Embodied Task Planning on RLBench Unseen domains
62.75Success RateTMoW
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TMoWBackbone=Llama-3.2-1B2026.01 | 62.75 | 8.95 | |
| SayCanPaySay Backbone=Llama-3.2-3B, Pay Backbone=Llama-3.2-1B2026.01 | 38.54 | 10.76 | |
| FLAREBackbone=Llama-3.2-3B2026.01 | 34.37 | 11.37 | |
| LLM-PlannerBackbone=Llama-3.2-3B, Adaptation=Few-shot2026.01 | 19.79 | 17.19 | |
| LLM+FTBackbone=Llama-3.2-1B, Training=Fine-tuned2026.01 | 15.63 | 17.44 | |
| ZSPBackbone=Llama-3.2-3B, Adaptation=Zero-shot2026.01 | 10.42 | 18.73 |