[anonymousDan]

Can anyone elaborate on what these attribution techniques might be? My understanding is that explainability is a bit of a holy grail for DNN research - are reliable techniques starting to emerge?