Teach a Reward Model to Correct Itself: Reward Guided Adversarial Failure Discovery for Robust Reward Modeling.

Pankayaraj Pathmanathan, Furong Huang

Browse the full ACL paper archive.