List your company

Mathematical Reasoning on GSM8K (Average Acceptance Length and Speedup Ratio)

6.97Average Acceptance Length (τ)PPOW
5.05645.55326.056.5468May 14, 2026

Evaluation results

MethodLinks
PPOWModel=Q3-8B, Temperature=0.02026.056.973.54
EAGLE-3Model=Q3-8B, Temperature=0.02026.056.863.47
PPOWModel=Q3-32B, Temperature=0.02026.056.624.05
PPOWModel=L31-8B, Temperature=0.02026.056.53.52
PPOWModel=L33-70B, Temperature=0.02026.056.474.54
PPOWModel=L33-70B, Temperature=1.02026.056.294.49
EAGLE-3Model=Q3-32B, Temperature=0.02026.056.213.94
EAGLE-3Model=L31-8B, Temperature=0.02026.056.123.41
GRIFFINModel=L31-8B, Temperature=0.02026.055.983.38
EAGLE-3Model=L33-70B, Temperature=1.02026.055.984.16
EAGLE-3Model=L33-70B, Temperature=0.02026.055.934.02
GRIFFINModel=L33-70B, Temperature=0.02026.055.93.99
PPOWModel=Q3-8B, Temperature=1.02026.055.763.01
PPOWModel=L31-8B, Temperature=1.02026.055.633.13
PPOWModel=Q3-32B, Temperature=1.02026.055.613.01
GRIFFINModel=L33-70B, Temperature=1.02026.055.543.87
EAGLE-3Model=Q3-8B, Temperature=1.02026.055.522.89
EAGLE-3Model=Q3-32B, Temperature=1.02026.055.232.81
EAGLE-3Model=L31-8B, Temperature=1.02026.055.162.89
GRIFFINModel=L31-8B, Temperature=1.02026.055.132.89