List your company
ResearchFormerly WizWand

Mathematical Reasoning on MATH 500 (Accuracy Statistics)

66.72Mean Accuracy (MATH 500)KTO
45.628851.104456.5862.0556Jun 8, 2026
Updated 4mo ago

Evaluation results

MethodLinks
KTOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0666.720.5866
RePO_detModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0663.682.0161.19
RePOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0663.641.362.03
RPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.06612.7657.57
TDPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0660.681.1959.2
IPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0659.921.5258.03
DPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0657.21.8354.93
KTOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0654.640.753.77
RePOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0653.21.0251.93
RePO_detModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0652.841.3151.22
IPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0651.841.4650.03
DPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0649.40.7348.49
RPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0647.961.7345.81
TDPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0646.442.0443.91