Critique on CriticBench (test)
94.1Math ScoreQwen3-32B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen3-32BModel=Qwen3-32B2026.05 | 94.1 | 67.36 | 94.41 | 56.12 | 78 | |
| CIPOBase Model=Qwen3-4B, Training=+CIPO2026.05 | 94 | 61.74 | 90.42 | 55.94 | 75.53 | |
| GRPOBase Model=Qwen3-4B, Training=+GRPO2026.05 | 93.87 | 61.55 | 90.51 | 54.79 | 75.18 | |
| InitialBase Model=Qwen3-4B2026.05 | 92.82 | 58.22 | 83.68 | 50.72 | 71.36 | |
| Qwen2.5-72B-InstructModel=Qwen2.5-72B-Instruct2026.05 | 89.81 | 64.94 | 91.99 | 63.58 | 77.58 |