PJMixers-Dev/Subtitles-rag-questions-r1

这是一个用于rag问题的字幕数据集，特别标记为r1。数据集的使用说明是仅保留最后一轮对话内容，以便训练模型学会在接收到任何用户输入时都输出思考。该数据集适用于类似R1模型的多轮对话训练。

This is a subtitle dataset for rag questions, specifically marked as r1. The dataset is used by masking out all content except the last turn of the conversation to train the model to always output thinking regardless of the users input. The dataset is suitable for multi-round conversation training similar to the R1 model.