Action Success Rate on ALFWorld
0.42Average Success RateGPT-4o
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GPT-4o2025.12 | 0.42 | 0.53 | 0.67 | 0.22 | 0 | 0.25 | 0.47 | |
| Qwen-2.5-VLModel Scale=72B2025.12 | 0.32 | 0.38 | 0.14 | 0.5 | 0.27 | 0.63 | 0.21 | |
| Qwen-2.5-VLModel Scale=32B2025.12 | 0.21 | 0.44 | 0.25 | 0 | 0.2 | 0.16 | 0.2 | |
| GTR2025.12 | 0.16 | 0.44 | 0.14 | 0 | 0.14 | 0.05 | 0 | |
| GTR-TurboGuidance=KL2025.12 | 0.15 | 0.4 | 0.09 | 0 | 0.14 | 0.08 | 0.07 | |
| LLaMA-AdapterNote=Reported in previous work2025.12 | 0.13 | 0.17 | 0.1 | 0.27 | 0.22 | 0 | 0 | |
| GTR-TurboGuidance=SFT2025.12 | 0.12 | 0.36 | 0 | 0.13 | 0 | 0.125 | 0 | |
| Qwen-2.5-VLModel Scale=7B, Variant=sft2025.12 | 0.08 | 0.26 | 0.07 | 0.09 | 0 | 0.05 | 0 | |
| RL4VLM2025.12 | 0.08 | 0.4 | 0 | 0 | 0 | 0 | 0 | |
| CNN+RLNote=Reported in previous work2025.12 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |