Self-GRIT/wikitext-2-raw-v1-preprocessed-1k-PI_KFI_claude-claude-3-5-sonnet-20240620

该数据集包含1000个训练样本，主要特征包括文本、消息和输出检索。文本字段为字符串类型，消息字段包含内容和角色两个子字段，均为字符串类型。输出检索字段包含更新后的段落、错误代码和结果列表，结果列表中包含解释、理想知识、关键事实信息、前缀意图、前缀重构和验证等子字段，均为字符串类型。数据集总大小为14255561字节，下载大小为3665796字节。默认配置指定了训练数据文件的路径。

The dataset includes multiple features such as text, messages, and output retrievals. The text feature is of string type, and the messages feature contains content and role, both of which are string types. The output retrievals feature is a struct that includes updated passage, error, results, etc., where results is a list containing explanations, ideal knowledge, key factual information, etc. The dataset is divided into a training set with 1000 samples, with a total size of 14255561 bytes.