语言大模型 / LLM

2502.17598 2025-10-21 cs.LG cs.AI cs.CL 75%

Hallucination Detection in LLMs Using Spectral Features of Attention Maps

Jakub Binkowski, Denis Janiak, Albert Sawczyn, Bogdan Gabrys, Tomasz Kajdanowicz

机构 * Wroclaw University of Science and Technology（沃拉日-克拉夫大学科学与技术学院）； University of Technology Sydney（悉尼技术大学）

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

Comments Accepted to EMNLP 2025. Code available at https://github.com/graphml-lab-pwr/lapeigvals

2510.15188 2025-10-21 cs.CR cs.LG 70%

OCR-APT: Reconstructing APT Stories from Audit Logs using Subgraph Anomaly Detection and LLMs

Ahmed Aly, Essam Mansour, Amr Youssef

机构 * Concordia University（康科德大学）

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

Comments This is the authors' extended version of the paper accepted for publication at the ACM SIGSAC Conference on Computer and Communications Security (CCS 2025). The final published version is available at https://doi.org/10.1145/3719027.3765219

2505.13946 2025-10-21 cs.AI 70%

Visual Instruction Bottleneck Tuning

Changdae Oh, Jiatong Li, Shawn Im, Sharon Li

机构 * Department of Computer Sciences, University of Wisconsin–Madison（计算机科学系，威斯康星大学麦迪逊分校）

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

Comments NeurIPS 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

2505.14671 2025-10-21 cs.CV 50%

UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens

Ruichuan An, Sihan Yang, Renrui Zhang, Zijun Shen, Ming Lu, Gaole Dai, Hao Liang, Ziyu Guo, Shilin Yan, Yulin Luo, Bocheng Zou, Chaoqun Yang, Wentao Zhang

机构 * Peking University（北京大学）； Xi’an JiaoTong University（西安交通大学）； CUHK（香港中文大学）； Intel Labs, China（中国英特尔实验室）； Nanjing University（南京大学）； University of Wisconsin-Madison（威斯康星大学麦迪逊分校）； Tsinghua University（清华大学）

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

Journal ref NeurIPS 2025

2510.17748 2025-10-21 cs.DB 89%

This is Going to Sound Crazy, But What If We Used Large Language Models to Boost Automatic Database Tuning Algorithms By Leveraging Prior History? We Will Find Better Configurations More Quickly Than Retraining From Scratch!

William Zhang, Wan Shen Lim, Andrew Pavlo

专题命中其他LLM ：large language model(title,abstract)；language model(title,abstract)；LLM(abstract)

Comments Accepted to SIGMOD2026

2510.16502 2025-10-21 cs.SE 88%

On the Use of Large Language Models for Qualitative Synthesis

Sebastián Pizard, Ramiro Moreira, Federico Galiano, Ignacio Sastre, Lorena Etcheverry

专题命中其他LLM ：large language model(title,abstract)；language model(title,abstract)

2412.11704 2025-10-21 cs.CL cs.AI 86%

Adapting Chat Language Models Using Only Target Unlabeled Language Data

Atsuki Yamaguchi, Terufumi Morishita, Aline Villavicencio, Nikolaos Aletras

机构 * University of Sheffield（谢菲尔德大学）； Hitachi, Ltd.（日立株式会社）； University of Exeter（埃克塞特大学）

专题命中其他LLM ：language model(title,abstract)；LLM(abstract)；large language model(abstract)；分类 cs.CL、cs.AI

Comments Accepted to TMLR

2510.16794 2025-10-21 cs.CR cs.LG 85%

Black-box Optimization of LLM Outputs by Asking for Directions

Jie Zhang, Meng Ding, Yang Liu, Jue Hong, Florian Tramèr

机构 * ETH Zurich（苏黎世联邦理工学院）； University at Buffalo（布法罗大学）； Bytedance, Security Research（字节跳动安全研究）

专题命中其他LLM ：LLM(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.LG

2510.15914 2025-10-21 cs.AR cs.AI cs.PL 85%

VeriGRAG: Enhancing LLM-Based Verilog Code Generation with Structure-Aware Soft Prompts

Jiayu Zhao, Song Chen

专题命中其他LLM ：LLM(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.AI

Comments 9 pages, 5 figures

2510.17534 2025-10-21 cs.HC 85%

NieNie: Adaptive Rhythmic System for Stress Relief with LLM-Based Guidance

Yichen Yu, Qiaoran Wang

专题命中其他LLM ：LLM(title,abstract)；large language model(abstract)；language model(abstract)

2410.14835 2025-10-21 cs.PL 85%

Towards Automated Verification of LLM-Synthesized C Programs

Prasita Mukherjee, Benjamin Delaware

专题命中其他LLM ：LLM(title,abstract)；large language model(abstract)；language model(abstract)

