Saved in:
Bibliographic Details
Main Authors: Park, Seonghwan, Mun, Jueun, Oh, Donghyun, Lee, Namhoon
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.16705
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911407871623168
author Park, Seonghwan
Mun, Jueun
Oh, Donghyun
Lee, Namhoon
author_facet Park, Seonghwan
Mun, Jueun
Oh, Donghyun
Lee, Namhoon
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16705
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations
Park, Seonghwan
Mun, Jueun
Oh, Donghyun
Lee, Namhoon
Machine Learning
Artificial Intelligence
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.
title An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2505.16705