Enhancing CBMs Through Binary Distillation with Applications to Test-Time Intervention

2025-03-09

AI Summary

To address the opacity of concept-to-label mappings in Concept Bottleneck Models (CBMs) and the persistent trade-off between interpretability and predictive performance, this paper proposes a binary distillation-based method that enhances CBMs with interpretable decision trees. Specifically, it introduces Fast Interpretable Greedy Sum-Trees (FIGS) into the CBM framework for the first time, distilling the opaque black-box concept-to-label mapping into a compact, human-readable binary tree. This tree enables adaptive, test-time intervention on salient concepts. Experiments across four benchmark datasets demonstrate that the method preserves high predictive accuracy while achieving high-fidelity attribution of concept interactions. Moreover, it significantly improves model performance under minimal human intervention—e.g., editing just a few concept values—thereby effectively supporting human-in-the-loop decision-making in real-world applications.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Interpretability, Explainability, and Transparency
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Concept bottleneck models~(CBM) aim to improve model interpretability by predicting human level ``concepts"in a bottleneck within a deep learning model architecture. However, how the predicted concepts are used in predicting the target still either remains black-box or is simplified to maintain interpretability at the cost of prediction performance. We propose to use Fast Interpretable Greedy Sum-Trees~(FIGS) to obtain Binary Distillation~(BD). This new method, called FIGS-BD, distills a binary-augmented concept-to-target portion of the CBM into an interpretable tree-based model, while mimicking the competitive prediction performance of the CBM teacher. FIGS-BD can be used in downstream tasks to explain and decompose CBM predictions into interpretable binary-concept-interaction attributions and guide adaptive test-time intervention. Across $4$ datasets, we demonstrate that adaptive test-time intervention identifies key concepts that significantly improve performance for realistic human-in-the-loop settings that allow for limited concept interventions.

Problem

• Improve interpretability of concept bottleneck models (CBM)
• Maintain prediction performance while enhancing interpretability
• Enable adaptive test-time intervention using interpretable binary concepts

Innovation

• Uses Fast Interpretable Greedy Sum-Trees (FIGS)
• Implements Binary Distillation (BD) for interpretability
• Enables adaptive test-time intervention via binary concepts