Introspective Planning on KitchenAmbig (OOD)
97.6Average AccuracyAdaptive CLIPR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Adaptive CLIPRelicitation turns=up to 15, early-stop signal=true2026.05 | 97.6 | 3 | |
| GATEelicitation turns=52026.05 | 96.6 | 4.1 | |
| CLIPR2026.05 | 96.1 | 4.8 | |
| CIPHER (Lev.)2026.05 | 93.7 | 2.8 | |
| TidyBot2026.05 | 93.2 | 4 | |
| CIPHER (Sem.)2026.05 | 92.7 | 3 | |
| ICL + Answersprotocol=In-Context Learning with answers2026.05 | 89.3 | 10.6 | |
| IP (full)2026.05 | 86.3 | 9.8 | |
| Zero-shotprotocol=Zero-shot2026.05 | 78 | 7.3 | |
| ICLprotocol=In-Context Learning2026.05 | 66.3 | 14.3 |