Mathematical Problem Solving on AIME 2024 (Avg@8)
84.17Avg@8 ScoreEvoTrainer
Evaluation Results
| Method | Links | |
|---|---|---|
| EvoTrainerConfiguration=Proposed Method2026.06 | 84.17 | |
| RAGEN v2 SNR FilteringConfiguration=Algorithmic Baseline2026.06 | 81.67 | |
| Human-engineered RLConfiguration=Reference2026.06 | 80.83 | |
| RAGEN v1 FilteringConfiguration=Algorithmic Baseline2026.06 | 80.42 | |
| GRPO + Clip-HigherConfiguration=Algorithmic Baseline2026.06 | 80 | |
| GSPO + Clip-HigherConfiguration=Algorithmic Baseline2026.06 | 79.17 | |
| Group-Variance FilteringConfiguration=Algorithmic Baseline2026.06 | 78.75 | |
| AutoResearchConfiguration=Reference2026.06 | 78.33 | |
| Base model (no RL)RL=None2026.06 | 77.5 |