Mathematical Reasoning on AIME 25 (Accuracy, Token Cost)
96AccuracyHyPER
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HyPERModel=Qwen3-Next2026.02 | 96 | 0.59 | |
| HyPERModel=Qwen3-30B2026.02 | 95.3 | 0.71 | |
| DeepConfModel=Qwen3-Next2026.02 | 94 | 0.53 | |
| DeepConfModel=Qwen3-30B2026.02 | 90.7 | 0.67 | |
| SCModel=Qwen3-Next2026.02 | 88 | 1 | |
| RoEModel=Qwen3-Next2026.02 | 86.7 | 1.46 | |
| SCModel=Qwen3-30B2026.02 | 86 | 1 | |
| RoEModel=Qwen3-30B2026.02 | 84.7 | 1.64 | |
| Self-CertaintyModel=Qwen3-Next2026.02 | 83.3 | 0.95 | |
| Self-CertaintyModel=Qwen3-30B2026.02 | 80 | 1.13 |