Jul 2026
Naming the Concepts Classifiers Rely On: Language-Anchored Decomposition for Faithful Explanation
Across natural-image, scene, and medical-imaging benchmarks, LAD produces spatially precise explanations that are decision-relevant under both concept insertion and deletion, while uniquely providing stable, human-interpretable concept names.
Ahsan Habib Akash, Dipkamal Bhusal, Stacey Jones et al.
· arXiv.org · 0 citations