Towards Worst-Case Guarantees with Scale-Aware Interpretability

TL;DR

This paper proposes a scale-aware interpretability framework for neural networks, leveraging physics-inspired renormalisation techniques to provide robustness and formal guarantees on feature influence across resolutions.

Contribution

It introduces a unifying research agenda combining physics-based methods with AI interpretability to improve robustness and faithfulness of explanations.

Findings

Framework based on renormalisation from physics for interpretability

Synthesis of interdisciplinary research into practical tools

Potential for formal robustness guarantees in model explanations

Abstract

Neural networks organize information according to the hierarchical, multi-scale structure of natural data. Methods to interpret model internals should be similarly scale-aware, explicitly tracking how features compose across resolutions and guaranteeing bounds on the influence of fine-grained structure that is discarded as irrelevant noise. We posit that the renormalisation framework from physics can meet this need by offering technical tools that can overcome limitations of current methods. Moreover, relevant work from adjacent fields has now matured to a point where scattered research threads can be synthesized into practical, theory-informed tools. To combine these threads in an AI safety context, we propose a unifying research agenda -- \emph{scale-aware interpretability} -- to develop formal machinery and interpretability tools that have robustness and faithfulness properties supported by statistical physics.