Reward Prediction on OSWorld Chrome
93.5Reward AccuracyPRORE
Evaluation Results
| Method | Links | |
|---|---|---|
| PROREEvaluator Agent=Claude-Sonnet-4.52025.09 | 93.5 | |
| WebRLEvaluator Agent=Claude-Sonnet-4.52025.09 | 87 | |
| StepCriticEvaluator Agent=Claude-Sonnet-4.52025.09 | 87 | |
| DigiRLEvaluator Agent=Claude-Sonnet-4.52025.09 | 84.8 | |
| DistRLEvaluator Agent=Claude-Sonnet-4.52025.09 | 82.6 |