Long-form writing on LongBench-Write-en ([4k, 20k))
21.5Sl ScoreQwen-2.5-14B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen-2.5-14Btuning=LongDPO2025.02 | 21.5 | 86.04 | |
| Qwen-2.5-14Btuning=Base2025.02 | 19.57 | 82.35 | |
| Qwen-2.5-14Btuning=DPO2025.02 | 18.33 | 80.2 | |
| GPT-4o2025.02 | 6.2 | 81.2 | |
| QWQ2025.02 | 0.26 | 89.39 | |
| Llama3.1-8B-instruct2025.02 | 0 | 57.6 | |
| Llama3.1-70B-instruct2025.02 | 0 | 78 |