S2J: Bridging the Gap Between Solving and Judging Ability in Generative Reward Models

Open in new window