List your company

Mathematical Reasoning on Countdown (CD4/CD34 metrics)

49.8CD4 ScoreDAPO
-1.47211.83925.1538.461May 31, 2026

Evaluation results

MethodLinks
DAPOBackbone=Qwen-3B, Finetuning=true, Runtime=5.6h, Rollouts=877k2026.0549.873.261.5
POPOBackbone=Qwen-3B, Finetuning=true, Runtime=3.2h, Rollouts=205k2026.0548.772.160.4
MoPPSBackbone=Qwen-3B, Finetuning=true, Runtime=3.4h, Rollouts=205k2026.0540.766.353.5
ARPOBackbone=Qwen-3B, Finetuning=true, Runtime=2.9h, Rollouts=205k2026.0539.764.852.3
GRPOBackbone=Qwen-3B, Finetuning=true, Runtime=2.8h, Rollouts=205k2026.053964.351.6
Qwen-3B2026.050.52.61.5