E-commerce Search Relevance Classification on Gold Benchmark Offline (offline)
49.48C1 F1 ScoreDyKnow-RAG
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DyKnow-RAGInitialization=SFT2025.10 | 49.48 | 64.1 | 45.64 | 82.02 | 60.33 | 73.26 | |
| DyKnow-RAGInitialization=DPO2025.10 | 47.96 | 65.21 | 44.98 | 83.66 | 60.45 | 75.19 | |
| RAG GRPO2025.10 | 47.54 | 62.72 | 42.82 | 80.18 | 58.33 | 72.55 | |
| RAG DPO2025.10 | 45.66 | 65.07 | 44.26 | 83.54 | 59.67 | 74.91 | |
| SFT-only2025.10 | 41.97 | 63.35 | 39.44 | 79.55 | 58.14 | 71.88 | |
| RAG SFT2025.10 | 40.47 | 62.59 | 39.21 | 80.37 | 57.06 | 71.92 |