语言大模型 / LLM - arXivDaily 专题

2510.04933 2025-10-07 cs.CL cs.AI cs.IT cs.LG cs.NE math.IT 89%

The Geometry of Truth: Layer-wise Semantic Dynamics for Hallucination Detection in Large Language Models

Amir Hameed Mir

机构 * Sirraya Labs（Sirraya实验室）

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

Comments Comments: 14 pages, 14 figures, 5 tables. Code available at: https://github.com/sirraya-tech/Sirraya_LSD_Code

2510.03799 2025-10-07 cs.CL cs.AI cs.CY 84%

Mechanistic Interpretability of Socio-Political Frames in Language Models

Hadi Asghari, Sami Nenno

机构 * Technische Universität Berlin, Berlin, Germany Humboldt Institute for Internet \& Society, Berlin, Germany

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

Comments Peer-reviewed and presented at Advances in Interpretable Machine Learning and Artificial Intelligence (AIMLAI) Workshop at ECML/PKDD 2024

2505.18562 2025-10-07 cs.CL cs.AI 84%

From Word to World: Evaluate and Mitigate Culture Bias in LLMs via Word Association Test

Xunlian Dai, Li Zhou, Benyou Wang, Haizhou Li

机构 * The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））； Shenzhen Research Institute of Big Data（深圳大数据研究院）

专题命中知识编辑与模型理解：large language model(abstract,comments)；language model(abstract,comments)；LLM(abstract)；prompting(abstract)

Comments Cultural Analysis, Cultural Alignment, Word Association Test, Large Language Models. Accepted by EMNLP 2025 (Oral)

2412.10246 2025-10-07 cs.LG 83%

Detecting LLM Hallucination Through Layer-wise Information Deficiency: Analysis of Ambiguous Prompts and Unanswerable Questions

Hazel Kim, Tom A. Lamb, Adel Bibi, Philip Torr, Yarin Gal

机构 * University of Oxford（牛津大学）

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

Comments Accepted to EMNLP(main)2025

2509.05801 2025-10-07 cs.LG cs.AI 82%

time2time: Causal Intervention in Hidden States to Simulate Rare Events in Time Series Foundation Models

Debdeep Sanyal, Aaryan Nagpal, Dhruv Kumar, Murari Mandal, Saurabh Deshpande

机构 * Birla AI Labs, Office of Ananya Birla（Birla AI实验室，Ananya Birla办公室）； BITS Pilani ； KIIT Bhubaneswar（KIIT巴尔班格斯）

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

Journal ref NeurIPS 2025 Workshop on Recent Advances in Time Series Foundation Models (BERT2S)

2501.12266 2025-10-07 cs.CV cs.AI cs.CL 81%

CBVLM: Training-free Explainable Concept-based Large Vision Language Models for Medical Image Classification

Cristiano Patrício, Isabel Rio-Torto, Jaime S. Cardoso, Luís F. Teixeira, João C. Neves

机构 * INESC TEC ； NOVA LINCS ； Universidade da Beira Interior（贝拉蒙特大学）； Universidade do Porto（波尔图大学）

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

Comments Accepted for publication in Computers in Biology and Medicine

2504.06303 2025-10-07 cs.CY cs.AI cs.CL cs.LG 80%

On the Effectiveness and Generalization of Race Representations for Debiasing High-Stakes Decisions

Dang Nguyen, Chenhao Tan

机构 * Department of Computer Science University of Chicago（计算机科学系芝加哥大学）

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

Comments 21 pages, 15 figures, 14 tables. Accepted as a conference paper at COLM 2025. Camera-ready

2510.04819 2025-10-07 cs.CV cs.CL 79%

Visual Representations inside the Language Model

Benlin Liu, Amita Kamath, Madeleine Grunde-McLaughlin, Winson Han, Ranjay Krishna

机构 * University of Washington（华盛顿大学）； University of California Los Angeles（加州大学洛杉矶分校）； Allen Institute for AI（人工智能研究院）

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

Comments Accepted to COLM 2025

2510.03911 2025-10-07 cs.LG 79%

THEMIS: Unlocking Pretrained Knowledge with Foundation Model Embeddings for Anomaly Detection in Time Series

Yadav Mahesh Lorik, Kaushik Sarveswaran, Nagaraj Sundaramahalingam, Aravindakumar Venugopalan

机构 * Comcast India Engineering Center（Comcast印度工程中心）

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

Comments Oral Presentation. AI4TS Workshop, IJCAI'25

2405.14715 2025-10-07 cs.CV cs.AI 79%

Towards Cross-modal Backward-compatible Representation Learning for Vision-Language Models

Young Kyun Jang, Ser-nam Lim

机构 * Google DeepMind（谷歌DeepMind）； University of Central Florida（中央佛罗里达大学）

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

2510.04506 2025-10-07 cs.CL cs.AI cs.IR 79%

GRACE: Generative Representation Learning via Contrastive Policy Optimization

Jiashuo Sun, Shixuan Liu, Zhaochen Su, Xianrui Zhong, Pengcheng Jiang, Bowen Jin, Peiran Li, Weijia Shi, Jiawei Han

