List your company

Mathematical Problem Solving on AIME 2025 (Avg@8)

73.33Avg@8EvoTrainer
67.266868.840970.41571.9891Jun 2, 2026

Evaluation results

MethodLinks
EvoTrainerConfiguration=Proposed Method2026.0673.33
RAGEN v2 SNR FilteringConfiguration=Algorithmic Baseline2026.0672.5
Human-engineered RLConfiguration=Reference2026.0671.67
RAGEN v1 FilteringConfiguration=Algorithmic Baseline2026.0670.83
AutoResearchConfiguration=Reference2026.0670.42
GSPO + Clip-HigherConfiguration=Algorithmic Baseline2026.0670
GRPO + Clip-HigherConfiguration=Algorithmic Baseline2026.0669.58
Group-Variance FilteringConfiguration=Algorithmic Baseline2026.0668.75
Base model (no RL)RL=None2026.0667.5