Math Reasoning on AIME 2025 (BA@20, BA@40, BA@60)
40.6Best Answer Rate @20RoRo
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RoRoSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.05 | 40.6 | 47 | 60.5 | |
| RoRoSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.05 | 38.9 | 45.7 | 59 | |
| RandomSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.05 | 35.4 | 43.1 | 50.8 | |
| SRM-onlySetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.05 | 32.4 | 32.4 | 32.4 |