PJMixers-Dev/MinervaAI_Aesir-Preview-Anon-qwq-all

MinervaAI_Aesir-Preview-Anon数据集是一个对话数据集，包含了角色扮演的对话内容。数据集中的每个样本都由多轮对话组成，对话中包含了系统、用户和模型的发言。在训练模型时，数据集被处理为只保留最后一轮模型的输出，用于训练模型始终输出思考内容，而不受用户输入的影响。

The MinervaAI_Aesir-Preview-Anon dataset is a conversational dataset containing role-playing dialogue content. Each sample in the dataset consists of multiple rounds of conversation, including speech from the system, user, and model. When training the model, the dataset is processed to retain only the models output in the last round, used to train the model to always output thoughts, regardless of the users input.