dogtooth/helpsteer2_binarized_filtered

该数据集包含对话提示(prompt)及其唯一标识(prompt_id)，以及对于每个提示所选(chosen)和拒绝(rejected)的回应内容及其角色(role)。每个回应还伴有分数(score_chosen和score_rejected)，表示选择的回应比拒绝的回应更合适程度的评分。数据集分为训练集，其中包含2507个示例，总文件大小约为13MB。

The dataset includes conversation prompts (prompt) and their unique identifiers (prompt_id), as well as the content and role of the chosen and rejected responses for each prompt. Each response is also accompanied by scores (score_chosen and score_rejected), indicating the degree to which the chosen response is more suitable than the rejected one. The dataset is split into a training set, which contains 2507 examples, with a total file size of approximately 13MB.