mlfoundations-dev/pdf_exp_train_data__pdf_sci_qa_unverified_filtered_popular_urls

该数据集包含对话信息，每个对话有from和价值两个字段，from字段代表消息的发送者，value字段代表消息的内容。数据集仅包含训练集部分，共有116457条对话记录，数据集大小为2.5GB。

The dataset includes conversation information, with each conversation having from and value fields. The from field represents the sender of the message, and the value field represents the content of the message. The dataset contains only the training set, with a total of 116,457 conversation records, and the dataset size is 2.5GB.