sfd-anonymous/sfd-v1
收藏资源简介：
SFD-v1是一个开放的数据集，基于美国证券交易委员会（SEC）EDGAR文件，重建为布局保真的MultiMarkdown（MMD）格式，旨在支持长上下文语言建模、金融推理、文档理解和评估任务。该数据集覆盖了2022年1月至2025年6月期间的约380万份文件，包含约1520亿个Qwen3-1.7B tokens。它处理多种源格式（如HTML、XML、纯文本等），并保留合并单元格表格、缩进和视觉层次结构，以提供高效的令牌表示。数据集包括多种文件类型（如10-K、10-Q、8-K等），并附带元数据（如SEC编号、年份、月份等），每个文件都是自包含的。
SFD-v1 is an open, layout-faithful reconstruction of U.S. Securities and Exchange Commission (SEC) EDGAR filings into token-efficient MultiMarkdown (MMD), targeted at long-context language modeling, financial reasoning, document understanding, and evaluation. This release covers filings from January 2022 through June 2025, with approximately 3.8 million filings and around 152 billion Qwen3-1.7B tokens. It handles multiple source formats (e.g., HTML, XML, plaintext) and preserves merged-cell tables, indentation, and visual hierarchy for efficient token representation. The dataset includes various filing types (such as 10-K, 10-Q, 8-K) and is accompanied by metadata (e.g., accession number, year, month), making each filing self-contained.
