danirodriguezz/nil-ojeda-dataset-dpo

这是一个用于偏好学习或对齐任务的数据集，包含423个训练示例和23个测试示例。每个示例由三个文本字段组成：prompt（提示）、rejected（被拒绝的回答）和chosen（被选择的回答），适用于训练模型区分更好和更差的响应。This is a dataset for preference learning or alignment tasks, containing 423 training examples and 23 test examples. Each example consists of three text fields: prompt, rejected (the dispreferred response), and chosen (the preferred response), suitable for training models to distinguish between better and worse responses.