语言大模型 / LLM - arXivDaily 专题

2309.03433 2023-09-08 cs.CL 89%

Improving Open Information Extraction with Large Language Models: A Study on Demonstration Uncertainty

Chen Ling, Xujiang Zhao, Xuchao Zhang, Yanchi Liu, Wei Cheng, Haoyu Wang, Zhengzhang Chen, Takao Osaki, Katsushi Matsuda, Haifeng Chen, Liang Zhao

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

2308.04076 2023-08-09 cs.HC cs.CL 89%

DataTales: Investigating the use of Large Language Models for Authoring Data-Driven Articles

Nicole Sultanum, Arjun Srinivasan

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

Comments 4 pages, 3 figures

2307.01881 2023-07-06 cs.CR cs.CL 89%

ProPILE: Probing Privacy Leakage in Large Language Models

Siwon Kim, Sangdoo Yun, Hwaran Lee, Martin Gubri, Sungroh Yoon, Seong Joon Oh

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

2304.12918 2023-04-26 cs.LG 89%

N2G: A Scalable Approach for Quantifying Interpretable Neuron Representations in Large Language Models

Alex Foote, Neel Nanda, Esben Kran, Ionnis Konstas, Fazl Barez

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

Comments To be published at ICLR 2023 Workshop on Trustworthy and Reliable Large-Scale Machine Learning Models

2303.15125 2023-03-28 cs.HC cs.CL 89%

LMCanvas: Object-Oriented Interaction to Personalize Large Language Model-Powered Writing Environments

Tae Soo Kim, Arghya Sarkar, Yoonjoo Lee, Minsuk Chang, Juho Kim

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

Comments Accepted to CHI 2023 Workshop on Generative AI and HCI

2110.01691 2022-03-21 cs.HC cs.CL 89%

AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts

Tongshuang Wu, Michael Terry, Carrie J. Cai

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

2507.00979 2025-07-02 cs.AI cs.CL cs.LG 89%

Enhancing LLM Agent Safety via Causal Influence Prompting

Dongyoon Hahm, Woogyeol Jin, June Suk Choi, Sungsoo Ahn, Kimin Lee

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

Comments Accepted at ACL 2025 Findings, Source code: https://github.com/HahmDY/causal_influence_prompting.git

2406.01506 2025-02-19 cs.CL cs.AI cs.LG stat.ML 89%

The Geometry of Categorical and Hierarchical Concepts in Large Language Models

Kiho Park, Yo Joong Choe, Yibo Jiang, Victor Veitch

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；LLM(comments)

Comments Accepted for an oral presentation at ICLR 2025. Best Paper Award at the ICML 2024 Workshop on Mechanistic Interpretability. Code is available at https://github.com/KihoPark/LLM_Categorical_Hierarchical_Representations

2404.07362 2024-04-12 cs.HC 89%

"We Need Structured Output": Towards User-centered Constraints on Large Language Model Output

Michael Xieyang Liu, Frederick Liu, Alexander J. Fiannaca, Terry Koo, Lucas Dixon, Michael Terry, Carrie J. Cai

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

Journal ref "We Need Structured Output": Towards User-centered Constraints on LLM Output. In Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA '24), May 11-16, 2024, Honolulu, HI, USA

2312.03140 2023-12-07 cs.LG cs.AI cs.CL cs.DC 89%

FlexModel: A Framework for Interpretability of Distributed Large Language Models

Matthew Choi, Muhammad Adil Asif, John Willes, David Emerson

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

Comments 14 pages, 8 figures. To appear at the Socially Responsible Language Modelling Research (SoLaR) Workshop, 37th Conference on Neural Information Processing Systems (NeurIPS 2023)

2310.18679 2023-11-09 cs.CL cs.AI cs.LG 89%

N-Critics: Self-Refinement of Large Language Models with Ensemble of Critics

Sajad Mousavi, Ricardo Luna Gutiérrez, Desik Rengarajan, Vineet Gundecha, Ashwin Ramesh Babu, Avisek Naug, Antonio Guillen, Soumyendu Sarkar

