Embodied Task Planning on VirtualHome Novel Apartment (Unseen)
82.9Simple Success RateGPT3.5-MCTS
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT3.5-MCTSbackbone=GPT-3.5, search=MCTS, mode=few-shot, prompt_candidates=2002023.05 | 82.9 | 71.9 | 79.3 | 70.4 | 38.8 | |
| GPT3.5 Policybackbone=GPT-3.5, mode=few-shot, prompt_candidates=2002023.05 | 74.3 | 43.3 | 67.8 | 54 | 6.9 | |
| finetuned GPT2 policybackbone=GPT-2, training_trajectories=10000, mode=fine-tuned2023.05 | 65.5 | 39.9 | 33.4 | 12.8 | 1.1 | |
| UCTsearch=UCT, commonsense knowledge=false2023.05 | 0 | 0 | 0 | 0 | 0 |