语言大模型 / LLM - arXivDaily 专题

2506.02964 2025-10-27 cs.CV cs.LG 57%

FORLA: Federated Object-centric Representation Learning with Slot Attention

Guiqiu Liao, Matjaz Jogan, Eric Eaton, Daniel A. Hashimoto

机构 * PCASO Laboratory, Dept. of Surgery, University of Pennsylvania（宾夕法尼亚大学外科部PCASO实验室）； Dept. of Computer and Information Science, University of Pennsylvania（宾夕法尼亚大学计算机与信息科学系）

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

Comments Accepted by Neurips2025

2510.19861 2025-10-24 cs.LG 57%

Some Attention is All You Need for Retrieval

Felix Michalak, Steven Abreu

机构 * University of Groningen（格罗宁根大学）

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

Comments 16 pages, 10 figures

2510.18037 2025-10-23 cs.LG q-bio.NC stat.ML 57%

Benchmarking Probabilistic Time Series Forecasting Models on Neural Activity

Ziyu Lu, Anna J. Li, Alexander E. Ladd, Pascha Matveev, Aditya Deole, Eric Shea-Brown, J. Nathan Kutz, Nicholas A. Steinmetz

机构 * Department of Applied Mathematics, University of Washington（应用数学系，华盛顿大学）； Department of Neurobiology and Biophysics, University of Washington（神经生物学与生物物理学系，华盛顿大学）； Allen Institute for Brain Science（脑科学研究所）； Department of Electrical and Computer Engineering, University of Washington（电气与计算机工程系，华盛顿大学）

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

Comments Accepted at the 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Data on the Brain & Mind

2505.11741 2025-10-22 cs.AI cs.CR 57%

MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs

Geigh Zollicoffer, Minh Vu, Manish Bhattarai

机构 * Los Alamos National Laboratory（洛斯阿拉莫斯国家实验室）

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

2501.04641 2025-10-22 cs.LG math.ST stat.ML stat.TH 57%

A Statistical Theory of Contrastive Pre-training and Multimodal Generative AI

Kazusato Oko, Licong Lin, Yuhang Cai, Song Mei

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

2410.12539 2025-10-22 cs.AI cs.MA 57%

Counterfactual Effect Decomposition in Multi-Agent Sequential Decision Making

Stelios Triantafyllou, Aleksa Sukovic, Yasaman Zolfimoselo, Goran Radanovic

机构 * Max Planck Institute for Software Systems（马克斯·普朗克软件系统研究所）

专题命中知识编辑与模型理解：LLM(abstract)；分类 cs.AI

Comments ICML 2025

2510.17771 2025-10-21 cs.AI cs.CV 57%

Seeing but Not Believing: Probing the Disconnect Between Visual Attention and Answer Correctness in VLMs

Zhining Liu, Ziyi Chen, Hui Liu, Chen Luo, Xianfeng Tang, Suhang Wang, Joy Zeng, Zhenwei Dai, Zhan Shi, Tianxin Wei, Benoit Dumoulin, Hanghang Tong

机构 * University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）； Amazon（亚马逊）； Penn State University（宾夕法尼亚州立大学）

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

Comments 21 pages, 10 figures, 6 tables

2510.15115 2025-10-20 cs.CL 57%

Measuring the Effect of Disfluency in Multilingual Knowledge Probing Benchmarks

Kirill Semenov, Rico Sennrich

机构 * University of Zurich（苏黎世大学）

专题命中知识编辑与模型理解：LLM(abstract)；分类 cs.CL

2510.13916 2025-10-20 cs.CL 57%

Element2Vec: Build Chemical Element Representation from Text for Property Prediction

Yuanhao Li, Keyuan Lai, Tianqi Wang, Qihao Liu, Jiawei Ma, Yuan-Chao Hu

机构 * Songshan Lake Materials Laboratory（松山湖材料实验室）； International Academic Center of Complex Systems, Beijing Normal University（复杂系统国际学术中心，北京师范大学）； Department of Applied Physics, The Hong Kong Polytechnic University（应用物理系，香港理工大学）； Department of Computer Science, City University of Hong Kong (Dongguan)（计算机科学系，香港城市大学（东莞））； Department of Computer Science, City University of Hong Kong（计算机科学系，香港城市大学）

