Reward Prediction on V-Droid trajectories
93.1AccuracyPRORE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PRORELast N States=Proactive2025.09 | 93.1 | 93.4 | |
| Step-CriticLast N States=Full2025.09 | 89.6 | 91 | |
| DistRLLast N States=Full2025.09 | 87 | 88.9 | |
| WebRLLast N States=Full2025.09 | 85.2 | 87 | |
| Step-CriticLast N States=12025.09 | 85.2 | 87 | |
| DigiRLLast N States=Full2025.09 | 84.5 | 87.1 | |
| WebRLLast N States=12025.09 | 82.6 | 85.1 | |
| DigiRLLast N States=12025.09 | 76.5 | 80.3 | |
| DistRLLast N States=12025.09 | 71.3 | 76.9 |