Fetching the paper…

AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories · Around