Mathematical Reasoning on AIME 2024 (Performance %)
12.92Performance (%)CARE
Evaluation Results
| Method | Links | |
|---|---|---|
| CAREBackbone=Phi4-mini-instruct2026.05 | 12.92 | |
| GTBackbone=Phi4-mini-instruct2026.05 | 11.56 | |
| OracleBackbone=Phi4-mini-instruct2026.05 | 10.93 | |
| EntropyBackbone=Phi4-mini-instruct2026.05 | 8.65 | |
| TTRLBackbone=Phi4-mini-instruct2026.05 | 8.44 | |
| ProbBackbone=Phi4-mini-instruct2026.05 | 8.02 | |
| RandomBackbone=Phi4-mini-instruct2026.05 | 7.71 | |
| VanillaBackbone=Phi4-mini-instruct2026.05 | 6.46 |