Safe Mobile Manipulation GPT-3.5 (test)
91SRIntrospective Planning
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Introspective PlanningPrediction Mode=Direct, LLM Backbone=GPT-3.52024.02 | 91 | 69 | 0 | 16.7 | 0.5 | 80 | 12 | 1 | |
| Introspective PlanningPrediction Mode=Conformal, LLM Backbone=GPT-3.52024.02 | 85 | 71.5 | 9 | 26.3 | 0 | 51 | 37 | 0 | |
| KnowNoPrediction Mode=Conformal, LLM Backbone=GPT-3.52024.02 | 84.5 | 94 | 88.5 | 27.3 | 0.5 | 10 | 83.5 | 12 | |
| Prompt SetPrediction Mode=Direct, LLM Backbone=GPT-3.52024.02 | 69.5 | 57 | 20.5 | 57.3 | 5.5 | 47 | 28.5 | 11.5 | |
| Prompt Set + CoTPrediction Mode=Direct, LLM Backbone=GPT-3.52024.02 | 69.5 | 63 | 6.4 | 48.3 | 4.5 | 55.5 | 24.5 | 6.5 | |
| Retrieval-Q-CoTPrediction Mode=Direct, LLM Backbone=GPT-3.52024.02 | 66 | 54 | 3.8 | 55.4 | 6.5 | 47 | 24 | 7.5 | |
| Auto-CoTPrediction Mode=Direct, LLM Backbone=GPT-3.52024.02 | 44 | 38 | 1.3 | 75.4 | 7 | 25 | 37.5 | 7.5 |