List your company

Reward Prediction on OSWorld

92Reward AccuracyPRORE
80.5683.5386.589.47Sep 26, 2025

Evaluation results

MethodLinks
PROREEvaluator Agent=Claude-Sonnet-4.52025.0992
DigiRLEvaluator Agent=Claude-Sonnet-4.52025.0988
DistRLEvaluator Agent=Claude-Sonnet-4.52025.0988
WebRLEvaluator Agent=Claude-Sonnet-4.52025.0986
StepCriticEvaluator Agent=Claude-Sonnet-4.52025.0981