AI RESEARCH
Towards more holistic interpretability: A lightweight disentangled Concept Bottleneck Model
arXiv CS.LG
•
ArXi:2510.15770v2 Announce Type: replace-cross Concept Bottleneck Models (CBMs) enhance interpretability by predicting human-understandable concepts as intermediate representations. However, existing CBMs often suffer from input-to-concept mapping bias and limited controllability, which restricts their practical utility and undermines the reliability of concept-based strategies.