billxbf/ultrafeedback-sft-iter1

该数据集包含以下字段：提示文本(prompt)、提示ID(prompt_id)、选中内容(chosen)、被拒绝内容(rejected)、消息(messages)和是否交换偏好的标记(swap_preferences)。选中内容和被拒绝内容都包含内容和角色两个子字段。消息字段也包括内容和角色两个子字段。数据集分为训练集(train)，共有20378个示例，大小为153060276字节。

The dataset includes the following fields: prompt, prompt_id, chosen, rejected, messages, and a swap_preferences marker. Both chosen and rejected contain subfields for content and role. The messages field also includes subfields for content and role. The dataset is split into a training set (train) with 20378 examples, totaling 153060276 bytes in size.