Problem
Research questions and friction points this paper is trying to address.
Concept Bottleneck Models
interpretability
concept entanglement
faithfulness
vision-language models
CHOQOLATE: Organizing Concept Bottleneck Latent Spaces with Choquet Integrals
🤖 AI Summary
This study addresses the issue of unfaithful representations caused by correlated concept entanglement in Concept Bottleneck Models by proposing an interpretable layer based on the 2-additive Choquet integral. By integrating the CLIP vision-language model with closed-form gradient derivation, this method drives concept organization without explicit supervision, merging correlated concepts into compact nodes whose weights directly map to Shapley values to support test-time interventions. Experiments across four datasets demonstrate that the proposed approach generates sparse and semantically coherent nodes, effectively optimizing the accuracy-interpretability trade-off while achieving debiasing performance comparable to methods requiring retraining.
📝 Abstract
Concept Bottleneck Models (CBMs) built on vision-language models such as CLIP represent a latent space as human-understandable concepts. These representations are unfaithful: related concepts are entangled, so individual scores do not reflect their intended meaning. We propose CHOQOLATE, an interpretable-by-design layer based on 2-additive Choquet integrals, which merges correlated concepts into compact nodes. Across four datasets, CHOQOLATE achieves a favorable accuracy-interpretability trade-off, with weight-sparse and semantically coherent nodes. A closed-form gradient derivation, backed by experiments, explains why Choquet layers drive this organization without explicit supervision. Choquet weights also map directly to Shapley values, which enables test-time intervention. On standard bias-mitigation benchmarks, suppressing spurious concepts after training performs on par with methods that require group annotations or retraining, while needing neither.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Concept Bottleneck Models
Choquet Integrals
Interpretability
Shapley Values
Bias Mitigation
🔎 Similar Papers
No similar papers found.