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

2510.14800 2025-10-17 cs.CV cs.AI 57%

Morphology-Aware Prognostic model for Five-Year Survival Prediction in Colorectal Cancer from H&E Whole Slide Images

Usama Sajjad, Abdul Rehman Akbar, Ziyu Su, Deborah Knight, Wendy L. Frankel, Metin N. Gurcan, Wei Chen, Muhammad Khalid Khan Niazi

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

2207.13929 2025-10-15 cs.CL 57%

MLRIP: Pre-training a military language representation model with informative factual knowledge and professional knowledge base

Hui Li, Xuekang Yang

机构 * School of Automation, Nanjing University of Science and Technology（自动化学院，南京理工大学）

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

Comments 12 pages, 6 figures

2510.10409 2025-10-14 cs.AI 57%

Trace Length is a Simple Uncertainty Signal in Reasoning Models

Siddartha Devic, Charlotte Peale, Arwen Bradley, Sinead Williamson, Preetum Nakkiran, Aravind Gollakota

机构 * University of Southern California（南加州大学）； Stanford University（斯坦福大学）； Apple（苹果公司）

专题命中知识编辑与模型理解：post-training(abstract)；分类 cs.AI

2510.10224 2025-10-14 cs.CL cs.IR 57%

Text2Token: Unsupervised Text Representation Learning with Token Target Prediction

Ruize An, Richong Zhang, Zhijie Nie, Zhanyu Wu, Yanzhao Zhang, Dingkun Long

机构 * CCSE, School of Computer Science and Engineering, Beihang University, Beijing, China（计算机科学与工程学院，北京航空航天大学，北京，中国）； Zhongguancun Laboratory, Beijing, China（中关村实验室，北京，中国）； Shen Yuan Honors College, Beihang University, Beijing, China（神元荣誉学院，北京航空航天大学，北京，中国）

专题命中知识编辑与模型理解：LLM(abstract)；分类 cs.CL

2510.10100 2025-10-14 cs.CV cs.LG 57%

Cooperative Pseudo Labeling for Unsupervised Federated Classification

Kuangpu Guo, Lijun Sheng, Yongcan Yu, Jian Liang, Zilei Wang, Ran He

机构 * University of Science and Technology of China（中国科学技术大学）； NLPR & MAIS, Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）； School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院）

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

Comments Accepted by ICCV 2025

2410.23684 2025-10-13 cs.CL 57%

Improbable Bigrams Expose Vulnerabilities of Incomplete Tokens in Byte-Level Tokenizers

Eugene Jang, Kimin Lee, Jin-Woo Chung, Keuntae Park, Seungwon Shin

机构 * Northeastern University（东北大学）； KAIST（韩国科学技术院）； S2W Inc.（S2W公司）

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

Comments EMNLP 2025 Main

2510.09135 2025-10-13 cs.CV cs.LG 57%

Training Feature Attribution for Vision Models

Aziz Bacha, Thomas George

机构 * Orange Research Châtillon, France（法国夏特隆橙色研究机构）； École polytechnique, Palaiseau, France（法国帕莱索高等理工学院）

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

2505.08348 2025-10-09 cs.CL 57%

Geometry of Semantics in Next-Token Prediction: How Optimization Implicitly Organizes Linguistic Representations

Yize Zhao, Christos Thrampoulidis

机构 * Department of Electrical and Computer Engineering（电气与计算机工程系）； University of British Columbia（不列颠哥伦比亚大学）

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

Comments Revised manuscript for improved clarity and readability

2509.19569 2025-10-06 cs.CL 57%

ExPe: Exact Positional Encodings for Generative Transformer Models with Extrapolating Capabilities

Aleksis Datseris, Sylvia Vassileva, Ivan Koychev, Svetla Boytcheva

机构 * Faculty of Mathematics and Informatics, Sofia University St. Kliment Ohridski（数学与信息学系，索菲亚大学）； Graphwise

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

2509.18997 2025-10-03 cs.LG 57%

Theoretical Foundations of Representation Learning using Unlabeled Data: Statistics and Optimization

