Mathematical Reasoning on AIME 2025 (Performance %)
9.58AIME 2025 PerformanceGT
Evaluation Results
| Method | Links | |
|---|---|---|
| GTBackbone=Phi4-mini-instruct2026.05 | 9.58 | |
| CAREBackbone=Phi4-mini-instruct2026.05 | 7.81 | |
| TTRLBackbone=Phi4-mini-instruct2026.05 | 6.67 | |
| OracleBackbone=Phi4-mini-instruct2026.05 | 5.94 | |
| RandomBackbone=Phi4-mini-instruct2026.05 | 5.83 | |
| EntropyBackbone=Phi4-mini-instruct2026.05 | 5.83 | |
| ProbBackbone=Phi4-mini-instruct2026.05 | 4.9 | |
| VanillaBackbone=Phi4-mini-instruct2026.05 | 3.33 |