Interactive Task Solving on ALFWorld
63.65Average task-level performanceATG
Evaluation Results
| Method | Links | |
|---|---|---|
| ATGBackbone=Llama-3-8B2026.07 | 63.65 | |
| ATGBackbone=Gemma-7B2026.07 | 58.71 | |
| ATGBackbone=Mistral-7B2026.07 | 55.73 | |
| ReActBackbone=GPT-42026.07 | 41.24 | |
| PoGBackbone=Gemma-7B2026.07 | 32.85 | |
| ToTBackbone=Gemma-7B2026.07 | 25.91 | |
| PoGBackbone=Mistral-7B2026.07 | 23.72 | |
| PoGBackbone=Llama-3-8B2026.07 | 21.17 | |
| ToTBackbone=Mistral-7B2026.07 | 18.25 | |
| ToTBackbone=Llama-3-8B2026.07 | 15.33 | |
| CAMELBackbone=Mistral-7B2026.07 | 11.68 | |
| ReflexionBackbone=Mistral-7B2026.07 | 8.84 | |
| CAMELBackbone=Llama-3-8B2026.07 | 8.76 | |
| CAMELBackbone=Gemma-7B2026.07 | 7.66 | |
| ReActBackbone=GPT-3.5-Turbo2026.07 | 7.3 | |
| ReActBackbone=Mistral-7B2026.07 | 6.57 | |
| ReflexionBackbone=Gemma-7B2026.07 | 5.11 | |
| ReActBackbone=Gemma-7B2026.07 | 4.38 | |
| ReflexionBackbone=Llama-3-8B2026.07 | 4.38 | |
| ReActBackbone=Llama-3-8B2026.07 | 3.29 |