List your company

Math Reasoning on Minerva Math (pass@1, pass@64)

42.8pass@1MaxRL
19.9225.8631.837.74May 23, 2026

Evaluation results

MethodLinks
MaxRLModel Backbone=Qwen3-4B-Base2026.0542.862.3
GRPOModel Backbone=Qwen3-4B-Base2026.0540.858.3
CurveRLModel Backbone=Qwen3-4B-Base2026.0539.962.9
GRPOModel Backbone=Qwen3-1.7B-Base2026.0531.457.7
MaxRLModel Backbone=Qwen3-1.7B-Base2026.0531.259.5
CurveRLModel Backbone=Qwen3-1.7B-Base2026.0530.360.3
BaseModel Backbone=Qwen3-4B-Base2026.0525.961.3
BaseModel Backbone=Qwen3-1.7B-Base2026.0520.856.9