---
license: cc-by-nc-4.0
pretty_name: SFD Archive 1B-Token Source-Format Sample
task_categories:
- text-generation
- question-answering
tags:
- sec
- edgar
- finance
- filings
- multimarkdown
---

# SFD Archive 1B-Token Source-Format Sample

Cleaned artifacts for an archive-wide SFD source-format analysis sample. The sanitized `filing_stats.jsonl.gz` contains 37,534 parsed filing rows and 997,469,365 final SFD tokens. The sampled manifest contains 100,000 candidate rows. `summary.json` is recomputed from the uploaded filing stats; `source_summary_checkpoint.json` preserves the original run checkpoint summary. Parser stdout tails, local paths, temporary raw SEC downloads, and process logs are omitted.