Towards Reasonable Concept Bottleneck Models

arXiv:2506.05014v2 Announce Type: replace-cross Abstract: We propose a novel, flexible, and efficient framework for designing Concept Bottleneck Models (CBMs) that enables practitioners to explicitly encode and extend their prior knowledge and beliefs about the concept-concept (C-C) and concept-task (C o Y) relationships within the model's reasoning when making predictions. The resulting extbf{C}oncept extbf{REA}soning extbf{M}odels (CREAMs) architecturally encode arbitrary types of C-C relationships such as mutual exclusivity, hierarchical associations, and/or correlations, as well as potentially sparse C o Y relationships. Moreover, CREAM can optionally incorporate a regularized side-channel to complement the potentially {incomplete concept sets}, achieving competitive task performance while encouraging predictions to be concept-grounded. To evaluate CBMs in such settings, we introduce a C o Y agnostic metric that quantifies interpretability when predictions partially rely on the side-channel. In our experiments, we show that, without additional computational overhead, CREAM models support efficient interventions, can avoid concept leakage, and achieve black-box-level performance under missing concepts. We further analyze how an optional side-channel affects interpretability and intervenability. Importantly, the side-channel enables CBMs to remain effective even in scenarios where only a limited number of concepts are available.

Related

• Countering the Over-Reliance Trap: Mitigating Object Hallucination for LVLMs via a Self-Validation Framework
• Towards an Appropriate Level of Reliance on AI: A Preliminary Reliance-Control Framework for AI in Software Engineering

Source: arXiv cs.AI | 2026-04-14