Skill retrieval on CHAMP (Recall@K)
22.5Recall@1Llama-3.3-70B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Llama-3.3-70BRetrieval Stage=LLM-based reranker, Candidate Pool=top-50 candidates retrieved by BM252026.04 | 22.5 | 47.8 | |
| Qwen3-32BRetrieval Stage=LLM-based reranker, Candidate Pool=top-50 candidates retrieved by BM252026.04 | 22.3 | 49.1 | |
| Qwen3-235BRetrieval Stage=LLM-based reranker, Candidate Pool=top-50 candidates retrieved by BM252026.04 | 22.1 | 50.2 | |
| Qwen3-4BRetrieval Stage=LLM-based reranker, Candidate Pool=top-50 candidates retrieved by BM252026.04 | 18.5 | 44 | |
| Mistral3.1-24BRetrieval Stage=LLM-based reranker, Candidate Pool=top-50 candidates retrieved by BM252026.04 | 18 | 49.3 | |
| Llama-3.1-8BRetrieval Stage=LLM-based reranker, Candidate Pool=top-50 candidates retrieved by BM252026.04 | 15.8 | 41.6 | |
| BM25Retrieval Stage=First-stage2026.04 | 13.2 | 36.1 | |
| HybridRetrieval Stage=First-stage2026.04 | 13.2 | 41.4 | |
| BGERetrieval Stage=First-stage2026.04 | 9.8 | 34 | |
| TF-IDFRetrieval Stage=First-stage2026.04 | 7.2 | 25.7 | |
| ContrieverRetrieval Stage=First-stage2026.04 | 3.7 | 29.3 |