Reward Prediction on OSWorld
92Reward AccuracyPRORE
Evaluation Results
| Method | Links | |
|---|---|---|
| PROREEvaluator Agent=Claude-Sonnet-4.52025.09 | 92 | |
| DigiRLEvaluator Agent=Claude-Sonnet-4.52025.09 | 88 | |
| DistRLEvaluator Agent=Claude-Sonnet-4.52025.09 | 88 | |
| WebRLEvaluator Agent=Claude-Sonnet-4.52025.09 | 86 | |
| StepCriticEvaluator Agent=Claude-Sonnet-4.52025.09 | 81 |