mytestdpo/llama3_sft_gsm8k_type12_math

该数据集包含文本选择相关的特征，其中有被选中的文本(chosen_txt)、被拒绝的文本(rejected_txt)、真实标签(gt)、提示文本(prompt)、被选中的标识(chosen)、被拒绝的标识(rejected)以及一个浮点数margin。数据集仅包含训练集split，共有2464个样本，数据集总大小为24161031字节。

The dataset includes text selection-related features, such as chosen text (chosen_txt), rejected text (rejected_txt), ground truth labels (gt), prompt text (prompt), chosen indicator (chosen), rejected indicator (rejected), and a floating-point margin. The dataset contains only a training set split with a total of 2,464 samples and a dataset size of 24,161,031 bytes.