List your company
ResearchFormerly WizWand

Mathematical Reasoning on Minerva (Accuracy Statistics)

26.76Mean AccuracyKTO
9.256813.800918.34522.8891Jun 8, 2026
Updated 4mo ago

Evaluation results

MethodLinks
KTOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0626.761.3125.14
DPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0624.561.5222.67
TDPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0622.281.2420.75
RePOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0621.691.0420.4
RePO_detModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0621.41.3619.71
RePOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0620.510.7119.64
RePO_detModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.06200.8119
RPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0619.341.7217.21
IPOModel (Backbone)=Qwen3-4B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0617.51.7215.37
KTOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0616.41.2414.86
IPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0616.250.9215.11
DPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0615.072.0112.57
RPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.0610.810.929.66
TDPOModel (Backbone)=Qwen3-1.7B, Decoding Strategy=sampling-based inference, Temperature=0.7, Top-p=0.952026.069.931.38.31