Reward Prediction on M3A trajectories
91.4AccuracyPRORE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PRORELast N States=Proactive2025.09 | 91.4 | 88.9 | |
| DigiRLLast N States=Full2025.09 | 82.6 | 80.8 | |
| Step-CriticLast N States=Full2025.09 | 82.6 | 79.6 | |
| DistRLLast N States=Full2025.09 | 81.7 | 79.6 | |
| WebRLLast N States=12025.09 | 81.7 | 79.2 | |
| WebRLLast N States=Full2025.09 | 81.7 | 79.2 | |
| Step-CriticLast N States=12025.09 | 81.7 | 79.4 | |
| DigiRLLast N States=12025.09 | 74.8 | 73.4 | |
| DistRLLast N States=12025.09 | 70.4 | 68.5 |