Mechanistic Interpretability

Overview

The field seeks to explain how a model’s internal computations produce particular outputs, rather than relying only on input-output behavior. Researchers use sparse autoencoders and transcoders to identify features, and probes, attribution graphs, activation patching, ablation, and steering to trace or test their roles. Studies examine language-model tasks such as factual recall, reasoning, and refusal, as well as vision, audio, and scientific models. A central challenge is that a signal can be easy to decode without being causally used, while circuit explanations and interventions can vary across prompts, models, and analysis choices.

Latest papers

CardsList

427 papers58 papers in the last four weeks, up 23% on the four weeks before. 0.4% of all new papers.

  1. Circuit Distillation

    Sep 29, 2025Somin Wadhwa, Silvio Amir, Byron C. WallaceLanguage Model DistillationMechanistic Interpretability

  2. Mechanistic Interpretability with Sparse Autoencoder Neural Operators

    Sep 3, 2025Bahareh Tolooshams, Ailsa Shen, Anima AnandkumarSparse AutoencodersMechanistic Interpretability

  3. Interpretability and Generalization Bounds for Learning Spatial Physics

    Jun 18, 2025Alejandro Francisco Queiruga, Theo Gutman-Solo, Shuai JiangNeural Network GeneralizationMechanistic Interpretability

  4. Interpreting Global Perturbation Robustness of Image Models using Axiomatic Spectral Importance Decomposition

    Aug 2, 2024Róisín Luo, James McDermott, Colm O'RiordanImage Corruption RobustnessNeural Network Robustness