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；foundation model(journal_ref)

Journal ref NeurIPS 2023 Workshop on Robustness of Few-shot and Zero-shot Learning in Foundation Models 2023(NeurIPS 2023)

2607.01457 2026-07-03 cs.CL cs.AI 新提交 89%

Grounded Optimization: A Layered Engineering Framework for Reducing LLM Hallucination in Automated Personal Document Rewriting

接地优化：一种减少自动个人文档重写中LLM幻觉的分层工程框架

Shashank Indukuri, Adarsh Agrawal

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

AI总结提出五层接地优化框架，通过时间验证、污染检测、结构不变性、提示接地和评估器，将简历重写中的幻觉率降至0.04-0.24。

Comments 13 pages, 1 figure. Equal contribution by both authors. Code and data: https://github.com/shashank-indukuri/grounded-optimization

2607.00415 2026-07-02 cs.CL cs.LG 新提交 89%

A Mechanistic View of Authority Hierarchy in LLM Sycophancy

LLM 谄媚中权威层级的机制性视角

Emil Joswin, Srujananjali Medicherla, Priyanka Mary Mammen

机构 * Independent Research（独立研究）

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

AI总结通过受控医疗QA实验，发现LLM按感知权威程度分级响应，机制是特定后期层中正确答案表征被权威信号主动擦除，且该擦除与权威水平成比例、抵抗均值向量干预、仅部分可逆。

2606.19353 2026-06-19 cs.CL cs.LG 新提交 89%

Quantifying Aleatoric Uncertainty of In-Context Learning for Robust Measure of LLM Prediction Confidence

量化上下文学习中的偶然不确定性以稳健衡量LLM预测置信度

Jinseok Chung, Minkyoung Song, Hyunji Jung, Namhoon Lee

机构 * POSTECH（浦项科技大学）

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

AI总结针对上下文学习（ICL）中预测对提示设计敏感的问题，提出基于贝叶斯观点和机制可解释性的自函数向量，直接估计偶然不确定性，并设计严格评估协议，在合成和真实数据集上验证了方法的可靠性及在幻觉检测等应用中的实用性。

Comments Accepted to ACL 2026

2606.10942 2026-06-10 cs.NI cs.AI cs.LG 新提交 89%

Generative Explainability for Next-Generation Networks: LLM-Augmented XAI with Mutual Feature Interactions

下一代网络的生成式可解释性：基于互特征交互的LLM增强XAI

Kiarash Rezaei, Omran Ayoub, Sebastian Troia, Francesco Lelli, Paolo Monti, Carlos Natalino

机构 * Swedish Innovation Agency（瑞典创新署）； Swiss Innovation Agency（瑞士创新署）

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

AI总结提出一种利用大语言模型和互特征交互数据生成自然语言解释的框架，在光传输质量估计用例中，相比基线方法，解释有用性和范围分别提升12.2%和6.2%，正确率达97.5%。

Comments 7 pages, with one page for appendix. Accepted for publication at the 2025 21th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob)

Journal ref Proc. WiMob, Marrakesh, Morocco, 2025

2606.04262 2026-06-04 cs.CL cs.AI 89%

Can I Take Another Dose? Evaluating LLM Decision-Making Under Temporal Uncertainty in OTC Dosing QA

我可以再服一剂吗？评估LLM在OTC剂量问答中时间不确定性下的决策能力

Maroof Kousar, Yibo Hu

机构 * Illinois Institute of Technology（伊利诺伊理工学院）

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

AI总结提出DOSEBENCH基准测试，评估大语言模型在非处方药剂量问答中处理时间推理、约束遵循和不确定性的能力。

Comments 16 pages, 7 figures

2603.21601 2026-03-24 cs.LG cs.AI 89%

Riemannian Geometry Speaks Louder Than Words: From Graph Foundation Model to Next-Generation Graph Intelligence

