Fetching the paper…

RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment · Around