Dataset
kl3m-data-edgar
A copyright-clean text collection released by the ALEA Institute as part of the KL3M Data Project, intended for training language models; full documentation is available via the linked paper and GitHub repository.
Data preview
Dataset card
kl3m-data-edgar is a small subset of the ALEA Institute KL3M Data Project, containing roughly 1,000 SEC EDGAR filings released as copyright-clean training material for large language models. It is distributed in Parquet format on Hugging Face under CC-BY terms.
What this dataset contains
The split holds about 1,000 rows drawn from SEC EDGAR documents, supplied as part of a broader KL3M corpus that the ALEA Institute describes as spanning over 132 million documents across 16 sources. Each row points to a single filing stored as text on S3, and the release is intended for pre-training and mid- or post-training of language models rather than for direct financial analysis.
Structure and fields
The Parquet files expose four schema columns. An identifier string gives the S3 path to the underlying document, dataset is fixed to the literal value "edgar", and mime_type records the document format (text/markdown in the available samples). A tokens field stores a pre-tokenized int64 sequence representing the document content, so consumers do not need to retokenize the text.
How analysts use it
Because the rows are small and the identifiers reference individual filings, the resource is best suited as a lightweight sample for evaluating EDGAR-based NLP pipelines, retrieval-augmented generation over filings, or text classification experiments before scaling to the full KL3M pipeline. Investment-fund style signal extraction is plausible only as a research starting point, since the schema carries no derived financial fields and the dataset is explicitly positioned as LLM training data.
Documentation & references 1
Related datasets
Browse moreSEC EDGAR 10-K & 10-Q financial statements for 2,600+ NASDAQ companies.
Real Estate listings (2.2M+) in the US broken by State and zip code
About 1.5 million US patent claims split into training and test partitions in CSV, organized for claim-level classification and summarization work (Apache 2.0). A large, ready-split corpus for IP-claims modeling.
ALEA Patent Claim Graph Reading questions about the dependency structure of real U.S. patent claim sets: which claim a dependent claim refers back to, the chain of claims it rests on, its independent claim, which and how many claims depend on a given claim, and yes / no questions about dependency. Every answer is computed deterministically from the claim text shown in the row. This dataset is part
A component of the ALEA Institute's KL3M Data Project supplying training material cleared of copyright concerns. The dataset card is currently a placeholder, directing readers to the GitHub repo and project paper for full documentation.
The KL3M Data Project from the ALEA Institute supplies copyright-clean training material, and the dataset listed here is one component pending further documentation on its dedicated page.
Discussion
Questions or observations about this dataset? Join the conversation on X.
No comments yet. Start the conversation on X.