The field of machine learning is moving towards increased interpretability, with a focus on developing models that provide transparent and explainable results. This shift is driven by the need for domain experts to understand and trust the predictions made by machine learning models, particularly in high-stakes applications such as medical diagnosis and decision-making. Recent work has centered on improving the interpretability of Concept Bottleneck Models (CBMs), which use human-understandable concepts to predict outcomes. Notable papers in this area include:

• Debugging Concept Bottleneck Models through Removal and Retraining, which introduces a novel method for converting concept-level user feedback into sample-level auxiliary labels to reduce bias.
• Uncertainty-Aware Concept Bottleneck Models with Enhanced Interpretability, which proposes a new classifier that learns binary class-level concept prototypes to improve robustness and interpretability.
• Automated Genomic Interpretation via Concept Bottleneck Models for Medical Robotics, which demonstrates the application of CBMs to genomic interpretation and medical automation.