As large language models are increasingly trained and fine-tuned, practitioners need methods to identify which training data drive specific behaviors, particularly unintended ones. Training Data Attribution (TDA) methods address this by estimating datapoint influence. Existing approaches like influence functions are both computationally expensive and attribute based on single test examples, which can bias results toward syntactic rather than semantic similarity. To address these issues of scalability and influence to abstract behavior, we leverage interpretable structures within the model during the attribution. First, we introduce Concept Influence which attribute model behavior to semantic directions (such as linear probes or sparse autoencoder features) rather than individual test examples. Second, we show that simple probe-based attribution methods are first-order approximations of Concept Influence that achieve comparable performance while being over an order-of-magnitude faster. We empirically validate Concept Influence and approximations across emergent misalignment benchmarks and real post-training datasets, and demonstrate they achieve comparable performance to classical influence functions while being substantially more scalable. More broadly, we show that incorporating interpretable structure within traditional TDA pipelines can enable more scalable, explainable, and better control of model behavior through data.
A Monosemantic Attribution Framework for Stable Interpretability in Clinical Neuroscience Transformer-Based Language Models
Department of Computer Science and Technology(计算机科学与技术系);Cancer Research UK(癌症研究英国公司);Cambridge Institute(剑桥研究所);University of Cambridge(剑桥大学);United Kingdom(英国);DIMES(迪梅斯);University of Calabria(卡拉布里亚大学);Italy(意大利);Department of Computer Automatic and Management Engineering (DIAG)(计算机自动与管理工程系);Sapienza Università di Roma(罗马大学萨皮恩扎);Department of Psychiatry(精神病学系);School of Computing(计算学院);University of Kent(肯特大学);Department of Psychology(心理学系)
Disentangling Hallucinations: Orthogonal Semantic Projection for Robust Interpretability
U2IS, ENSTA, Institut Polytechnique de Paris(巴黎综合理工学院ENSTA学院U2IS实验室);ISIR, Université Sorbonne, Pierre et Marie Curie(索邦大学皮埃尔和玛丽·居里分校ISIR实验室);AMIAD, Pôle Recherche(AMIAD研究部)
I Can Find You in Seconds! Leveraging Large Language Models for Code Authorship Attribution
University of Central Florida(中佛罗里达大学);Nanyang Technological University(南洋理工大学);Technical University of Munich(慕尼黑工业大学);Propulsion and Space Research Center, Technology Innovation Institute(技术创新研究所推进与空间研究中心)
ProxySPEX: Inference-Efficient Interpretability via Sparse Feature Interactions in LLMs
Department of EECS UC Berkeley(电子工程与计算机科学系,伯克利大学);Department of Statistics UC Berkeley(统计学系,伯克利大学);Departments of Statistics and EECS UC Berkeley(统计学系和电子工程与计算机科学系,伯克利大学)
SafeSeek: Universal Attribution of Safety Circuits in Language Models
University of Science;United Arab Emirates University;Nanyang Technological University;Zayed University;Intelligent Science \& Technology Academy of CASIC
Department of Computer & Information Science & Engineering, University of Florida(佛罗里达大学计算机与信息科学与工程系);School of Computer Science, University of Oklahoma(俄克拉荷马大学计算机科学学院);Department of Electrical Engineering and Computer Science, United States Military Academy(美国军事学院电子工程与计算机科学系)
Backdoor Attribution: Elucidating and Controlling Backdoor in Language Models
University of Science and Technology of China(中国科学技术大学);Nanyang Technological University(南洋理工大学);United Arab Emirates University(阿拉伯联合酋长国大学);University of California San Diego(加州大学圣地亚哥分校);Abel AI;National University of Singapore(新加坡国立大学);Squirrel Ai Learning
The Metagame of Interpretability and Meta-Attributions
University of Warsaw(华沙大学);Centre for Credible AI, Warsaw University of Technology(可信AI中心,华沙技术大学);Bielefeld University(比勒菲尔德大学);LMU Munich, MCML(慕尼黑大学LMU,MCML)
Comprehensive and Reliable Feature Attribution for Diverse Modalities and Models via Frequency-Domain Insights
School of Computational Science, Wuhan University(武汉大学计算科学学院);Brain Research Center, Wuhan University(武汉大学脑科学研究中心);College of Information Science and Technology (School of Cyber Science and Technology), Shihezi University(石河子大学信息科学学院(网络安全科学与技术学院));Xinjiang Production and Construction Corps Key Laboratory of Computing Intelligence and Network Information Security Open Fund(新疆生产建设兵团计算智能与网络信息安全重点实验室开放基金)
Mechanistic Interpretability of LLM Jailbreaks via Internal Attribution Graphs
Department of Computer Science, University of South Dakota(南达科他大学计算机科学系);School of Information and Artificial Intelligence, Yangzhou University(扬州大学信息与人工智能学院)
Hybrid Attribution Priors for Explainable and Robust Model Training
School of Computer Science, Peking University(北京大学计算机科学学院);Key Lab of High Confidence Software Technologies, Peking University(北京大学高可信软件技术重点实验室);Kling Team(Kling团队);Department of Systems Engineering and Engineering Management, CUHK(CUHK系统工程与工程管理系)
A Unified Framework for Evaluating and Enhancing the Transparency of Explainable AI Methods via Perturbation-Gradient Consensus Attribution
Department of Computer Science, American International University-Bangladesh(美国国际大学-孟加拉国计算机科学系);Thomas Lord Department of Computer Science, Viterbi School of Engineering, University of Southern California(南加州大学维特比工程学院托马斯·洛德计算机科学系);Department of Computer Science and Engineering, Techno International New Town(Techno International New Town计算机科学与工程系)
FedXDS: Leveraging Model Attribution Methods to counteract Data Heterogeneity in Federated Learning
Fraunhofer Heinrich Hertz Institute(弗劳恩霍夫海因里希·赫兹研究所);Technical University of Berlin(柏林工业大学);Berlin Institute for the Foundations of Learning and Data (BIFOLD)(柏林学习与数据基础研究所)
Contrastive Explanations for Model Interpretability
Bar Ilan University(巴伊兰大学);Allen Institute for Artificial Intelligence(艾伦人工智能研究所);Paul G. Allen School of Computer Science and Engineering, University of Washington(华盛顿大学保罗·G·艾伦计算机科学与工程学院)
Unifying Perplexing Behaviors in Modified BP Attributions through Alignment Perspective
University of Science and Technology of China(中国科学技术大学);National Lab of Pattern Recognition, Institute of Automation, CAS(中国科学院自动化研究所模式识别国家重点实验室);University of Chinese Academy of Sciences(中国科学院大学)