Towards Compositionality in Concept Learning

TL;DR

This paper introduces Compositional Concept Extraction (CCE), a method to automatically discover more compositional and interpretable concept representations in foundation models, improving downstream task performance.

Contribution

The paper proposes CCE, a novel unsupervised method that identifies more compositional concepts, enhancing interpretability and accuracy in concept-based models.

Findings

CCE finds more compositional concepts than baselines

CCE improves accuracy on four downstream classification tasks

CCE is evaluated on five datasets across image and text data

Abstract

Concept-based interpretability methods offer a lens into the internals of foundation models by decomposing their embeddings into high-level concepts. These concept representations are most useful when they are compositional, meaning that the individual concepts compose to explain the full sample. We show that existing unsupervised concept extraction methods find concepts which are not compositional. To automatically discover compositional concept representations, we identify two salient properties of such representations, and propose Compositional Concept Extraction (CCE) for finding concepts which obey these properties. We evaluate CCE on five different datasets over image and text data. Our evaluation shows that CCE finds more compositional concept representations than baselines and yields better accuracy on four downstream classification tasks. Code and data are available at https://github.com/adaminsky/compositional_concepts .