机构 * University of Illinois Urbana–Champaign（伊利诺伊大学厄巴纳-香槟分校）； Australian National University（澳大利亚国立大学）； Hong Kong University of Science and Technology（香港科学与技术大学）； University of Wisconsin–Madison（威斯康星大学麦迪逊分校）； University of Washington（华盛顿大学）

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

Comments 23 pages, 7 figures, 7 tables

2510.03767 2025-10-07 cs.CV 78%

CoPA: Hierarchical Concept Prompting and Aggregating Network for Explainable Diagnosis

Yiheng Dong, Yi Lin, Xin Yang

机构 * School of Electronic Information and Communications, Huazhong University of Science and Technology, Wuhan, China（电子信息学院，华中科技大学，武汉）； Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China（计算机科学与工程系，香港科技大学，香港）

专题命中知识编辑与模型理解：prompting(title,abstract)

Comments Accepted by MICCAI2025

2510.04849 2025-10-07 cs.CL 77%

When Models Lie, We Learn: Multilingual Span-Level Hallucination Detection with PsiloQA

Elisei Rykov, Kseniia Petrushina, Maksim Savkin, Valerii Olisov, Artem Vazhentsev, Kseniia Titova, Alexander Panchenko, Vasily Konovalov, Julia Belikova

机构 * Skoltech ； AIRI ； MWS AI ； Sber AI Lab ； Moscow Institute of Physics and Technology（莫斯科物理技术学院）

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

2510.04439 2025-10-07 cs.CL 77%

On the Role of Unobserved Sequences on Sample-based Uncertainty Quantification for LLMs

Lucie Kunitomo-Jacquin, Edison Marrese-Taylor, Ken Fukuda

机构 * National Institute of Advanced Industrial Science and Technology (AIST)（国家先进工业科学与技术研究院）

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

Comments Accepted to UncertaiNLP workshop of EMNLP 2025

2510.03659 2025-10-07 cs.LG cs.AI cs.CL stat.ML 75%

Does higher interpretability imply better utility? A Pairwise Analysis on Sparse Autoencoders

Xu Wang, Yan Hu, Benyou Wang, Difan Zou

机构 * School of Computing and Data Science, The University of Hong Kong（计算与数据科学学院，香港大学）； School of Data Science, The Chinese University of Hong Kong, Shenzhen（数据科学学院，香港中文大学（深圳））

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

Comments 24 pages

2507.20836 2025-10-07 cs.LG cs.AI 73%

First Hallucination Tokens Are Different from Conditional Ones

Jakob Snel, Seong Joon Oh

机构 * University of Tübingen, Germany（图宾根大学）； Tübingen AI Center（图宾根人工智能中心）

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

Comments 4.5 pages, 3 figures, Dataset, Knowledge Paper, Hallucination, Trustworthiness

2510.04285 2025-10-07 cs.CL cond-mat.stat-mech cs.LG stat.ML 73%

Probing Geometry of Next Token Prediction Using Cumulant Expansion of the Softmax Entropy

Karthik Viswanathan, Sang Eon Park

机构 * Institute of Physics, University of Amsterdam（阿姆斯特丹大学物理研究所）； Massachusetts Institute of Technology（麻省理工学院）

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

Comments 14 pages, 7 figures. Poster at HiLD 2025: 3rd Workshop on High-dimensional Learning Dynamics

2510.03351 2025-10-07 cs.LG cs.AI eess.IV 73%

Interpretable Neuropsychiatric Diagnosis via Concept-Guided Graph Neural Networks

Song Wang, Zhenyu Lei, Zhen Tan, Jundong Li, Javier Rasero, Aiying Zhang, Chirag Agarwal

机构 * University of Central Florida（中央佛罗里达大学）； University of Virginia（弗吉尼亚大学）

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

2510.03295 2025-10-07 cs.CV cs.CL cs.LG 73%

Multimodal Arabic Captioning with Interpretable Visual Concept Integration

Passant Elchafei, Amany Fashwan

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

2403.03645 2025-10-07 cs.AI 70%

Graph Generation Powered with LLMs for Boosting Multivariate Time-Series Representation Learning

Yucheng Wang, Min Wu, Ruibing Jin, Xiaoli Li, Lihua Xie, Zhenghua Chen

机构 * Institute for Infocomm Research, A ∗ STAR, Singapore（信息通信研究所，A*STAR，新加坡）； School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore（电气电子工程学院，南洋理工大学，新加坡）； College of Computing and Data Science, Nanyang Technological University, Singapore（计算与数据科学学院，南洋理工大学，新加坡）

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

2402.10612 2025-10-07 cs.CL 70%

Rowen: Adaptive Retrieval-Augmented Generation for Hallucination Mitigation in LLMs

