Seed42Lab/AI-paper-crawl

该数据集名为AI-paper-crawl，包含11个分割，每个分割对应一个会议。每个分割包含四个字段：index（索引号）、text（论文的纯文本内容）、year（论文的发表年份，字符串格式）、No（年份内的索引号，字符串格式）。特别提到ICLR分割可能缺少约20%-25%的论文，因为这些论文是通过在arxiv上搜索收集的，可能返回0个或多个结果。

The dataset contains 11 splits, corresponding to 11 conferences. Each split includes four fields: index (a primary key starting from 0), text (the content of the paper in plain text form, with newline characters converted to 3 spaces unless - is detected), year (the publication year of the paper as a string, such as 2018, which can be converted to an integer if needed), and No (a string-formatted index number within a year, 1-indexed. In the ECCV split, No is the index number throughout the entire split, providing only a reference of the order these papers were accessed, not the actual publication order). The ICLR split may be missing approximately 20%-25% of the papers, as it is collected by searching on arxiv, which may return 0 or more than 1 result.