List your company

Mathematical Reasoning on AIME 2025 (pass@1, pass@64)

19.6Pass@1CurveRL
2.446.89511.3515.805May 23, 2026

Evaluation results

MethodLinks
CurveRLBackbone=Qwen3-4B-Base2026.0519.655.6
GRPOBackbone=Qwen3-4B-Base2026.0519.148.6
MaxRLBackbone=Qwen3-4B-Base2026.0518.952.6
CurveRLBackbone=Qwen3-1.7B-Base2026.057.740.9
BaseBackbone=Qwen3-4B-Base2026.056.944.3
MaxRLBackbone=Qwen3-1.7B-Base2026.056.736.5
GRPOBackbone=Qwen3-1.7B-Base2026.054.829.3
BaseBackbone=Qwen3-1.7B-Base2026.053.134.6