Skill Routing on Skill Routing Dataset
76Hit@1SR-Emb-8B + SR-Rank-8B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SR-Emb-8B + SR-Rank-8BEncoder=SR-Emb-8B, Reranker=SR-Rank-8B, Rank Input=full2026.03 | 76 | 80.8 | 66.6 | 71.9 | |
| SR-Emb-0.6B + SR-Rank-8BEncoder=SR-Emb-0.6B, Reranker=SR-Rank-8B, Rank Input=full2026.03 | 74.7 | 80.4 | 65.5 | 70.7 | |
| SR-Emb-0.6B + SR-Rank-0.6BEncoder=SR-Emb-0.6B, Reranker=SR-Rank-0.6B, Rank Input=full2026.03 | 74 | 79.1 | 64 | 70.4 | |
| SR-Emb-0.6B + Qwen3-Rank-8BEncoder=SR-Emb-0.6B, Reranker=Qwen3-Rank-8B, Rank Input=full2026.03 | 71.4 | 77.6 | 64.9 | 72.7 | |
| SR-Emb-0.6B + Qwen3-Rank-0.6BEncoder=SR-Emb-0.6B, Reranker=Qwen3-Rank-0.6B, Rank Input=full2026.03 | 70.7 | 76.9 | 64.9 | 72.4 | |
| Qwen3-Emb-0.6B + Qwen3-Rank-8BEncoder=Qwen3-Emb-0.6B, Reranker=Qwen3-Rank-8B, Rank Input=full2026.03 | 68.7 | 73.3 | 59.1 | 62.4 | |
| Qwen3-Emb-8B + Qwen3-Rank-8BEncoder=Qwen3-Emb-8B, Reranker=Qwen3-Rank-8B, Rank Input=full2026.03 | 68 | 74.5 | 62.3 | 69.2 | |
| SR-Emb-0.6B + GPT-4o-miniEncoder=SR-Emb-0.6B, Reranker=GPT-4o-mini, Rank Input=full2026.03 | 67.3 | — | — | — | |
| SR-Emb-0.6B + GPT-5.4-miniEncoder=SR-Emb-0.6B, Reranker=GPT-5.4-mini, Rank Input=full2026.03 | 66 | — | — | — | |
| SR-Emb-0.6BEncoder=SR-Emb-0.6B, Reranker=None, Rank Input=N/A2026.03 | 65.4 | 72.3 | 61.6 | 68.5 | |
| Qwen3-Emb-8B + GPT-4o-miniEncoder=Qwen3-Emb-8B, Reranker=GPT-4o-mini, Rank Input=full2026.03 | 64.7 | — | — | — | |
| Qwen3-Emb-8BEncoder=Qwen3-Emb-8B, Reranker=None, Rank Input=N/A2026.03 | 64 | 69.8 | 58.1 | 65.9 | |
| Qwen3-Emb-0.6B + Qwen3-Rank-0.6BEncoder=Qwen3-Emb-0.6B, Reranker=Qwen3-Rank-0.6B, Rank Input=full2026.03 | 64 | 68.4 | 55.8 | 60.4 | |
| Qwen3-Emb-8B + GPT-5.4-miniEncoder=Qwen3-Emb-8B, Reranker=GPT-5.4-mini, Rank Input=full2026.03 | 59.4 | — | — | — | |
| Qwen3-Emb-8B + Qwen3-Rank-0.6BEncoder=Qwen3-Emb-8B, Reranker=Qwen3-Rank-0.6B, Rank Input=full2026.03 | 58 | 67.2 | 58.8 | 69.4 | |
| Qwen3-Emb-0.6BEncoder=Qwen3-Emb-0.6B, Reranker=None, Rank Input=N/A2026.03 | 56 | 63.8 | 52.6 | 59.5 | |
| Qwen3-Emb-0.6B + GPT-5.4-miniEncoder=Qwen3-Emb-0.6B, Reranker=GPT-5.4-mini, Rank Input=full2026.03 | 56 | — | — | — | |
| Qwen3-Emb-0.6B + GPT-4o-miniEncoder=Qwen3-Emb-0.6B, Reranker=GPT-4o-mini, Rank Input=full2026.03 | 55.4 | — | — | — | |
| Qwen3-Emb-0.6B + GPT-5.4-miniEncoder=Qwen3-Emb-0.6B, Reranker=GPT-5.4-mini, Rank Input=nd2026.03 | 33.3 | — | — | — | |
| Qwen3-Emb-0.6B + Qwen3-Rank-8BEncoder=Qwen3-Emb-0.6B, Reranker=Qwen3-Rank-8B, Rank Input=nd2026.03 | 30.7 | 42.5 | 38.1 | 52.5 | |
| Qwen3-Emb-8B + GPT-5.4-miniEncoder=Qwen3-Emb-8B, Reranker=GPT-5.4-mini, Rank Input=nd2026.03 | 30.7 | — | — | — | |
| Qwen3-Emb-0.6B + Qwen3-Rank-0.6BEncoder=Qwen3-Emb-0.6B, Reranker=Qwen3-Rank-0.6B, Rank Input=nd2026.03 | 26.7 | 39.2 | 37.4 | 52.4 | |
| Qwen3-Emb-8B + Qwen3-Rank-8BEncoder=Qwen3-Emb-8B, Reranker=Qwen3-Rank-8B, Rank Input=nd2026.03 | 24 | 39.2 | 35.7 | 53 | |
| Qwen3-Emb-8B + Qwen3-Rank-0.6BEncoder=Qwen3-Emb-8B, Reranker=Qwen3-Rank-0.6B, Rank Input=nd2026.03 | 21.3 | 38.5 | 38.5 | 60.3 | |
| Qwen3-Emb-0.6B + GPT-4o-miniEncoder=Qwen3-Emb-0.6B, Reranker=GPT-4o-mini, Rank Input=nd2026.03 | 20.7 | — | — | — | |
| Qwen3-Emb-8B + GPT-4o-miniEncoder=Qwen3-Emb-8B, Reranker=GPT-4o-mini, Rank Input=nd2026.03 | 19.3 | — | — | — |