Paper

TRUER: Test-Time Reward Correction Using an Error-Aware Rectifier for Reinforcement Learning · ByteTrail