Interactive Task Learning on ALFWorld (val_seen)
97.7Pick Success RateGiGPOw/std
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GiGPOw/stdBase Model=Qwen2.5-7B-Instruct2026.06 | 97.7 | 98.8 | 89.3 | 82.7 | 83.7 | 79.2 | 90.8 | |
| ECPOBase Model=Qwen2.5-1.5B-Instruct2026.06 | 96.7 | 95.7 | 87 | 94.4 | 100 | 81.7 | 92.7 | |
| ECPOBase Model=Qwen2.5-7B-Instruct2026.06 | 94.4 | 100 | 79.7 | 82.2 | 91.2 | 93.3 | 91.9 | |
| GLM-5.1Model Family=Closed-source LLMs2026.06 | 92.9 | 82.9 | 66.7 | 72.7 | 85.4 | 65 | 77.3 | |
| PPOBase Model=Qwen2.5-7B-Instruct2026.06 | 92.3 | 92.5 | 80.3 | 64 | 89.5 | 68.8 | 80.4 | |
| Gemini-3.5-FlashModel Family=Closed-source LLMs2026.06 | 91.2 | 67.1 | 33.3 | 51.5 | 33.3 | 57.5 | 57.6 | |
| GRPOBase Model=Qwen2.5-7B-Instruct2026.06 | 90.8 | 89.3 | 72.5 | 66.1 | 74.7 | 64.7 | 77.6 | |
| RLOOBase Model=Qwen2.5-1.5B-Instruct2026.06 | 88.3 | 71 | 66.4 | 52.8 | 62.8 | 56.9 | 69.7 | |
| RLOOBase Model=Qwen2.5-7B-Instruct2026.06 | 87.6 | 87.3 | 71.9 | 78.2 | 81.3 | 48.9 | 75.5 | |
| GRPOBase Model=Qwen2.5-1.5B-Instruct2026.06 | 85.3 | 84.5 | 59.7 | 53.7 | 78.2 | 53.5 | 72.8 | |
| GiGPOw/stdBase Model=Qwen2.5-1.5B-Instruct2026.06 | 83.2 | 94.6 | 84.1 | 82.2 | 84.2 | 91.7 | 87.5 | |
| PPOBase Model=Qwen2.5-1.5B-Instruct2026.06 | 64.8 | 57.1 | 46.4 | 40.5 | 60.6 | 47.4 | 54.4 | |
| ReflexionBase Model=Qwen2.5-7B-Instruct2026.06 | 62 | 44.9 | 36.3 | 41.6 | 30.9 | 23.8 | 42.7 | |
| ReActBase Model=Qwen2.5-7B-Instruct2026.06 | 48.5 | 34.3 | 18.2 | 35.4 | 13.2 | 17.6 | 31.2 | |
| ReflexionBase Model=Qwen2.5-1.5B-Instruct2026.06 | 35.3 | 21.7 | 19.4 | 22.2 | 13.6 | 3.7 | 21.8 | |
| PromptingBase Model=Qwen2.5-7B-Instruct2026.06 | 33.4 | 19.3 | 2.8 | 21.6 | 6.9 | 3.2 | 14.8 | |
| ReActBase Model=Qwen2.5-1.5B-Instruct2026.06 | 17.4 | 15.7 | 7.7 | 20.5 | 6.2 | 2 | 12.8 | |
| PromptingBase Model=Qwen2.5-1.5B-Instruct2026.06 | 5.9 | 3.3 | 4.2 | 5.5 | 9.7 | 0 | 4.1 |