Adding Instructions during Pretraining: Effective way of Controlling Toxicity in Language Models.

Shrimai Prabhumoye, Mostofa Patwary, Mohammad Shoeybi, Bryan Catanzaro

Browse the full EACL paper archive.