Relevance Prediction on Balanced Eval Set B-Eval
70.19Class-1 ScoreTaoSR-AGRL
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TaoSR-AGRLBackbone=TbStar-DPO2025.10 | 70.19 | 73.63 | 49.28 | 83.94 | 86.18 | 69.26 | 78.03 | |
| GRPO-PRBackbone=TbStar-DPO2025.10 | 69.88 | 73.29 | 47.11 | 83.63 | 86.16 | 68.48 | 77.64 | |
| GRPOBackbone=TbStar-DPO2025.10 | 69.8 | 72.5 | 49.13 | 83.85 | 85.97 | 68.82 | 77.74 | |
| TbStar-DPOBackbone=TbStar (42B MoE)2025.10 | 69.3 | 72.86 | 44.5 | 83.04 | 86.02 | 67.43 | 76.94 |