Introspective Planning on KitchenAmbig (In-Distribution)
95.7Average AccuracyGATE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GATEelicitation turns=52026.05 | 95.7 | 9.6 | |
| Adaptive CLIPRelicitation turns=up to 15, early-stop signal=true2026.05 | 94.3 | 9.3 | |
| CLIPR2026.05 | 94.3 | 9.3 | |
| TidyBot2026.05 | 92.9 | 0 | |
| ICL + Answersprotocol=In-Context Learning with answers2026.05 | 88.6 | 8.1 | |
| IP (full)2026.05 | 77.1 | 3.2 | |
| CIPHER (Lev.)2026.05 | 74.3 | 3.9 | |
| CIPHER (Sem.)2026.05 | 74.3 | 3.9 | |
| Zero-shotprotocol=Zero-shot2026.05 | 52.9 | 9.6 | |
| ICLprotocol=In-Context Learning2026.05 | 40 | 3.9 |