2510.15976 2025-10-21 cs.CR cs.AI 84%

Learning to Watermark: A Selective Watermarking Framework for Large Language Models via Multi-Objective Optimization

Chenrui Wang, Junyi Shu, Billy Chiu, Yu Li, Saleh Alharbi, Min Zhang, Jing Li

机构 * Harbin Institute of Technology, Shenzhen, China（哈尔滨工业大学（深圳））； Lingnan University（岭大大学）； Zhejiang University（浙江大学）； Shaqra University, Saudi Arabia（沙迦大学）

专题命中其他LLM ：large language model(title)；language model(title)；分类 cs.AI

Comments 28 pages, 11 figures, NeurIPS 2025 Poster

2506.08164 2025-10-21 cs.LG 83%

BLUR: A Bi-Level Optimization Approach for LLM Unlearning

Hadi Reisizadeh, Jinghan Jia, Zhiqi Bu, Bhanukiran Vinzamuri, Anil Ramakrishna, Kai-Wei Chang, Volkan Cevher, Sijia Liu, Mingyi Hong

专题命中其他LLM ：LLM(title)；large language model(abstract)；language model(abstract)；分类 cs.LG

2510.17457 2025-10-21 cs.LG 79%

Deeper with Riemannian Geometry: Overcoming Oversmoothing and Oversquashing for Graph Foundation Models

Li Sun, Zhenhao Huang, Ming Zhang, Philip S. Yu

机构 * North China Electric Power University（华北电力大学）； University of Illinois Chicago（伊利诺伊大学芝加哥分校）

专题命中其他LLM ：foundation model(title,abstract)；分类 cs.LG

Comments Accept by NeurIPS 25

2510.17000 2025-10-21 cs.CR cs.CL cs.LG 79%

Bits Leaked per Query: Information-Theoretic Bounds on Adversarial Attacks against LLMs

Masahiro Kaneko, Timothy Baldwin

机构 * MBZUAI Abu Dhabi, UAE（阿布扎赫德MBZUAI）

专题命中其他LLM ：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.LG

Comments NeurIPS 2025 (spotlight)

2510.16870 2025-10-21 cs.CV 78%

Uncovering Brain-Like Hierarchical Patterns in Vision-Language Models through fMRI-Based Neural Encoding

Yudan Ren, Xinlong Wang, Kexin Wang, Tian Xia, Zihan Ma, Zhaowei Li, Xiangrong Bi, Xiao Li, Xiaowei He

专题命中其他LLM ：language model(title,abstract)

Comments 14 pages, 7 figures

2510.17006 2025-10-21 cs.CL 70%

Online Learning Defense against Iterative Jailbreak Attacks via Prompt Optimization

Masahiro Kaneko, Zeerak Talat, Timothy Baldwin

机构 * MBZUAI（马克斯·普朗克人工智能研究所）； University of Edinburgh（爱丁堡大学）

专题命中其他LLM ：large language model(abstract)；language model(abstract)；分类 cs.CL

2509.13772 2025-10-21 cs.CR cs.IR cs.LG 70%

Who Taught the Lie? Responsibility Attribution for Poisoned Knowledge in Retrieval-Augmented Generation

Baolei Zhang, Haoran Xin, Yuxi Chen, Zhuqing Liu, Biao Yi, Tong Li, Lihai Nie, Zheli Liu, Minghong Fang

机构 * Nankai University（南开大学）； University of North Texas（北德克萨斯大学）； University of Louisville（路易斯维尔大学）

专题命中其他LLM ：large language model(abstract)；language model(abstract)；分类 cs.LG

Comments To appear in the IEEE Symposium on Security and Privacy, 2026

2505.18752 2025-10-21 cs.CL 70%

Unifying Attention Heads and Task Vectors via Hidden State Geometry in In-Context Learning

Haolin Yang, Hakaze Cho, Yiqiao Zhong, Naoya Inoue

机构 * University of Chicago（芝加哥大学）； JAIST ； University of Wisconsin - Madison（威斯康星大学麦迪逊分校）； RIKEN（日本研究机构）

专题命中其他LLM ：large language model(abstract)；language model(abstract)；分类 cs.CL

Comments 52 pages, 70 figures, 24 tables, NeurIPS 2025

2510.16989 2025-10-21 cs.CV 67%

Training-free Online Video Step Grounding

Luca Zanella, Massimiliano Mancini, Yiming Wang, Alessio Tonioni, Elisa Ricci

机构 * University of Trento（特伦托大学）； Fondazione Bruno Kessler（布鲁诺·凯斯勒基金会）； Google（谷歌）

专题命中其他LLM ：large language model(abstract)；language model(abstract)

