Skill retrieval on MedCalc-Bench
92.3nDCG@1Qwen3-235B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-235BRetrieval Stage=Reranker, Candidate selection=top-50 candidates retrieved by BM252026.04 | 92.3 | 92.4 | |
| Qwen3-32BRetrieval Stage=Reranker, Candidate selection=top-50 candidates retrieved by BM252026.04 | 91.2 | 91.9 | |
| Mistral3.1-24BRetrieval Stage=Reranker, Candidate selection=top-50 candidates retrieved by BM252026.04 | 91.1 | 91.9 | |
| Llama-3.3-70BRetrieval Stage=Reranker, Candidate selection=top-50 candidates retrieved by BM252026.04 | 89.5 | 91.3 | |
| Llama-3.1-8BRetrieval Stage=Reranker, Candidate selection=top-50 candidates retrieved by BM252026.04 | 86 | 88.9 | |
| Qwen3-4BRetrieval Stage=Reranker, Candidate selection=top-50 candidates retrieved by BM252026.04 | 85.9 | 88.9 | |
| BGERetrieval Stage=First-stage2026.04 | 41.4 | 54.5 | |
| TF-IDFRetrieval Stage=First-stage2026.04 | 37.5 | 52 | |
| ContrieverRetrieval Stage=First-stage2026.04 | 34.9 | 49.4 | |
| BM25Retrieval Stage=First-stage2026.04 | 29.3 | 44.7 | |
| HybridRetrieval Stage=First-stage2026.04 | 29.3 | 47.4 |