cs.CLNov 10, 2026

Mechanistic Interpretability Needs a Logic of Inference

Authors: Zhimin Hu, Chen Jiang, Hanbo Xie, N. Apurva Ratan Murty, Sashank Varma

Organizations: Georgia Tech · McGill University

Abstract

Neural networks are increasingly studied not merely as predictive systems, but as complex systems whose internal computations are themselves objects of scientific inquiry. Recent progress in mechanistic interpretability (MI) has made it possible to identify neurons, features, circuits, population structures, and dynamical patterns associated with model behavior. As these tools improve, attention is shifting from what can be observed inside a model to how those observations support mechanistic explanations. We examine the parallel efforts of neuroscience and MI to address this question and identify recurring inferential bottlenecks when inferring mechanisms from complex neural systems. We propose five principles and organize them into a recursive logic for a more rigorous procedure of mechanistic inference. These principles provide a framework that supports scientific discovery in two senses: it helps turn local interpretability findings into cumulative knowledge of mechanisms within models, and it makes those mechanisms sufficiently precise to generate testable hypotheses about domains beyond the models. Evidence beyond MI is ultimately required to establish what model mechanisms reveal about the world.

Figures & tables

Explore similar work

CardsList
  1. Tracking Equivalent Mechanistic Interpretations Across Neural Networks

    Mar 31, 2026Alan Sun, Mariya TonevaMechanistic InterpretabilityTransformer Interpretability

  2. From Mechanistic to Compositional Interpretability

    May 9, 2026Ward Gauderis, Thomas Dooms, Steven T. Homer +2Mechanistic InterpretabilityNeural Network Interpretability

  3. From Circuit Evidence to Mechanistic Theory: An Inductive Logic Approach

    May 20, 2026Nura Aljaafari, Danilo S. Carvalho, Andre FreitasCircuit DiscoveryMechanistic Interpretability