A central goal of explainable AI is to express large language model (LLM) decision logic symbolically and ground it in internal mechanisms. Existing rule-extraction methods usually learn ungrounded symbolic surrogates, while mechanistic interpretability links behavior to neurons but often requires hand-crafted hypotheses and costly interventions. We introduce MechaRule, a pipeline that grounds rule extraction in LLM circuits by localizing sparse agonist activations whose ablation disrupts rule-related behavior. MechaRule rests on two findings. First, in a fixed baseline/flip regime, sparse agonist effects can exhibit overtopping: a few high-effect activations remain detectable within larger groups, dominate weaker ones, and flip many of the same examples. In such regimes, adaptive group testing with confidence-guided conservative pruning requires O(k log(N/k) + k) interventions over N candidates when k << N are agonists. Second, agonists are localized more reliably on data splits aligned with close-to-faithful rule behavior; spectral splits provide a rule-free fallback, whereas unfaithful splits degrade localization. Empirically, on arithmetic and jailbreaking, MechaRule recalls 97.0% of highest-effect agonists in matched brute-force validations at only 2.14% of exhaustive-ablation cost on average. Ablating the localized agonists eliminates 97.6--100.0% of eligible correct arithmetic answers and jailbreaks, and can correct arithmetic errors or induce jailbreaks by up to 72.8% and 32.5%.
Dissecting Role Cognition in Medical LLMs via Neuronal Ablation
School of Information, Renmin University of China(中国人民大学信息学院);Center of Machine Learning Research, Peking University(北京大学机器学习研究中心);School of Artificial Intelligence, Shanghai Jiaotong University(上海交通大学人工智能学院);Health Informatics Committee of the China Society for Scientific and Technical Information, Renmin University of China(中国科学技术信息学会健康信息技术委员会,中国人民大学);Institute for Advanced Algorithms Research (IAAR), Shanghai, China(先进算法研究所(IAAR),上海,中国);Memtensor Research Center, Shanghai, China(Memtensor研究中心,上海,中国)
SCAN: Sparse Circuit Anchor Interpretable Neuron for Lifelong Knowledge Editing
New Laboratory of Pattern Recognition (NLPR), State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS), Institute of Automation, Chinese Academy of Sciences(模式识别新实验室(NLPR)、多模态人工智能系统国家重点实验室(MAIS)、自动化研究所、中国科学院);Cuiying Honors College, Lanzhou University(cuiying荣誉学院、兰州大学);Research Institute of Intelligent Complex Systems, Fudan University(智能复杂系统研究院、复旦大学);Zhongguancun Academy(中关村学院)
Linguistic Neuron Overlap Patterns to Facilitate Cross-lingual Transfer on Low-resource Languages
School of Information Science and Technology, Beijing Foreign Studies University(信息科学与技术学院,北京外国语大学);Department of Informatics, King’s College London(信息学院,伦敦国王学院)
Under the Hood of Neural Networks: Characterizing Learned Representations by Functional Neuron Populations and Network Ablations
Institute of Technologies and Management of Digital Transformation(数字化转型技术与管理研究所);Institute of Information Management in Mechanical Engineering(机械工程信息管理研究所)
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
Operating Within the Operational Design Domain: Zero-Shot Perception with Vision-Language Models
Volkswagen Aktiengesellschaft(大众汽车股份有限公司);L3S Research Center(莱比锡大学汉诺威研究中心);Faculty of Information Technology(信息科技学院);MOIA GmbH(MOIA公司);Motor AI GmbH(Motor AI公司)