Can we Constrain Concept Bottleneck Models to Learn Semantically Meaningful Input Features?

Open in new window