List your company

Mathematical Reasoning on GSM8K Qwen2.5-Instruct (test)

93.7AccuracyASCoT
74.14879.22484.389.376Aug 7, 2025

Evaluation results

MethodLinks
ASCoTScale=14B, Ratio=0.72025.0893.7221.170.71
ASCoTScale=14B, Ratio=0.92025.0893.6269.760.86
ASCoTScale=14B, Ratio=0.82025.0893.5246.570.79
ASCoTScale=14B, Ratio=1.02025.0893.3312.30.99
OriginalScale=14B2025.0893.1313.11—
ASCoTScale=14B, Ratio=0.62025.0892.8177.40.57
ASCoTScale=7B, Ratio=1.02025.0891.8294.740.99
ASCoTScale=14B, Ratio=0.52025.0891.6158.210.51
OriginalScale=7B2025.0891.4297.83—
ASCoTScale=7B, Ratio=0.92025.0891.3251.710.85
ASCoTScale=7B, Ratio=0.82025.0890.4234.950.79
ASCoTScale=7B, Ratio=0.72025.0890.3216.730.73
ASCoTScale=7B, Ratio=0.62025.0888.2178.280.6
ASCoTScale=7B, Ratio=0.52025.0886.3150.980.51
OriginalScale=3B2025.0883.7314.87—
ASCoTScale=3B, Ratio=1.02025.0883.5318.321.01
ASCoTScale=3B, Ratio=0.92025.0883.2262.310.83
ASCoTScale=3B, Ratio=0.82025.0882250.750.79
ASCoTScale=3B, Ratio=0.72025.0880.1233.150.74
ASCoTScale=3B, Ratio=0.62025.0877.6199.440.63
ASCoTScale=3B, Ratio=0.52025.0874.9170.870.54