Mathematical Problem Solving on CNMO 2024 (Avg@8)
81.94Avg@8EvoTrainer
Evaluation Results
| Method | Links | |
|---|---|---|
| EvoTrainerConfiguration=Proposed Method2026.06 | 81.94 | |
| RAGEN v2 SNR FilteringConfiguration=Algorithmic Baseline2026.06 | 80.56 | |
| GRPO + Clip-HigherConfiguration=Algorithmic Baseline2026.06 | 79.17 | |
| AutoResearchConfiguration=Reference2026.06 | 78.47 | |
| Human-engineered RLConfiguration=Reference2026.06 | 77.78 | |
| RAGEN v1 FilteringConfiguration=Algorithmic Baseline2026.06 | 77.78 | |
| GSPO + Clip-HigherConfiguration=Algorithmic Baseline2026.06 | 76.39 | |
| Group-Variance FilteringConfiguration=Algorithmic Baseline2026.06 | 76.39 | |
| Base model (no RL)RL=None2026.06 | 75 |