Text Generation and Reranking on RJUA
34.1Rouge-LEAPO
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| EAPOBackbone=Qwen2.5-7B-Instruct2026.05 | 34.1 | 97.7 | 38.8 | 99.4 | 67.5 | |
| W-REINFORCEBackbone=Qwen2.5-7B-Instruct2026.05 | 33.6 | 97.3 | 37.1 | 99.2 | 66.8 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.05 | 32.3 | 96.8 | 34.6 | 99 | 65.7 | |
| PSRBackbone=Qwen2.5-7B-Instruct2026.05 | 28.7 | 94.4 | 29.4 | 97 | 62.4 | |
| NSRBackbone=Qwen2.5-7B-Instruct2026.05 | 24.6 | 85.8 | 35.9 | 99.4 | 61.4 |