The Stanford EDGAR Filings Dataset: Reconstructing U.S. Corporate and Financial Disclosures into Layout-Faithful and Token-Efficient Pretraining Data

2026-06-16

AI Summary

High-quality, long-context public financial training data remain scarce, with existing corpora often being costly, synthetically generated, or narrowly scoped. This work proposes SEFD-v1, the first open-source, large-scale corpus of SEC disclosure filings that preserves document layout fidelity. Leveraging document parsing and structured reconstruction techniques, EDGAR filings in PDF and HTML formats are converted into layout-preserving MultiMarkdown. After rigorous deduplication and cleaning, the resulting core corpus comprises 152 billion tokens (approximately 550 billion tokens across 18.5 million documents total) and exhibits less than 0.1% overlap with Common Crawl. Additionally, two new benchmarks—EDGAR-Forecast for numerical forecasting and EDGAR-OCR for table recognition—are introduced to significantly enhance pretraining and document understanding capabilities of financial language models.

Abstract

As high-quality public web corpora become increasingly exhausted, clean long-context documents have become a scarce and expensive source of training data for large language models (LLMs). Existing long-context corpora are often proprietary and costly to acquire, synthetically generated, or concentrated in narrow domains such as programming. We introduce the Stanford EDGAR Filings Dataset (SEFD), an open reconstruction of SEC filings into layout-faithful MultiMarkdown for financial language modeling and evaluation. SEFD makes audited financial statements, risk disclosures, ownership reports, accounting notes, and market-moving event filings usable as long-context pretraining data and as a basis for financial reasoning, forecasting, compliance, and document understanding. The resulting corpus is token-efficient, model-ready, and has less than 0.1% overlap with Common Crawl-derived corpora. We release SEFD-v1, a 152B-token initial public snapshot, and provide corpus-level analyses of a larger 18.5M-filing archive estimated at 550B tokens. We further introduce two SEFD-derived benchmarks: EDGAR-Forecast, which evaluates filing-grounded numerical forecasting after model knowledge cutoffs, and EDGAR-OCR, which evaluates transcription of complex financial tables.

Problem

• long-context pretraining data
• financial language modeling
• SEC filings
• layout-faithful documents
• token-efficient corpus

Innovation

• layout-faithful
• token-efficient
• financial language modeling
• long-context pretraining
• SEC filings