Text-based reasoning on ScienceWorld
88Running Max ReturnICRL Preset
Evaluation results
| Method | Links | |
|---|---|---|
| ICRL Presetconfiguration=Preset2025.05 | 88 | |
| ICRL Autonomousconfiguration=Autonomous2025.05 | 87 | |
| Self-Refine2025.05 | 83 | |
| Best-of-N2025.05 | 75 | |
| Reflexion2025.05 | 74 | |
| ReAct2025.05 | 69 |