Open-ended Chat on Arena-Hard v2.0
39.2ScoreRubicon-Preview
Evaluation Results
| Method | Links | |
|---|---|---|
| Rubicon-PreviewBackbone=Qwen3-30B-A3B, Training Method=rubric-anchor RL2026.05 | 39.2 | |
| Qwen3-30B-A3B + Hybrid RLBackbone=Qwen3-30B-A3B, Training Method=Hybrid RL2026.05 | 38.5 | |
| Qwen3-30B-A3BBackbone=Qwen3-30B-A3B2026.05 | 30.6 |