Mending the Holes: Mitigating Reward Hacking in Reinforcement Learning for Multilingual Translation.

Yifeng Liu, Siqi Ouyang, Yatish Hosmane Revanasiddappa, Lei Li

Browse the full ACL paper archive.