List your company

Mathematical Reasoning on Average (AIME 2025, BeyondAIME, HMMT 02/25, HMMT 02/26, MATH-500)

26.7Pass@1CurveRL
11.93215.76619.623.434May 23, 2026

Evaluation results

MethodLinks
CurveRLBackbone=Qwen3-4B-Base2026.0526.753.4
MaxRLBackbone=Qwen3-4B-Base2026.0526.448.7
GRPOBackbone=Qwen3-4B-Base2026.0525.143.7
CurveRLBackbone=Qwen3-1.7B-Base2026.0517.139.9
MaxRLBackbone=Qwen3-1.7B-Base2026.0516.937.7
BaseBackbone=Qwen3-4B-Base2026.0516.541.4
GRPOBackbone=Qwen3-1.7B-Base2026.0516.132.6
BaseBackbone=Qwen3-1.7B-Base2026.0512.533.8