Search Relevance Ranking on Manually Annotated Search Relevance (test)
0.902NDCG@1RAG-Match
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RAG-MatchNumber of Parameters=8B2026.05 | 0.902 | 0.907 | 0.916 | 0.935 | 0.833 | |
| GLM-4-9B-0414Number of Parameters=9B, Fine-tuning protocol=label-only supervised fine-tuning2026.05 | 0.846 | 0.877 | 0.891 | 0.915 | 0.801 | |
| MiniMax-SynLogicNumber of Parameters=8B, Fine-tuning protocol=label-only supervised fine-tuning2026.05 | 0.837 | 0.87 | 0.891 | 0.908 | 0.792 | |
| Qwen3-8B-RerankerNumber of Parameters=8B, Fine-tuning protocol=label-only supervised fine-tuning2026.05 | 0.826 | 0.878 | 0.892 | 0.912 | 0.796 | |
| DeepSeek-R1-0528Number of Parameters=8B, Fine-tuning protocol=label-only supervised fine-tuning2026.05 | 0.815 | 0.867 | 0.883 | 0.906 | 0.783 |