osunlp/QUEST-Mid-Training-Data
收藏资源简介:
QUEST Mid-Training Data 是一个用于QUEST模型中期训练的数据集,以Parquet分片格式提供。目前发布了一个分支:context_summarization(上下文摘要)。每行数据包含一个messages字段,该字段是一个聊天格式的列表,其中每个元素是一个字典,包含角色(role)和内容(content)键值对,用于表示对话消息。数据集旨在支持模型在研究和摘要任务中的训练,特别是通过合成任务来训练深度研究代理。
QUEST Mid-Training Data is a dataset designed for the mid-stage training of the QUEST model, provided in Parquet shard format. Currently, one branch has been released: context_summarization (context summarization). Each data row includes a `messages` field, which is a chat-formatted list. Each entry in this list is a dictionary with `role` and `content` key-value pairs to represent individual conversation messages. This dataset aims to support model training for research and summarization tasks, particularly for training deep research agents via synthetic tasks.




