Open-ended writing on DeepresearchBench
46.93Overall ScoreQwen3-8B + R2-Write-SFT + RLp
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen3-8B + R2-Write-SFT + RLpTraining=SFT + PPO2026.04 | 46.93 | 44.16 | 44.03 | 49.98 | 49.54 | |
| Qwen3-8B + R2-Write-SFTTraining=SFT2026.04 | 46 | 43.32 | 42.86 | 48.75 | 49.08 | |
| Qwen3-8B + RLTraining=PPO2026.04 | 44.97 | 43.8 | 43.03 | 45.62 | 47.43 | |
| Qwen3-4B + R2-Write-SFT + RLpTraining=SFT + PPO2026.04 | 44.95 | 42.24 | 42.08 | 46.68 | 48.79 | |
| LongWriter-Zero-32BTraining=SFT + GRPO2026.04 | 44.69 | 43.66 | 42.4 | 45.21 | 47.48 | |
| Qwen3-4B + R2-Write-SFTTraining=SFT2026.04 | 44.18 | 42.31 | 41.89 | 47.56 | 47.06 | |
| Reverse-Engineering-8BTraining=SFT2026.04 | 42.42 | 41.2 | 39.44 | 44.18 | 44.86 | |
| Qwen3-8BTraining=-2026.04 | 40.52 | 38.74 | 36.27 | 45.85 | 44.18 | |
| Qwen3-4B + RLTraining=PPO2026.04 | 39.74 | 40.25 | 37.8 | 41.22 | 39.68 | |
| Longwriter-9BTraining=SFT2026.04 | 38.68 | 36.84 | 34.45 | 43.18 | 40.26 | |
| Qwen3-4BTraining=-2026.04 | 38.11 | 36.32 | 33.41 | 43.66 | 42.33 |