List your company

Mathematical Reasoning on MATH-500 (held-out)

89.4Pass@1 ScoreHSD
77.9680.9383.986.87Jun 14, 2026

Evaluation results

MethodLinks
HSDBase Model=Qwen3-32B-Base2026.0689.4
GRPO + OPSDBase Model=Qwen3-32B-Base2026.0687.9
RLSDBase Model=Qwen3-32B-Base2026.0687.7
SDPOBase Model=Qwen3-32B-Base2026.0687.5
OPSDBase Model=Qwen3-32B-Base2026.0687
DAPOBase Model=Qwen3-32B-Base2026.0686.8
GSPOBase Model=Qwen3-32B-Base2026.0686.4
Dr. GRPOBase Model=Qwen3-32B-Base2026.0685.9
GRPOBase Model=Qwen3-32B-Base2026.0685.2
HSDBase Model=Qwen3-8B-Base2026.0683.6
GRPO + OPSDBase Model=Qwen3-8B-Base2026.0681.9
RLSDBase Model=Qwen3-8B-Base2026.0681.6
SDPOBase Model=Qwen3-8B-Base2026.0681.4
OPSDBase Model=Qwen3-8B-Base2026.0680.9
DAPOBase Model=Qwen3-8B-Base2026.0680.1
GSPOBase Model=Qwen3-8B-Base2026.0679.7
Dr. GRPOBase Model=Qwen3-8B-Base2026.0679.2
GRPOBase Model=Qwen3-8B-Base2026.0678.4