Average Reward on AlfWorld (Agent Task)
59Average RewardCeiling Model
Evaluation Results
| Method | Links | |
|---|---|---|
| Ceiling Model2025.07 | 59 | |
| W2SG with MCTS2025.07 | 57.5 | |
| W2SG with Tree DPO2025.07 | 56 | |
| SFT Strong Model + Best of N2025.07 | 55.2 | |
| SFT Strong ModelBase Model=Llama-2-13b+ETO2025.07 | 53.7 | |
| SFT Strong ModelBase Model=Llama-2-13b+SFT2025.07 | 51.5 | |
| SFT Weak ModelBase Model=Llama-2-7b+SFT2025.07 | 44.8 |