Fetching the paper…

Wrapper Boxes: Faithful Attribution of Model Predictions to Training Data · Around