Causal abstraction provides a theoretical foundation for mechanistic interpretability, the field concerned with providing intelligible algorithms that are faithful simplifications of the known, but opaque low-level details of black box AI models. Our contributions are (1) generalizing the theory of causal abstraction from mechanism replacement (i.e., hard and soft interventions) to arbitrary mechanism transformation (i.e., functionals from old mechanisms to new mechanisms), (2) providing a flexible, yet precise formalization for the core concepts of polysemantic neurons, the linear representation hypothesis, modular features, and graded faithfulness, and (3) unifying a variety of mechanistic interpretability methods in the common language of causal abstraction, namely, activation and path patching, causal mediation analysis, causal scrubbing, causal tracing, circuit analysis, concept erasure, sparse autoencoders, differential binary masking, distributed alignment search, and steering.
Interpretability without actionability: mechanistic methods cannot correct language model errors despite near-perfect internal representations
University of California San Francisco（加州大学旧金山分校）；Waymark；University of Pennsylvania（宾夕法尼亚大学）；Virginia Commonwealth University（弗吉尼亚 Commonwealth 大学）；Stanford University（斯坦福大学）
MINAR: Mechanistic Interpretability for Neural Algorithmic Reasoning
Pacific Northwest National Laboratory, Richland, WA（太平洋西北国家实验室）；Halıcıoğlu Data Science Institute, University of California, San Diego, San Diego, CA（哈利奇奥格鲁数据科学研究所，加州大学圣地亚哥分校）；Department of Computer and Information Science, University of Pennsylvania, Pennsylvaina, PA（计算机与信息科学系，宾夕法尼亚大学）；Department of Mathematics, University of Washington, Seattle, WA（数学系，华盛顿大学）
A Distributional View for Visual Mechanistic Interpretability: KL-Minimal Soft-Constraint Principle
School of Mathematics and Statistics（数学与统计学学院）；Ministry of Education Key Lab of Intelligent Networks and Network Security（教育部智能网络与网络安全重点实验室）；Shanghai Innovation Institute（上海创新研究院）；Fudan University（复旦大学）
Prisma: An Open Source Toolkit for Mechanistic Interpretability in Vision and Video
Mila Quebec（蒙特利尔大学）；McGill University（麦吉尔大学）；Meta；Université de Montréal（蒙特利尔大学）；Imperial College London（伦敦帝国理工学院）；Fraunhofer Heinrich Hertz Institute（弗劳恩霍夫 Heinrich Hertz 研究所）；Technological University Dublin（都柏林技术大学）；Apollo Research（Apollo 研究所）
Mechanistic Interpretability of Fine-Tuned Vision Transformers on Distorted Images: Decoding Attention Head Behavior for Transparent and Trustworthy AI
Krembil Research Institute, University Health Network（肯贝尔研究所，大学健康网络）
The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysis
Boston University（波士顿大学）；University of Mannheim（曼海姆大学）；Northeastern University（东北大学）；Anthropic（Anthropic公司）；Massachusetts Institute of Technology（麻省理工学院）；Stanford University（斯坦福大学）；Technion – Israel Institute of Technology（技术学院–以色列理工学院）
Beyond Static Interpretability: Anticipating Post-SFT Mechanisms from Pre-SFT Parameters for Better Tuning
College of Computing and Data Science（计算与数据科学学院）；Nanyang Technological University（南洋理工大学）；School of Computer Science and Engineering（计算机科学与工程学院）；The Chinese University of Hong Kong（香港中文大学）
Mechanistic Interpretability of Antibody Language Models Using SAEs
Department of Statistics, University of Oxford, UK（英国牛津大学统计系）；Reticular, San Francisco, USA（美国旧金山Reticular公司）；EECS, MIT, Cambridge MA, USA（美国麻省理工学院电子工程与计算机科学系）；Leyden Laboratories BV, Leiden, The Netherlands（荷兰莱顿实验室）
A Unified Theory of Sparse Dictionary Learning in Mechanistic Interpretability: Piecewise Biconvexity and Spurious Minima
National University of Singapore（新加坡国立大学）；Indian Institute of Technology, Dhanbad（印度德里理工学院）；Chinese University of Hong Kong（香港中文大学）；Hong Kong University of Science and Technology（香港科学与技术大学）；Peking University（北京大学）
