Fig. 5From: Using test-time augmentation to investigate explainable AI: inconsistencies between method, model and human intuitionComparison of the canonical and averaged values of in-between model and in-between method cosine distances. Cosine distances between different encoder-decoder models for the same method (in-between model) and different methods for the same encoder-decoder model (in-between method) of canonical and averaged atom attributionsBack to article page