Vision Transformer Interpretability via Prediction of Image Reflected Relevance Among Tokens.

Kento Sago, Kazuhiro Hotta

Browse the full ICPRAM paper archive.