Open-ended Writing on WritingBench
75.76ScoreEvoRubric
Evaluation Results
| Method | Links | |
|---|---|---|
| EvoRubricBackbone=Qwen3-14B2026.05 | 75.76 | |
| EvoRubricBackbone=Qwen3-8B2026.05 | 75.52 | |
| Static Rubric-RLBackbone=Qwen3-14B2026.05 | 74.91 | |
| External Evolving-RLBackbone=Qwen3-14B2026.05 | 74.49 | |
| Static Rubric-RLBackbone=Qwen3-8B2026.05 | 74.37 | |
| External Evolving-RLBackbone=Qwen3-8B2026.05 | 74.28 | |
| Gemini-2.5-proBackbone=Proprietary2026.05 | 73.69 | |
| Base ModelBackbone=Qwen3-14B2026.05 | 72.99 | |
| Base ModelBackbone=Qwen3-8B2026.05 | 71.22 | |
| GPT-4oBackbone=Proprietary2026.05 | 67.29 | |
| ARES-RL2026.05 | 38.24 | |
| Webscale2026.05 | 37.09 | |
| ARES-SFT2026.05 | 37.05 | |
| CPT2026.05 | 36.98 | |
| Natural Reasoning2026.05 | 36.77 | |
| WASHTime=1.85 ×, n=52026.05 | 4.26 | |
| WatermarkedTime=1.00 ×2026.05 | 4.1 | |
| De-markTime=43.97 ×2026.05 | 4.04 | |
| RandomWalkTime=10.03 ×2026.05 | 3.61 | |
| ToBlendTime=7.89 ×2026.05 | 2.32 |