Skip to content
conftrace_
2025 JMLR JMLR 2025

Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability

Abstract

Atticus Geiger, Duligur Ibeling, Amir Zur, Maheep Chaudhary, Sonakshi Chauhan, Jing Huang, Aryaman Arora, Zhengxuan Wu, Noah Goodman, Christopher Potts, Thomas Icard; 26(83):1−64, 2025.

Authors

Atticus Geiger, Duligur Ibeling, Amir Zur, Maheep Chaudhary, Sonakshi Chauhan, Jing Huang, Aryaman Arora, Zhengxuan Wu, Noah Goodman, Christopher Potts, Thomas Icard

Topics

Artificial Intelligence > Core AI > Causal Inference Artificial Intelligence > Core AI > Interpretability

Keywords

causal inference mechanistic interpretability causal abstraction
Download PDF

More from JMLR 2025

All JMLR 2025 papers →
GraphNeuralNetworks.jl: Deep Learning on Graphs with Julia
Towards Optimal Branching of Linear and Semidefinite Relaxations for Neural Network Robustness Certification
Implicit vs Unfolded Graph Neural Networks
Random Pruning Over-parameterized Neural Networks Can Improve Generalization: A Training Dynamics Analysis
On Inference for the Support Vector Machine
Integral Probability Metrics Meet Neural Networks: The Radon-Kolmogorov-Smirnov Test