List your company
ResearchFormerly WizWand

Math reasoning on MATH-500 (in-distribution)

62.4Pass@1 AccuracyG2D
45.13649.61854.158.582May 20, 2026
Updated 5mo ago

Evaluation results

MethodLinks
G2DModel=Qwen2.5-7B, K (warm-up steps)=150, GPU-hrs=3.8, Decoding strategy=greedy decoding2026.0562.4—
G2DModel=Qwen2.5-7B, K (warm-up steps)=300, GPU-hrs=7.0, Decoding strategy=greedy decoding2026.0559.2—
G2DModel=Qwen2.5-7B, K (warm-up steps)=500, GPU-hrs=10.4, Decoding strategy=greedy decoding2026.0558.5—
G2DModel=Qwen2.5-7B, K (warm-up steps)=1000, GPU-hrs=20.8, Decoding strategy=greedy decoding2026.0557.6—
G2DModel=Qwen2.5-7B, K (warm-up steps)=700, GPU-hrs=14.5, Decoding strategy=greedy decoding2026.0556.4—
DPOModel=Qwen2.5-7B, K (warm-up steps)=0, GPU-hrs=2.2, Decoding strategy=greedy decoding2026.0556.2—
GRPO (G = 4)Model=Qwen2.5-7B, K (warm-up steps)=1000, GPU-hrs=52.2, Decoding strategy=greedy decoding2026.0553.2—
GRPO (G = 2)Model=Qwen2.5-7B, K (warm-up steps)=1000, GPU-hrs=26.6, Decoding strategy=greedy decoding2026.0551.6—
G2DModel=Llama-3.1-8B, K (warm-up steps)=500, GPU-hrs=11.4, Decoding strategy=greedy decoding2026.0549.4—
G2DModel=Llama-3.1-8B, K (warm-up steps)=700, GPU-hrs=16.8, Decoding strategy=greedy decoding2026.0549.2—
SFTModel=Qwen2.5-7B, Decoding strategy=greedy decoding2026.0549.1—
G2DModel=Llama-3.1-8B, K (warm-up steps)=1000, GPU-hrs=21.8, Decoding strategy=greedy decoding2026.0547.9—
GRPO (G = 4)Model=Llama-3.1-8B, K (warm-up steps)=1000, GPU-hrs=54.2, Decoding strategy=greedy decoding2026.0547.5—
G2DModel=Llama-3.1-8B, K (warm-up steps)=300, GPU-hrs=7.8, Decoding strategy=greedy decoding2026.0547.4—
DPOModel=Llama-3.1-8B, K (warm-up steps)=0, GPU-hrs=2.6, Decoding strategy=greedy decoding2026.0547.1—
G2DModel=Llama-3.1-8B, K (warm-up steps)=150, GPU-hrs=4.5, Decoding strategy=greedy decoding2026.0546.4—
GRPO (G = 2)Model=Llama-3.1-8B, K (warm-up steps)=1000, GPU-hrs=27.2, Decoding strategy=greedy decoding2026.0546.1—
SFTModel=Llama-3.1-8B, Decoding strategy=greedy decoding2026.0545.8—
CalibRLTraining Data=9k-sample subset of MATH dataset2026.02—70.2
GRPOTraining Data=9k-sample subset of MATH dataset2026.02—68.5
Qwen2.5-VL-7BTraining Data=9k-sample subset of MATH dataset2026.02—64.8