List your company

Mathematical Reasoning on GSM8K (Acc, Tokens, Latency, ActRatio)

92.17AccuracyOriginal
3.93421626.84158349.7489572.656317Oct 9, 2025

Evaluation results

MethodLinks
OriginalBackbone=Qwen2.5-7B Series, Ratio=-2025.1092.17298.633.831—
UCoTBackbone=Qwen2.5-7B Series, Ratio=0.92025.1091.72244.883.3210.85
PromptBackbone=Qwen2.5-7B Series, Ratio=0.92025.1091.47281.653.7410.97
TokenskipBackbone=Qwen2.5-7B Series, Ratio=0.92025.1091.131277.213.6210.93
CoDBackbone=Qwen2.5-7B Series, Ratio=0.92025.1090.981257.773.4310.89
PromptBackbone=Qwen2.5-7B Series, Ratio=0.72025.1090.441287.323.7710.99
PromptBackbone=Qwen2.5-7B Series, Ratio=0.52025.1089.692283.23.6910.98
UCoTBackbone=Qwen2.5-7B Series, Ratio=0.72025.1087.984194.632.5710.65
CoDBackbone=Qwen2.5-7B Series, Ratio=0.72025.1087.544223.022.9710.77
PromptBackbone=Llama3.1-8B Series, Ratio=0.92025.1087.32207.892.4010.98
OriginalBackbone=Llama3.1-8B Series, Ratio=-2025.1087.26212.132.441—
PromptBackbone=Llama3.1-8B Series, Ratio=0.52025.1087.16189.322.1510.89
PromptBackbone=Llama3.1-8B Series, Ratio=0.72025.1086.79205.882.3310.97
UCoTBackbone=Qwen2.5-7B Series, Ratio=0.52025.1086.555140.361.8620.48
UCoTBackbone=Llama3.1-8B Series, Ratio=0.92025.1086.32180.312.1910.85
CoDBackbone=Llama3.1-8B Series, Ratio=0.92025.1085.821197.282.2810.93
TokenskipBackbone=Llama3.1-8B Series, Ratio=0.92025.1085.791197.162.3710.94
TokenskipBackbone=Qwen2.5-7B Series, Ratio=0.72025.1085.696223.832.8710.75
CoDBackbone=Qwen2.5-7B Series, Ratio=0.52025.1084.967150.611.9910.52
UCoTBackbone=Llama3.1-8B Series, Ratio=0.72025.1084.173140.011.7310.66
CoDBackbone=Llama3.1-8B Series, Ratio=0.72025.1083.903154.851.8410.73
UCoTBackbone=Llama3.1-8B Series, Ratio=0.52025.1083.623101.821.2610.48
TokenskipBackbone=Qwen2.5-7B Series, Ratio=0.52025.1083.478157.582.0710.53
TokenskipBackbone=Llama3.1-8B Series, Ratio=0.72025.1082.574161.211.8610.76
TokenskipBackbone=Llama3.1-8B Series, Ratio=0.52025.1081.805135.761.5610.64
TruncationBackbone=Qwen2.5-7B Series, Ratio=0.92025.1079.6812272.323.4710.91
CoDBackbone=Llama3.1-8B Series, Ratio=0.52025.1073.4613110.311.2710.52
TruncationBackbone=Llama3.1-8B Series, Ratio=0.92025.1062.5524193.912.2210.91
TruncationBackbone=Qwen2.5-7B Series, Ratio=0.72025.1052.4239210.552.6810.7
TruncationBackbone=Llama3.1-8B Series, Ratio=0.72025.1027.4859148.491.7210.7
TruncationBackbone=Qwen2.5-7B Series, Ratio=0.52025.1012.3079154.682.0110.52
TruncationBackbone=Llama3.1-8B Series, Ratio=0.52025.107.3279108.191.2510.51