Comments NeurIPS 2025. Project website at https://lucazanella.github.io/baglm/

2510.16990 2025-10-21 cs.LG 57%

Graph4MM: Weaving Multimodal Learning with Structural Information

Xuying Ning, Dongqi Fu, Tianxin Wei, Wujiang Xu, Jingrui He

机构 * University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）； Meta AI ； Rutgers University（罗格斯大学）

专题命中其他LLM ：foundation model(abstract)；分类 cs.LG

Comments ICML 2025

2510.04774 2025-10-21 cs.RO cs.AI cs.MA 57%

Online automatic code generation for robot swarms: LLMs and self-organizing hierarchy

Weixu Zhu, Marco Dorigo, Mary Katherine Heinrich

专题命中其他LLM ：LLM(abstract)；分类 cs.AI

Comments This abstract was accepted to and presented at the "Multi-Agent Cooperative Systems and Swarm Robotics in the Era of Generative AI" (MACRAI) workshop at the 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025)

2510.16257 2025-10-21 cs.CL 57%

Towards Low-Resource Alignment to Diverse Perspectives with Sparse Feedback

Chu Fei Luo, Samuel Dahan, Xiaodan Zhu

机构 * Department of Electrical and Computer Engineering & Ingenuity Labs Research Institute（电气与计算机工程系及创新实验室研究机构）； Conflict Analytics Lab, Queen’s University（冲突分析实验室，女王大学）； Cornell Law School（康奈尔法学院）

专题命中其他LLM ：language model(abstract)；分类 cs.CL

Comments Findings of EMNLP 2025, 5 pages

2510.17686 2025-10-21 cs.CV 50%

Towards 3D Objectness Learning in an Open World

Taichi Liu, Zhenyu Wang, Ruofeng Liu, Guang Wang, Desheng Zhang

机构 * Rutgers University（罗格斯大学）； Tsinghua University（清华大学）； Michigan State University（密歇根州立大学）； Florida State University（佛罗里达州立大学）

专题命中其他LLM ：foundation model(abstract)

Comments Accepted by NeurIPS 2025

2510.17323 2025-10-21 astro-ph.HE 50%

A Common Synchrotron Origin for Prompt Gamma-Ray and Soft X-Ray Emission in GRBs: Evidence from Joint Spectral Analysis

Ziming Wang, Chenyu Wang, He Gao, Hua Feng, An Li, Lin Lin, Songyu Shen

专题命中其他LLM ：prompting(abstract)

Comments 25 pages, 24 figures

1. 知识编辑与模型理解 13 篇

Hallucination Detection in LLMs Using Spectral Features of Attention Maps

OCR-APT: Reconstructing APT Stories from Audit Logs using Subgraph Anomaly Detection and LLMs

Visual Instruction Bottleneck Tuning

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

UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens

2. 其他LLM 21 篇

This is Going to Sound Crazy, But What If We Used Large Language Models to Boost Automatic Database Tuning Algorithms By Leveraging Prior History? We Will Find Better Configurations More Quickly Than Retraining From Scratch!

On the Use of Large Language Models for Qualitative Synthesis

Adapting Chat Language Models Using Only Target Unlabeled Language Data

Black-box Optimization of LLM Outputs by Asking for Directions

VeriGRAG: Enhancing LLM-Based Verilog Code Generation with Structure-Aware Soft Prompts

NieNie: Adaptive Rhythmic System for Stress Relief with LLM-Based Guidance

Towards Automated Verification of LLM-Synthesized C Programs

Learning to Watermark: A Selective Watermarking Framework for Large Language Models via Multi-Objective Optimization

BLUR: A Bi-Level Optimization Approach for LLM Unlearning

Deeper with Riemannian Geometry: Overcoming Oversmoothing and Oversquashing for Graph Foundation Models

Bits Leaked per Query: Information-Theoretic Bounds on Adversarial Attacks against LLMs

Uncovering Brain-Like Hierarchical Patterns in Vision-Language Models through fMRI-Based Neural Encoding

Online Learning Defense against Iterative Jailbreak Attacks via Prompt Optimization

Who Taught the Lie? Responsibility Attribution for Poisoned Knowledge in Retrieval-Augmented Generation

Unifying Attention Heads and Task Vectors via Hidden State Geometry in In-Context Learning

Training-free Online Video Step Grounding

Graph4MM: Weaving Multimodal Learning with Structural Information

Online automatic code generation for robot swarms: LLMs and self-organizing hierarchy

Towards Low-Resource Alignment to Diverse Perspectives with Sparse Feedback

Towards 3D Objectness Learning in an Open World

A Common Synchrotron Origin for Prompt Gamma-Ray and Soft X-Ray Emission in GRBs: Evidence from Joint Spectral Analysis