Search Relevance Evaluation on INDUSTRY
73.07ACC (5-way)Qwen3-32B-Mode-Balanced-RL
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen3-32B-Mode-Balanced-RLOptimization=Mode-Balanced RL2026.02 | 73.07 | 85.57 | 82.4 | 82.2 | 69.4 | 72.95 | |
| Qwen3-32B-Stepwise-GRPOOptimization=Stepwise-GRPO2026.02 | 73.06 | 85.02 | 81.89 | 82 | 68.62 | 72.71 | |
| Qwen3-32B-GRPOOptimization=GRPO2026.02 | 72.75 | 84.9 | 82.11 | 82.06 | 70.08 | 72.73 | |
| Qwen3-32B-SFT (AFRL)Training Strategy=AFRL2026.02 | 70.34 | 82.47 | 79.43 | 80.87 | 67.06 | 69.31 | |
| Qwen2.5-72B-InstructScoring Strategy=Standard2026.02 | 68.53 | 82.07 | 82.32 | 82.6 | 65.93 | 68.36 |