Embodied Task Planning on ALFWorld (unseen domains)
68.83Success Rate (SR)TMoW
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TMoWBackbone=Llama-3.2-1B2026.01 | 68.83 | 37.44 | |
| SayCanPaySay Backbone=Llama-3.2-3B, Pay Backbone=Llama-3.2-1B2026.01 | 42.04 | 40.64 | |
| LLM+FTBackbone=Llama-3.2-1B, Training=Fine-tuned2026.01 | 39.61 | 41.24 | |
| FLAREBackbone=Llama-3.2-3B2026.01 | 11.31 | 42.85 | |
| LLM-PlannerBackbone=Llama-3.2-3B, Adaptation=Few-shot2026.01 | 8.46 | 43.54 | |
| ZSPBackbone=Llama-3.2-3B, Adaptation=Zero-shot2026.01 | 2.08 | 49.68 |