Pascal Esser, Maximilian Fleissner, Debarghya Ghoshdastidar

机构 * Ludwig-Maximilians-Universität München（慕尼黑路德维希-马克西米利安大学）； Technical University of Munich（慕尼黑技术大学）； TUM School of Computation, Information and Technology（慕尼黑技术大学计算、信息与技术学院）； Munich Data Science Institute (MDSI)（慕尼黑数据科学研究所）； Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）

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

2505.11627 2025-10-03 cs.LG 57%

Enhancing Electricity-System Resilience with Adaptive Robust Optimization and Conformal Uncertainty Characterization

Shuyi Chen, Shixiang Zhu, Ramteen Sioshansi

机构 * Heinz College of Information Systems and Public Policy, Carnegie Mellon University（信息系统与公共政策学院，卡内基梅隆大学）； Carnegie Mellon Electricity Industry Center（卡内基梅隆电力产业中心）； Wilton E. Scott Institute for Energy Innovation（威利特·E·斯科特能源创新研究所）； Department of Engineering and Public Policy, Carnegie Mellon Electricity Industry Center（工程与公共政策系，卡内基梅隆电力产业中心）； Department of Electrical and Computer Engineering, Heinz College of Information Systems and Public Policy（电气与计算机工程系，信息系统与公共政策学院）； Department of Integrated Systems Engineering, The Ohio State University（整合系统工程系，俄亥俄州立大学）

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

2509.24431 2025-09-30 cs.LG 57%

Semantic Compression via Multimodal Representation Learning

Eleonora Grassucci, Giordano Cicchetti, Aurelio Uncini, Danilo Comminiello

机构 * Dept. of Information Engineering, Electronics, and Telecomm.（信息工程、电子与电信系）

专题命中知识编辑与模型理解：post-training(abstract)；分类 cs.LG

2509.00310 2025-09-30 cs.RO cs.AI 57%

TReF-6: Inferring Task-Relevant Frames from a Single Demonstration for One-Shot Skill Generalization

Yuxuan Ding, Shuangge Wang, Tesca Fitzgerald

机构 * Yale University（耶鲁大学）

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

2509.23717 2025-09-30 cs.AI 57%

Measuring Sparse Autoencoder Feature Sensitivity

Claire Tian, Katherine Tian, Nathan Hu

机构 * The Harker School（哈克尔学校）； 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.AI

Comments NeurIPS 2025 Workshop on Mechanistic Interpretability Camera Ready

2508.12341 2025-09-30 cs.CV cs.AI 57%

Semantic Discrepancy-aware Detector for Image Forgery Identification

Ziye Wang, Minghang Yu, Chunyan Xu, Zhen Cui

机构 * Nanjing University of Science and Technology（南京理工大学）； Beijing Normal University（北京师范大学）

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

Comments 10 pages, 5 figures

2505.15963 2025-09-30 cs.CV cs.CL 57%

OViP: Online Vision-Language Preference Learning for VLM Hallucination

Shujun Liu, Siyuan Wang, Zejun Li, Jianxiang Wang, Cheng Zeng, Zhongyu Wei

机构 * Fudan University（复旦大学）； University of Southern California（南加州大学）； ByteDance（字节跳动）

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

2509.21695 2025-09-29 cs.LG 57%

Wav2Arrest 2.0: Long-Horizon Cardiac Arrest Prediction with Time-to-Event Modeling, Identity-Invariance, and Pseudo-Lab Alignment

Saurabh Kataria, Davood Fattahi, Minxiao Wang, Ran Xiao, Matthew Clark, Timothy Ruchti, Mark Mai, Xiao Hu

机构 * Nell Hodgson Woodruff School of Nursing, Emory University（埃默里大学护理学院）； Department of Pediatrics, Emory School of Medicine（埃默里医学院儿科部）

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

Comments Submitted to BPSC

2506.02147 2025-09-26 cs.CL 57%

BabyLM's First Constructions: Causal probing provides a signal of learning

Joshua Rozner, Leonie Weissweiler, Cory Shain

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

Comments Accepted to EMNLP 2025

2509.20065 2025-09-25 cs.CL 57%

From Input Perception to Predictive Insight: Modeling Model Blind Spots Before They Become Errors