黎曼几何胜过言语：从图基础模型到下一代图智能

Philip S. Yu, Li Sun

机构 * University of Illinois Chicago（伊利诺伊大学香槟分校）； Beijing University of Posts and Telecommunications（北京邮电大学）

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

AI总结本文提出黎曼基础模型（RFM），通过内在几何捕捉复杂结构模式，推动图智能发展，实现从设计图模型到解决图结构应用的范式转变。

Comments 7 pages

2502.17516 2025-02-26 cs.LG cs.AI 89%

A Survey on Mechanistic Interpretability for Multi-Modal Foundation Models

Zihao Lin, Samyadeep Basu, Mohammad Beigi, Varun Manjunatha, Ryan A. Rossi, Zichao Wang, Yufan Zhou, Sriram Balasubramanian, Arman Zarei, Keivan Rezaei, Ying Shen, Barry Menglong Yao, Zhiyang Xu, Qin Liu, Yuxiang Zhang, Yan Sun, Shilong Liu, Li Shen, Hongxuan Li, Soheil Feizi, Lifu Huang

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

Comments 30 pages, 4 Figures, 10 Tables

2312.12141 2024-09-26 cs.CL cs.LG 89%

Neuron-Level Knowledge Attribution in Large Language Models

Zeping Yu, Sophia Ananiadou

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.LG

Comments Accepted by EMNLP 2024 main. This paper aims to identify the important neurons in large language models

2606.08444 2026-08-17 cs.SE 版本更新 89%

When LLMs Invent Rust Crates: An Empirical Study of Hallucination Patterns and Mitigation

当LLM发明Rust包：幻觉模式与缓解措施的实证研究

Jieming Zheng, Hao Guan, Yepang Liu

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

AI总结本研究首次大规模实证分析LLM生成Rust代码中的包幻觉问题，发现不同模型幻觉率惊人一致且对参数不敏感，并探索了提示工程缓解策略。

Comments The work has been accepted by the 17th International Conference on Internetware

2607.15626 2026-07-27 cs.HC 版本更新 89%

Understanding Fortunetelling with Large Language Models in China: User Practices, Perceptions, and Impacts on Beliefs and Decisions

理解中国大语言模型的算命现象：用户实践、认知及其对信念和决策的影响

Xueer Lin, Chenyu Li, Shuai Ma, Yuhan Lyu, Zhenhui Peng

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

AI总结研究中国用户用大语言模型算命的情况，通过分析社交媒体帖子和访谈用户，发现用户把它当情感支持工具，结果少改信念决策，但引发思维模式转变和小行为调整，探讨了从中获益的意义。

Comments Accepted at ICWSM 2027

2509.18127 2026-07-20 cs.LG cs.AI cs.CL 89%

Safe-SAIL: Towards a Fine-grained Safety Landscape of Large Language Models via Sparse Autoencoder Interpretation Framework

Safe-SAIL: 通过稀疏自编码解释框架构建大语言模型的细粒度安全景观

Jiaqi Weng, Han Zheng, Hanyu Zhang, Ej Zhou, Qinqin He, Jialing Tao, Hui Xue, Zhixuan Chu, Xiting Wang

机构 * Alibaba Group（阿里巴巴集团）； The State Key Laboratory of Blockchain and Data Security, Zhejiang University（浙江大学区块链与数据安全国家重点实验室）； Language Technology Lab, University of Cambridge（剑桥大学语言技术实验室）； Renmin University of China（中国人民大学）

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

AI总结本文提出Safe-SAIL框架，通过稀疏自编码解释方法高效识别安全领域特征，减少解释成本55%，并系统评估1758个安全相关特征，揭示风险特征识别和安全关键实体编码机制。

Journal ref Findings of the Association for Computational Linguistics: ACL 2026, pages 18916-18935 (2026)

2508.12620 2026-07-20 cs.SE cs.PL 版本更新 89%

Improving Code Understanding in Large Language Models through Concept-Aware Consistency Learning

通过概念感知一致性学习提高大语言模型中的代码理解能力

Xiaoning Ren, Qiang Hu, Wei Ma, Chongyang Liu, Yan Li, Yao Zhang, Lingxiao Jiang, Yongqiang Lyu, Yinxing Xue

