haeunkim/dpo_flagged

该数据集包含多个文本字段，如提示(prompt)、指令(instruction)、输入(input)、选择(chosen)、拒绝(rejected)以及一个布尔字段(chosen_non_empty)。此外，还有一个解析(parse)字段。数据集分为三个训练部分：train_low、train_medium和train_high，分别具有不同的示例数量和大小。总下载大小为50MB，总数据集大小为341MB。

The dataset includes multiple text fields such as prompt, instruction, input, chosen, rejected, and a boolean field chosen_non_empty. Additionally, there is a parse field. The dataset is split into three training parts: train_low, train_medium, and train_high, each with different numbers of examples and sizes. The total download size is 50MB, and the total dataset size is 341MB.