability of Sparse Autoencoders

Masked Omics Modeling for Multimodal Representation Learning across Histopathology and Molecular Profiles

SAGA: Open-World Mobile Manipulation via Structured Affordance Grounding

MindCraft: How Concept Trees Take Shape In Deep Models

Weight-sparse transformers have interpretable circuits

When Facts Change: Probing LLMs on Evolving Knowledge with evolveQA

Open Banking Foundational Model: Learning Language Representations from Few Financial Transactions

Using Self-Supervised Auxiliary Tasks to Improve Fine-Grained Facial Representation

S$^2$Drug: Bridging Protein Sequence and 3D Structure in Contrastive Representation Learning for Virtual Screening

Universal Spectral Tokenization via Self-Supervised Panchromatic Representation Learning

Skill Path: Unveiling Language Skills from Circuit Graphs

Harnessing Structured Knowledge: A Concept Map-Based Approach for High-Quality Multiple Choice Question Generation with Effective Distractors

BrowseConf: Confidence-Guided Test-Time Scaling for Web Agents

Why LVLMs Are More Prone to Hallucinations in Longer Responses: The Role of Context

Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models

Prompt Optimization Meets Subspace Representation Learning for Few-shot Out-of-Distribution Detection

Investigating VLM Hallucination from a Cognitive Psychology Perspective: A First Step Toward Interpretation with Intriguing Observations

Geometry-Aware Backdoor Attacks: Leveraging Curvature in Hyperbolic Embeddings

Object Centric Concept Bottlenecks

A Graph-Based Framework for Interpretable Whole Slide Image Analysis

The Argument is the Explanation: Structured Argumentation for Trust in Agents

Understanding Adversarial Transfer: Why Representation-Space Attacks Fail Where Data-Space Attacks Succeed

MetaFaith: Faithful Natural Language Uncertainty Expression in LLMs

TextCAM: Explaining Class Activation Map with Text

PaECTER: Patent-level Representation Learning using Citation-informed Transformers

Causal Interventions Reveal Shared Structure Across English Filler-Gap Constructions

Inducing Uncertainty on Open-Weight Models for Test-Time Privacy in Image Recognition

AdversariaL attacK sAfety aLIgnment(ALKALI): Safeguarding LLMs through GRACE: Geometric Representation-Aware Contrastive Enhancement- Introducing Adversarial Vulnerability Quality Index (AVQI)

FusionDTI: Fine-grained Binding Discovery with Token-level Fusion for Drug-Target Interaction

Causal Understanding by LLMs: The Role of Uncertainty