Fetching the paper…

Interpretability Beyond Classification Output: Semantic Bottleneck Networks · Around