OpenLLMAI/preference_700K

该数据集主要用于偏好学习，包含700,000个样本。每个样本包含四个主要特征：rejected、rejected_score、chosen_score和chosen。其中，rejected和chosen是列表类型，每个列表包含content和role两个字段，数据类型均为字符串。rejected_score和chosen_score的数据类型为float64。数据集只有一个训练集（train），总大小为2,802,733,004字节，下载大小为1,433,169,364字节。

This dataset is primarily used for preference learning and contains 700,000 samples. Each sample includes four main features: rejected, rejected_score, chosen_score, and chosen. Among these, rejected and chosen are list types, each containing two fields: content and role, both of which are of string type. rejected_score and chosen_score are of float64 type. The dataset has only one training set (train), with a total size of 2,802,733,004 bytes and a download size of 1,433,169,364 bytes.