遇见数据集

FinAudit-500

收藏
Zenodo2026-07-07 更新2026-08-01 收录
官方服务:

资源简介:

To conduct a controlled benchmark without releasing proprietary engagement information (EI), we create FinAudit-500. FinAudit-500 is a controlled benchmark generated by misstatement injection into actual public financial statements. The benchmark starts with 500 public companies, representing 12 industry groups from the GICS classification, ranging from the technology, healthcare, industrials, consumer discretionary and consumer staples, energy, financials, materials, utilities, real estate, and communication services industry groups to two cross-sector conglomerate industry groups representing the largest number of companies. CFIs are captured in their native XBRL format directly from SEC EDGAR, along with PDF/HTML exhibits, and subsequently standardized using a common taxonomy to ensure that account-level information remains comparable across companies that use different reporting rules. Consistencies in XBRL tagging have been well discussed, and to prevent errors from being added to the report, the pipeline includes a taxonomy-aligning stage that attempts to link firm-specific extension tags to the nearest standard US-GAAP tags before any misstatement is injected into the report.

提供机构:
Zenodo
创建时间:
2026-07-07
二维码
社区交流群
二维码
科研交流群
商业服务