Reversed Attention: On The Gradient Descent Of Attention Layers In GPT.

Shahar Katz, Lior Wolf

Browse the full NAACL paper archive.