AI RESEARCH

RewardBench 2: Advancing Reward Model Evaluation

arXiv CS.CL

ArXi:2506.01937v2 Announce Type: replace Reward models are used throughout the post-