Maggie Mi, Aline Villavicencio, Nafise Sadat Moosavi

机构 * University of Sheffield（谢菲尔德大学）； University of Exeter（埃克塞特大学）； The Alan Turing Institute（艾伦·图灵研究所）； UFRN, Brazil（巴西UFRN）

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

Comments EMNLP 2025

2503.13390 2025-09-25 cs.CL 57%

Aligned Probing: Relating Toxic Behavior and Model Internals

Andreas Waldis, Vagrant Gautam, Anne Lauscher, Dietrich Klakow, Iryna Gurevych

机构 * Ubiquitous Knowledge Processing Lab (UKP Lab)（通用知识处理实验室）； Technical University of Darmstadt（德累斯顿技术大学）； Information Systems Research Lab（信息系统研究实验室）； Lucerne University of Applied Sciences and Arts（卢塞恩应用科学与艺术大学）； Spoken Language Systems（语音语言系统）； Saarland University（萨尔兰大学）； Data Science Group（数据科学组）； University of Hamburg（汉堡大学）

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

2509.18691 2025-09-24 cs.SD cs.AI eess.AS 57%

An overview of neural architectures for self-supervised audio representation learning from masked spectrograms

Sarthak Yadav, Sergios Theodoridis, Zheng-Hua Tan

机构 * Department of Electronic Systems, Aalborg University（电子系统系，奥尔堡大学）； Pioneer Centre for Artificial Intelligence（人工智能先锋中心）； National and Kapodistrian University of Athens（雅典国家与卡波蒂斯坦大学）

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

1. 知识编辑与模型理解 7583 篇

FORLA: Federated Object-centric Representation Learning with Slot Attention

Some Attention is All You Need for Retrieval

Benchmarking Probabilistic Time Series Forecasting Models on Neural Activity

MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs

A Statistical Theory of Contrastive Pre-training and Multimodal Generative AI

Counterfactual Effect Decomposition in Multi-Agent Sequential Decision Making

Seeing but Not Believing: Probing the Disconnect Between Visual Attention and Answer Correctness in VLMs

Measuring the Effect of Disfluency in Multilingual Knowledge Probing Benchmarks

Element2Vec: Build Chemical Element Representation from Text for Property Prediction

Morphology-Aware Prognostic model for Five-Year Survival Prediction in Colorectal Cancer from H&E Whole Slide Images

MLRIP: Pre-training a military language representation model with informative factual knowledge and professional knowledge base

Trace Length is a Simple Uncertainty Signal in Reasoning Models

Text2Token: Unsupervised Text Representation Learning with Token Target Prediction

Cooperative Pseudo Labeling for Unsupervised Federated Classification

Improbable Bigrams Expose Vulnerabilities of Incomplete Tokens in Byte-Level Tokenizers

Training Feature Attribution for Vision Models

Geometry of Semantics in Next-Token Prediction: How Optimization Implicitly Organizes Linguistic Representations

ExPe: Exact Positional Encodings for Generative Transformer Models with Extrapolating Capabilities

Theoretical Foundations of Representation Learning using Unlabeled Data: Statistics and Optimization

Enhancing Electricity-System Resilience with Adaptive Robust Optimization and Conformal Uncertainty Characterization

Semantic Compression via Multimodal Representation Learning

TReF-6: Inferring Task-Relevant Frames from a Single Demonstration for One-Shot Skill Generalization

Measuring Sparse Autoencoder Feature Sensitivity

Semantic Discrepancy-aware Detector for Image Forgery Identification

OViP: Online Vision-Language Preference Learning for VLM Hallucination

Wav2Arrest 2.0: Long-Horizon Cardiac Arrest Prediction with Time-to-Event Modeling, Identity-Invariance, and Pseudo-Lab Alignment

BabyLM's First Constructions: Causal probing provides a signal of learning

From Input Perception to Predictive Insight: Modeling Model Blind Spots Before They Become Errors

Aligned Probing: Relating Toxic Behavior and Model Internals

An overview of neural architectures for self-supervised audio representation learning from masked spectrograms