Large language model (LLM) hallucinations, meaning fluent but factually incorrect generations, fall into two types: faithfulness violations, where the model misuses provided context, and factuality violations, where answers reflect errors in internal knowledge. Proper mitigation depends on knowing which source drives each answer. We study contributive attribution, i.e. the classification of the dominant knowledge source behind each output, and show that a simple linear probe trained on hidden representations can reliably identify it. We introduce AttriWiki, a self-supervised pipeline that automatically generates labelled training data by prompting models to recall withheld entities from memory or read them from context without relying on knowledge conflicts. Probes trained on AttriWiki achieve up to 0.96 Macro-$F_1$ on Llama-3.1-8B, Mistral-7B, and Qwen-7B, transfer to SQuAD and WebQuestions with 0.94-0.99 Macro-$F_1$, and generalise zero-shot to Tighidet et al. (2024)'s benchmark, outperforming their probe on conflicting settings without retraining. Furthermore, attribution mismatches raise error rates by up to 70%, though correct attribution does not guarantee correct answers, pointing to the need for broader detection frameworks.
Source-Aware Training Enables Knowledge Attribution in Language Models
University of Michigan（密歇根大学）；Allen Institute for AI（艾伦人工智能研究所）；Carnegie Mellon University（卡内基梅隆大学）；University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Can LLMs Evaluate Complex Attribution in QA? Automatic Benchmarking using Knowledge Graphs
School of Computer Science and Engineering, Southeast University（东南大学计算机科学与工程学院）；Key Laboratory of New Generation Artificial Intelligence Technology and its Interdisciplinary Applications (Southeast University), Ministry of Education（教育部新一代人工智能技术及其交叉应用关键实验室（东南大学））；The University of Manchester（曼彻斯特大学）；The University of Edinburgh（爱丁堡大学）；Law and Innovation Lab, School of Law, Southeast University（东南大学法学院创新与法律实验室）；State Key Lab. for Novel Software Technology, Nanjing University（南京大学新型软件技术国家重点实验室）；The Chinese University of Hong Kong（香港中文大学）；Monash University（莫纳什大学）
KG-TRACES: Enhancing Large Language Models with Knowledge Graph-constrained Trajectory Reasoning and Attribution Supervision
Zhejiang University（浙江大学）；Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；University of Science and Technology of China（中国科学技术大学）；Fudan University（复旦大学）
ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall
HKUST(GZ) Deep Interdisciplinary Intelligence Lab（香港科技大学(广州)深 interdisciplinary 智能实验室）；BUAA（北京航空航天大学）；HKUST(GZ) Institute of Deep Perception Technology, JITRI Deep Interdisciplinary Intelligence Lab（香港科技大学(广州)深度感知技术研究所，JITRI 深 interdisciplinary 智能实验室）
Birds have four legs?! NumerSense: Probing Numerical Commonsense Knowledge of Pre-trained Language Models
University of Southern California（南加州大学）
最近的工作表明，预训练语言模型（PTLM），如BERT，拥有一定的常识和事实知识。他们认为，通过预测屏蔽词，将PTLM用作“神经知识库”是有希望的。令人惊讶的是，我们发现这可能不适用于数字常识知识（例如，鸟通常有两条腿。在本文中，我们调查是否以及在何种程度上，我们可以从PTLM以及这个过程的鲁棒性的数值常识知识。为了研究这一点，我们引入了一个新的探测任务，诊断数据集NumerSense包含13.6k个屏蔽词预测探测器（10.5k用于微调，3.1k用于测试）。我们的分析表明：（1）在任何微调之前，BERT及其更强的变体RoBERTa在诊断数据集上表现不佳;（2）远程监督的微调带来了一些改进;（3）与人类表现相比，最好的监督模型仍然表现不佳（准确率为54.06% vs 96.3%）。
Context Variance Evaluation of Pretrained Language Models for Prompt-based Biomedical Knowledge Probing
College of Information and Computer Science, University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校信息与计算机科学学院）；Department of Computer Science, University of Massachusetts Lowell（马萨诸塞大学洛厄尔分校计算机科学系）；Department of Medicine, University of Massachusetts Medical School（马萨诸塞大学医学院医学系）；Center for Healthcare Organization and Implementation Research, Bedford Veterans Affairs Medical Center（贝德福德退伍军人事务医疗中心医疗保健组织与实施研究中心）
Give Me the Facts! A Survey on Factual Knowledge Probing in Pre-trained Language Models
University of Mannheim（曼海姆大学）；University of Marburg（马尔堡大学）；Institute for AI in Medicine (IKIM)（医学人工智能研究所（IKIM））；University Hospital Essen（埃森大学医院）；University of Duisburg-Essen（杜伊斯堡-埃森大学）
How Large Language Models Encode Context Knowledge? A Layer-Wise Probing Study
School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University（上海交通大学电子信息与电气工程学院）；Shandong University（山东大学）；Southeast University（东南大学）；Leiden University（莱顿大学）
本文利用ChatGPT构建探测数据集并以 $\mathcal V$-usable information 为指标，揭示LLM在不同层编码上下文知识的位置、扩散与遗忘规律。
Beyond Prompt or Skill? Attribution-Guided Optimization of Modular LLM Programs
The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））；Shenzhen Future Network of Intelligence Institute (FNii-Shenzhen)（深圳未来智能网络研究院（深圳FNii））；Sun Yat-sen University（中山大学）
Explaining Black-box Language Models with Knowledge Probing Systems: A Post-hoc Explanation Perspective
School of Computer and Information Technology, Shanxi University, China（山西大学计算机与信息学院）；Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education, Shanxi University, China（教育部计算智能与中文信息处理重点实验室）；Institute for Infocomm Research, A*Star, Singapore（A*Star信息与通信研究所）
Demystifying Scientific Problem-Solving in LLMs by Probing Knowledge and Reasoning
Department of Computer Science, Yale University；Department of Molecular \& Cellular Biology, Harvard University；Allen Institute for AI；Department of Computer Science, Northwestern University
Decomposed Prompting: Probing Multilingual Linguistic Structure Knowledge in Large Language Models
Center for Information and Language Processing (CIS)（信息与语言处理中心）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；ScaDS.AI and TU Dresden（ScaDS.AI 和 梅克伦堡-前波美拉尼亚技术大学）；Department of Statistics, LMU Munich（统计学系，慕尼黑大学）；University of Maryland, College Park（马里兰大学 College Park 分校）
