Embodied Task Planning on ALFWorld (seen domains)
72.05Success Rate (SR)TMoW
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TMoWBackbone=Llama-3.2-1B2026.01 | 72.05 | 6.94 | |
| LLM+FTBackbone=Llama-3.2-1B, Training=Fine-tuned2026.01 | 51.78 | 13.22 | |
| SayCanPaySay Backbone=Llama-3.2-3B, Pay Backbone=Llama-3.2-1B2026.01 | 51.48 | 13.19 | |
| FLAREBackbone=Llama-3.2-3B2026.01 | 21.22 | 34.4 | |
| LLM-PlannerBackbone=Llama-3.2-3B, Adaptation=Few-shot2026.01 | 11.67 | 37.19 | |
| ZSPBackbone=Llama-3.2-3B, Adaptation=Zero-shot2026.01 | 2.32 | 49.34 |