Task-oriented Interaction on To Do List Multi-turn
100Pass RateSemantic XPath
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Semantic XPathBackbone=GPT-5 mini, Scorer=Sem2026.03 | 100 | 6,132 | |
| Semantic XPathBackbone=GPT-5 mini, Scorer=Entail2026.03 | 100 | 5,649 | |
| Semantic XPathBackbone=Gemini 3 Flash, Scorer=Sem2026.03 | 100 | 5,872 | |
| Semantic XPathBackbone=Gemini 3 Flash, Scorer=Entail2026.03 | 100 | 5,526 | |
| In-contextBackbone=GPT-5 mini2026.03 | 70 | 20,398 | |
| In-contextBackbone=Gemini 3 Flash2026.03 | 70 | 23,602 | |
| Flat RAGBackbone=GPT-5 mini2026.03 | 50 | 3,043 | |
| Flat RAGBackbone=Gemini 3 Flash2026.03 | 50 | 2,959 |