Hanxing Ding, Liang Pang, Zihao Wei, Huawei Shen, Xueqi Cheng

机构 * State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences（人工智能安全国家重点实验室，计算技术研究所，中国科学院）

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

Comments Accepted at SIGIR-AP 2025

2506.16876 2025-10-07 cs.SE 67%

Revolutionizing Validation and Verification: Explainable Testing Methodologies for Intelligent Automotive Decision-Making Systems

Halit Eris, Stefan Wagner

专题命中知识编辑与模型理解：large language model(abstract)；language model(abstract)

Comments Preprint to be published at SE4ADS

Journal ref 2025 IEEE/ACM 1st International Workshop on Software Engineering for Autonomous Driving Systems (SE4ADS), Ottawa, ON, Canada, 2025, pp. 34-37

2507.09747 2025-10-07 cs.NE 67%

BrainFLORA: Uncovering Brain Concept Representation via Multimodal Neural Embeddings

Dongyang Li, Haoyang Qin, Mingyang Wu, Chen Wei, Quanying Liu

专题命中知识编辑与模型理解：large language model(abstract)；language model(abstract)

Comments ACM MM 2025

2408.16357 2025-10-07 cs.CV 67%

Law of Vision Representation in MLLMs

Shijia Yang, Bohan Zhai, Quanzeng You, Jianbo Yuan, Hongxia Yang, Chenfeng Xu

机构 * Stanford University（斯坦福大学）； UC Berkeley（加州大学伯克利分校）； The Hong Kong Polytechnic University（香港理工大学）

专题命中知识编辑与模型理解：large language model(abstract)；language model(abstract)

Comments The code is available at https://github.com/bronyayang/Law_of_Vision_Representation_in_MLLMs

2510.03315 2025-10-07 cs.CL cs.AI cs.LG 67%

Decomposing Attention To Find Context-Sensitive Neurons

Alex Gibson

机构 * University of Cambridge（剑桥大学）

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

Comments 10 pages, 7 figures. Submitted to the Mechanistic Interpretability Workshop at NeurIPS 2025

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

2503.07399 2025-10-07 cs.CV 50%

Exploring Representation Invariance in Finetuning

Wenqiang Zu, Shenghao Xie, Hao Chen, Zhiqiang Chen, Liwen Hu, Yuanhao Xi, Yiming Liang, Junliang Ye, Bo Lei, Tiejun Huang, Guoqi Li, Lei Ma

机构 * Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）； Peking University（北京大学）； University of Chinese Academy of Sciences（中国科学院大学）； University of Göttingen（哥廷根大学）； Tsinghua University（清华大学）； Beijing Academy of Artificial Intelligence（北京人工智能研究院）

专题命中知识编辑与模型理解：foundation model(abstract)

1. 知识编辑与模型理解 27 篇

The Geometry of Truth: Layer-wise Semantic Dynamics for Hallucination Detection in Large Language Models

Mechanistic Interpretability of Socio-Political Frames in Language Models

From Word to World: Evaluate and Mitigate Culture Bias in LLMs via Word Association Test

Detecting LLM Hallucination Through Layer-wise Information Deficiency: Analysis of Ambiguous Prompts and Unanswerable Questions

time2time: Causal Intervention in Hidden States to Simulate Rare Events in Time Series Foundation Models

CBVLM: Training-free Explainable Concept-based Large Vision Language Models for Medical Image Classification

On the Effectiveness and Generalization of Race Representations for Debiasing High-Stakes Decisions

Visual Representations inside the Language Model

THEMIS: Unlocking Pretrained Knowledge with Foundation Model Embeddings for Anomaly Detection in Time Series

Towards Cross-modal Backward-compatible Representation Learning for Vision-Language Models

GRACE: Generative Representation Learning via Contrastive Policy Optimization

CoPA: Hierarchical Concept Prompting and Aggregating Network for Explainable Diagnosis

When Models Lie, We Learn: Multilingual Span-Level Hallucination Detection with PsiloQA

On the Role of Unobserved Sequences on Sample-based Uncertainty Quantification for LLMs

Does higher interpretability imply better utility? A Pairwise Analysis on Sparse Autoencoders

First Hallucination Tokens Are Different from Conditional Ones

Probing Geometry of Next Token Prediction Using Cumulant Expansion of the Softmax Entropy

Interpretable Neuropsychiatric Diagnosis via Concept-Guided Graph Neural Networks

Multimodal Arabic Captioning with Interpretable Visual Concept Integration

Graph Generation Powered with LLMs for Boosting Multivariate Time-Series Representation Learning

Rowen: Adaptive Retrieval-Augmented Generation for Hallucination Mitigation in LLMs

Revolutionizing Validation and Verification: Explainable Testing Methodologies for Intelligent Automotive Decision-Making Systems

BrainFLORA: Uncovering Brain Concept Representation via Multimodal Neural Embeddings

Law of Vision Representation in MLLMs

Decomposing Attention To Find Context-Sensitive Neurons

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

Exploring Representation Invariance in Finetuning