ConceptTracer: Interactive Analysis of Concept Saliency and Selectivity in Neural Representations

TL;DR

ConceptTracer is an interactive tool that uses information-theoretic measures to analyze and interpret neural network representations by identifying neurons responsive to human-understandable concepts.

Contribution

It introduces a novel framework combining saliency and selectivity measures for exploring concept encoding in neural networks, demonstrated on TabPFN.

Findings

Facilitates discovery of interpretable neurons in neural networks

Enables analysis of concept saliency and selectivity in representations

Provides a practical framework for understanding concept encoding

Abstract

Neural networks deliver impressive predictive performance across a variety of tasks, but they are often opaque in their decision-making processes. Despite a growing interest in mechanistic interpretability, tools for systematically exploring the representations learned by neural networks in general, and tabular foundation models in particular, remain limited. In this work, we introduce ConceptTracer, an interactive application for analyzing neural representations through the lens of human-interpretable concepts. ConceptTracer integrates two information-theoretic measures that quantify concept saliency and selectivity, enabling researchers and practitioners to identify neurons that respond strongly to individual concepts. We demonstrate the utility of ConceptTracer on representations learned by TabPFN and show that our approach facilitates the discovery of interpretable neurons. Together, these capabilities provide a practical framework for investigating how neural networks like TabPFN encode concept-level information. ConceptTracer is available at https://github.com/ml-lab-htw/concept-tracer.