An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations

TL;DR

This paper systematically studies noise in concept bottleneck models, revealing its impact on performance and interpretability, and proposes a two-stage framework using sharpness-aware minimization and uncertainty-based correction to enhance robustness.

Contribution

It is the first to analyze noise effects in CBMs and introduces a novel two-stage mitigation framework combining sharpness-aware training and uncertainty-based concept correction.

Findings

Moderate noise significantly degrades CBM performance and interpretability.

Certain concepts are more susceptible to noise, causing major performance drops.

The proposed framework improves robustness and maintains interpretability under noisy conditions.

Abstract

Concept bottleneck models (CBMs) ensure interpretability by decomposing predictions into human interpretable concepts. Yet the annotations used for training CBMs that enable this transparency are often noisy, and the impact of such corruption is not well understood. In this study, we present the first systematic study of noise in CBMs and show that even moderate corruption simultaneously impairs prediction performance, interpretability, and the intervention effectiveness. Our analysis identifies a susceptible subset of concepts whose accuracy declines far more than the average gap between noisy and clean supervision and whose corruption accounts for most performance loss. To mitigate this vulnerability we propose a two-stage framework. During training, sharpness-aware minimization stabilizes the learning of noise-sensitive concepts. During inference, where clean labels are unavailable, we rank concepts by predictive entropy and correct only the most uncertain ones, using uncertainty as a proxy for susceptibility. Theoretical analysis and extensive ablations elucidate why sharpness-aware training confers robustness and why uncertainty reliably identifies susceptible concepts, providing a principled basis that preserves both interpretability and resilience in the presence of noise.