Long-form Writing on Creative-W.B.
83.17ScoreQwen2.5-14B-Writing-RL
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-14B-Writing-RLSFT=✓, RL=PPO2025.06 | 83.17 | |
| Llama3.1-8B-Writing-RLSFT=✓, RL=PPO2025.06 | 82.67 | |
| GPT-4oSFT=–, RL=–2025.06 | 80.45 | |
| Qwen2.5-14B-WritingBench-SFTSFT=✓, RL=✗2025.06 | 79.27 | |
| Llama3.1-8B-WritingBench-SFTSFT=✓, RL=✗2025.06 | 77.7 | |
| Qwen-PlusSFT=–, RL=–2025.06 | 76.78 | |
| Llama3.1-8B-Reference-SFTSFT=✓, RL=✗2025.06 | 76.7 | |
| Qwen2.5-7B-Writing-RLSFT=✓, RL=GRPO2025.06 | 75.41 | |
| Qwen2.5-7B-Writing-RLSFT=✓, RL=PPO2025.06 | 73.19 | |
| Qwen2.5-7B-WritingBench-SFT (12k)SFT=✓, RL=✗2025.06 | 70.24 | |
| Qwen2.5-7B-WritingBench-SFT (24k)SFT=✓, RL=✗2025.06 | 69.6 | |
| Qwen2.5-7B-Reference-SFTSFT=✓, RL=✗2025.06 | 68.89 | |
| Longwriter-Zero-32BSFT=✓, RL=GRPO2025.06 | 61.14 | |
| Qwen2.5-14B-InstructSFT=✗, RL=✗2025.06 | 56.89 | |
| Qwen2.5-7B-InstructSFT=✗, RL=✗2025.06 | 49.29 | |
| Llama3.1-8B-InstructSFT=✗, RL=✗2025.06 | 48.83 | |
| Longwriter-9BSFT=✓, RL=DPO2025.06 | 44.15 | |
| Suri-7BSFT=✓, RL=✗2025.06 | 18.44 |