Preference-aligned decision making on AmbiK (test)
84.6AccuracyAdaptive CLIPR
Evaluation results
| Method | Links | |
|---|---|---|
| Adaptive CLIPRFeedback=Adaptive2026.05 | 84.6 | |
| CLIPRFeedback=Iterative learning approach2026.05 | 82.6 | |
| ICL + AnswersMode=In-context learning with answers2026.05 | 82.4 | |
| TidyBot2026.05 | 78.6 | |
| CIPHER (Sem.)Distance Metric=Semantic2026.05 | 71.7 | |
| CIPHER (Lev.)Distance Metric=Levenshtein2026.05 | 69.4 | |
| GATE (15-turn)Interaction turns=152026.05 | 62.3 | |
| ICLMode=In-context learning2026.05 | 53.8 | |
| IPMethod Identity=Introspective Planning2026.05 | 51.7 | |
| Zero-shotProtocol=Zero-shot2026.05 | 50.2 |