Granular Concept Circuits: Toward a Fine-Grained Circuit Discovery for Concept Representations

TL;DR

This paper introduces Granular Concept Circuits (GCC), a novel method for fine-grained discovery of concept-specific circuits in deep vision models, enhancing interpretability by identifying how specific visual concepts are encoded.

Contribution

The paper presents the first approach to automatically discover circuits tied to specific visual concepts at a fine-grained level in deep models.

Findings

GCC effectively identifies concept-specific circuits across various models.

The method improves interpretability of deep vision models.

GCC captures multiple concepts within a single model.

Abstract

Deep vision models have achieved remarkable classification performance by leveraging a hierarchical architecture in which human-interpretable concepts emerge through the composition of individual neurons across layers. Given the distributed nature of representations, pinpointing where specific visual concepts are encoded within a model remains a crucial yet challenging task. In this paper, we introduce an effective circuit discovery method, called Granular Concept Circuit (GCC), in which each circuit represents a concept relevant to a given query. To construct each circuit, our method iteratively assesses inter-neuron connectivity, focusing on both functional dependencies and semantic alignment. By automatically discovering multiple circuits, each capturing specific concepts within that query, our approach offers a profound, concept-wise interpretation of models and is the first to identify circuits tied to specific visual concepts at a fine-grained level. We validate the versatility and effectiveness of GCCs across various deep image classification models.