Query Reformulation on 8,500 Body queries Production Search
94.2Precision@1Gold (n=100)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gold (n=100)Relevance Criteria=Loose Relevance2026.01 | 94.2 | 93.7 | 93 | |
| PRECISE-PPI (λ=0.95)Relevance Criteria=Loose Relevance2026.01 | 91.4 | 90.4 | 89.5 | |
| Sonnet-Unlb (binary)Relevance Criteria=Strict Relevance2026.01 | 83.1 | 82.7 | 82 | |
| Sonnet-Unlb (binary)Relevance Criteria=Loose Relevance2026.01 | 83.1 | 82.7 | 82 | |
| Sonnet-Unlb (prob)Relevance Criteria=Strict Relevance2026.01 | 74.9 | 74.9 | 74.9 | |
| Sonnet-Unlb (prob)Relevance Criteria=Loose Relevance2026.01 | 74.9 | 74.9 | 74.9 | |
| Gold (n=100)Relevance Criteria=Strict Relevance2026.01 | 60.6 | 60 | 61.1 | |
| PRECISE-PPI (λ=0.95)Relevance Criteria=Strict Relevance2026.01 | 55.1 | 54.6 | 55.3 |