List your company

Math Reasoning on Minerva (Acc@1)

36.76Acc@1DAPO
21.846425.718229.5933.4618Mar 2, 2026

Evaluation results

MethodLinks
DAPOBackbone=Qwen3-8B-Base2026.0336.76
NL GuidanceBackbone=Qwen3-8B-Base2026.0336.39
Ctrl-RBackbone=Qwen3-8B-Base, beta=0.22026.0336.39
Reward ShapingBackbone=Qwen3-8B-Base2026.0334.92
Ctrl-RBackbone=Qwen3-1.7B-Base, beta=1, Importance Sampling (IS)=true2026.0324.26
NL GuidanceBackbone=Qwen3-1.7B-Base2026.0323.52
DAPOBackbone=Qwen3-1.7B-Base2026.0323.16
Reward ShapingBackbone=Qwen3-1.7B-Base2026.0323.16
Ctrl-RBackbone=Qwen3-1.7B-Base, beta=0.22026.0323.16
Ctrl-RBackbone=Qwen3-1.7B-Base, beta=0, Importance Sampling (IS)=false2026.0322.42