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

AI总结研究针对大语言模型对基本编程概念理解浅的问题，引入结合概念感知调整的反事实代码增强框架，经多模型和基准综合评估，证明此方法能引导大语言模型增强概念理解，有效提升其在代码相关任务中的表现。

Comments To appear at IJCAI 2026

2606.30945 2026-07-01 cs.CY 新提交 89%

Free-form Association Tasks Reveal Stereotype Hallucination in Large Language Models

自由联想任务揭示大型语言模型中的刻板印象幻觉

Xinrui Chloe Zhao, Douglas Guilbeault, Amir Goldberg

专题命中知识编辑与模型理解：LLM(summary_cn,abstract_cn)；large language model(title)；language model(title)

AI总结通过抽象艺术和罗夏墨迹测试，比较人类与多模态LLM在自由联想中的刻板印象结构，发现LLM存在“刻板印象幻觉”，即生成不反映真实群体差异的强烈二阶刻板印象。

2606.07537 2026-06-11 cs.CL cs.AI cs.LG 交叉投稿 89%

From Architecture to Output: Structural Origins of Hallucination in Large Language Models and the Amplifying Role of Data

从架构到输出：大语言模型中幻觉的结构性起源及数据的放大作用

Md. Rejaul Korim Sadi, Toufiqur Rahman Tasin, Golam Mostofa Naeem

机构 * University of Science and Technology of China（中国科学技术大学）

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

AI总结本文分析大语言模型幻觉的结构性根源，指出自注意力、最大似然估计训练目标和自回归解码三个架构决策构成复合失效系统，并揭示数据病理如何放大这些脆弱性。

Comments 11 pages, 7 figures, 15 references

2404.14928 2026-06-01 cs.LG cs.AI cs.CL cs.SI 89%

Graph Machine Learning in the Era of Large Language Models (LLMs)

大语言模型时代的图机器学习

Shijie Wang, Jiani Huang, Zhikai Chen, Yu Song, Wenzhuo Tang, Haitao Mao, Wenqi Fan, Hui Liu, Xiaorui Liu, Dawei Yin, Qing Li

机构 * The Hong Kong Polytechnic University（香港理工大学）； Michigan State University（密歇根州立大学）； North Carolina State University（北卡罗来纳州立大学）； Baidu Inc（百度公司）

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

AI总结本文综述了大语言模型如何增强图机器学习的泛化、迁移和少样本学习能力，以及图如何提升大语言模型的推理和可解释性。

Comments Accepted by TIST

2605.19220 2026-05-20 cs.CL cs.AI cs.LG 89%

Position: Uncertainty Quantification in LLMs is Just Unsupervised Clustering

位置：在LLM中的不确定性量化仅仅是无监督聚类

Tiejin Chen, Longchao Da, Xiaoou Liu, Hua Wei

机构 * School of Computing（计算学院）； Augmented Intelligence, Arizona State University（智能增强与亚利桑那州立大学）

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

AI总结本文指出，当前LLM的不确定性量化方法本质上是无监督聚类算法，无法有效评估模型的外部正确性，导致无法检测出自信但错误的回答。文章提出了改进的不确定性量化方法，以确保模型的自信度能可靠地反映现实。

Comments Accepted by ICML 2026 Position Paper Track

2605.14749 2026-05-15 cs.CL cs.AI cs.LG 89%

Non-linear Interventions on Large Language Models

非线性干预大语言模型

Sangwoo Kim

机构 * Department of Linguistics, Seoul National University, Republic of Korea（韩国首尔国立大学语言系）

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

AI总结本文提出非线性干预方法，用于处理大语言模型中非线性特征，通过隐式特征干预提升拒绝绕过性能。

2605.12384 2026-05-13 cs.CL cs.AI cs.LG 89%

Scalable Token-Level Hallucination Detection in Large Language Models

可扩展的令牌级幻觉检测方法在大语言模型中

Rui Min, Tianyu Pang, Chao Du, Minhao Cheng, Yi R. Fung

机构 * Sea AI Lab（Sea AI实验室）； Hong Kong University of Science and Technology（香港科学与技术大学）； Pennsylvania State University（宾夕法尼亚州立大学）

