List your company

Mathematical Reasoning on GSM8K (test) (Pass@1, Efficiency Metrics)

93.3Pass@1DRGRPO
84.77286.98689.291.414Feb 19, 2026

Evaluation results

MethodLinks
DRGRPOBackbone=Qwen3-4B-Instruct2026.0293.3146.521.12
DRGRPOBackbone=Qwen2.5-7B-Instruct2026.0292.6511.055.63
GRPOBackbone=Qwen2.5-7B-Instruct2026.0292.4166.976.12
WS-GRPOBackbone=Qwen3-4B-Instruct2026.0291.7114.532.06
GRPOBackbone=Qwen3-4B-Instruct2026.0290.6161.177.79
WS-GRPOBackbone=Qwen2.5-7B-Instruct2026.0285.179.262.17