List your company

Mathematical Reasoning on (AIME24, MATH500, GSM8k)

79.8Average AccuracyGRPO
62.74467.17271.676.028May 18, 2026

Evaluation results

MethodLinks
GRPOBackbone=Qwen3-4B2026.0579.8
CorR-POBackbone=Qwen3-4B2026.0579.8
DAPOBackbone=Qwen3-4B2026.0579.5
GSPOBackbone=Qwen3-4B2026.0578.3
Dr.GRPOBackbone=Qwen3-4B2026.0574.4
BaseBackbone=Qwen3-4B2026.0563.4