Fake it till You Make it: Reward Modeling as Discriminative Prediction

Re-assign community