scottgeng00/olmo-3-preference-mix-deltas-yolo_og_no_multilingual

该数据集包含用户交互信息，其中包括提示文本（prompt）、用户选择的内容（chosen）和被拒绝的内容（rejected）。每个条目还包括用户的语言、国家、IP哈希值、设备头部信息等上下文信息。此外，还包含是否含有有毒内容（toxic）的标记、对话的轮次标识（turn_identifier）等信息。数据集的训练集大小为约14.06GB，共有约1698028个示例。

The dataset includes user interaction information, such as prompt text, chosen content, and rejected content. Each entry also includes contextual information such as user language, country, IP hash, device header information, etc. Additionally, it includes a tag indicating the presence of toxic content, a turn identifier for the conversation, and more. The training set of the dataset is approximately 14.06GB in size, with a total of about 1,698,028 examples.