Task-oriented Plan Retrieval on Meal Kit Recommendation Single-turn
80Pass RateSemantic XPath
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Semantic XPathModel=GPT-5 mini, Scorer=Entail2026.03 | 80 | 6,728 | |
| In-contextModel=GPT-5 mini, Scorer=–2026.03 | 75 | 61,069 | |
| In-contextModel=Gemini 3 Flash, Scorer=–2026.03 | 70 | 52,849 | |
| Semantic XPathModel=Gemini 3 Flash, Scorer=Entail2026.03 | 65 | 6,644 | |
| Semantic XPathModel=GPT-5 mini, Scorer=Sem2026.03 | 60 | 7,013 | |
| Semantic XPathModel=Gemini 3 Flash, Scorer=Sem2026.03 | 60 | 6,966 | |
| Flat RAGModel=Gemini 3 Flash, Scorer=–2026.03 | 30 | 3,800 | |
| Flat RAGModel=GPT-5 mini, Scorer=–2026.03 | 25 | 3,892 |