Bayesian Concept Bottleneck Models with LLM Priors

TL;DR

This paper introduces BC-LLM, a Bayesian concept bottleneck model leveraging LLM priors to improve interpretability, accuracy, and robustness across diverse datasets, while providing rigorous uncertainty quantification.

Contribution

It proposes a novel Bayesian framework that uses LLMs as both concept extractors and priors, enabling flexible, interpretable modeling without predefined concept sets.

Findings

Outperforms interpretable baselines and some black-box models

Converges faster to relevant concepts

More robust to out-of-distribution samples

Abstract

Concept Bottleneck Models (CBMs) have been proposed as a compromise between white-box and black-box models, aiming to achieve interpretability without sacrificing accuracy. The standard training procedure for CBMs is to predefine a candidate set of human-interpretable concepts, extract their values from the training data, and identify a sparse subset as inputs to a transparent prediction model. However, such approaches are often hampered by the tradeoff between exploring a sufficiently large set of concepts versus controlling the cost of obtaining concept extractions, resulting in a large interpretability-accuracy tradeoff. This work investigates a novel approach that sidesteps these challenges: BC-LLM iteratively searches over a potentially infinite set of concepts within a Bayesian framework, in which Large Language Models (LLMs) serve as both a concept extraction mechanism and prior. Even though LLMs can be miscalibrated and hallucinate, we prove that BC-LLM can provide rigorous statistical inference and uncertainty quantification. Across image, text, and tabular datasets, BC-LLM outperforms interpretable baselines and even black-box models in certain settings, converges more rapidly towards relevant concepts, and is more robust to out-of-distribution samples.