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

AI总结本文提出TokenHD方法，通过可扩展的数据引擎和重要性加权策略，实现令牌级幻觉检测，实验显示小型模型在训练后性能优于大模型，且检测能力随模型规模扩大而提升。

2604.25591 2026-04-29 eess.AS cs.AI cs.CL cs.LG cs.SD 89%

Walking Through Uncertainty: An Empirical Study of Uncertainty Estimation for Audio-Aware Large Language Models

穿越不确定性：音频感知大语言模型不确定性估计的实证研究

Chun-Yi Kuan, Wei-Ping Huang, Hung-yi Lee

机构 * Graduate Institute of Communication Engineering, National Taiwan University, Taiwan（台湾大学通讯工程研究所）； Artificial Intelligence Center of Research Excellence (AI-CoRE), National Taiwan University, Taiwan（台湾大学人工智能卓越研究中心）

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

AI总结本文研究了音频感知大语言模型的不确定性估计，通过多种方法对比发现语义层面方法在通用音频推理中表现更优，且在可靠性导向任务中效果依赖模型和基准。

Comments Manuscript in progress

1. 知识编辑与模型理解 7505 篇

Improving Open Information Extraction with Large Language Models: A Study on Demonstration Uncertainty

DataTales: Investigating the use of Large Language Models for Authoring Data-Driven Articles

ProPILE: Probing Privacy Leakage in Large Language Models

N2G: A Scalable Approach for Quantifying Interpretable Neuron Representations in Large Language Models

LMCanvas: Object-Oriented Interaction to Personalize Large Language Model-Powered Writing Environments

AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts

Enhancing LLM Agent Safety via Causal Influence Prompting

The Geometry of Categorical and Hierarchical Concepts in Large Language Models

"We Need Structured Output": Towards User-centered Constraints on Large Language Model Output

FlexModel: A Framework for Interpretability of Distributed Large Language Models

N-Critics: Self-Refinement of Large Language Models with Ensemble of Critics

Grounded Optimization: A Layered Engineering Framework for Reducing LLM Hallucination in Automated Personal Document Rewriting

A Mechanistic View of Authority Hierarchy in LLM Sycophancy

Quantifying Aleatoric Uncertainty of In-Context Learning for Robust Measure of LLM Prediction Confidence

Generative Explainability for Next-Generation Networks: LLM-Augmented XAI with Mutual Feature Interactions

Can I Take Another Dose? Evaluating LLM Decision-Making Under Temporal Uncertainty in OTC Dosing QA

Riemannian Geometry Speaks Louder Than Words: From Graph Foundation Model to Next-Generation Graph Intelligence

A Survey on Mechanistic Interpretability for Multi-Modal Foundation Models

Neuron-Level Knowledge Attribution in Large Language Models

When LLMs Invent Rust Crates: An Empirical Study of Hallucination Patterns and Mitigation

Understanding Fortunetelling with Large Language Models in China: User Practices, Perceptions, and Impacts on Beliefs and Decisions

Safe-SAIL: Towards a Fine-grained Safety Landscape of Large Language Models via Sparse Autoencoder Interpretation Framework

Improving Code Understanding in Large Language Models through Concept-Aware Consistency Learning

Free-form Association Tasks Reveal Stereotype Hallucination in Large Language Models

From Architecture to Output: Structural Origins of Hallucination in Large Language Models and the Amplifying Role of Data

Graph Machine Learning in the Era of Large Language Models (LLMs)

Position: Uncertainty Quantification in LLMs is Just Unsupervised Clustering

Non-linear Interventions on Large Language Models

Scalable Token-Level Hallucination Detection in Large Language Models

Walking Through Uncertainty: An Empirical Study of Uncertainty Estimation for Audio-Aware Large Language Models