alea-institute/alea-fr-headers

ALEA Federal Register headers

Questions and JSON extraction over Federal Register document headers (2000–2023): document type, agency and sub-agency, action, CFR parts, docket number, RIN, FR citation, publication date, effective date and comment deadline, with null or 'not stated' when the header does not say.

18,046 documents, 6,543,144 tokens (alea-v2b-sw32k, the ALEA R4 tokenizer, counting each document's start and end tokens).

Part of the ALEA R4 training mix

This dataset contains exactly the documents of this family that ALEA's R4 model trained on in its stable pretraining mixture (second data stage, one pass each). Every row is byte-identical to the trained text; text_sha256 is the SHA-256 of text.

Held-out evaluation items generated by the same code are not part of this release.

Composition

task	documents	tokens
field_qa	8,036	2,299,131
json_partial	4,060	1,593,485
json_full	3,737	2,030,886
dates_qa	2,213	619,642

6,325 documents use ALEA's chat template.

How it was made

Written by deterministic code; no language model was used. Gold comes from the XML elements of the daily issue; the citation page must agree between the issue's table of contents and the page tracking. Effective dates and comment deadlines are read only from simple DATES sentences; documents with complex date language never get date questions. Headers are shown printed (with or without the running head) or as XML. Contact blocks are not shown.

Sources

• Federal Register daily XML, 2000–2023 (U.S. government works, public domain)

Verification

• Every field is taken from its element; a scripted audit over all rows checked the date fields.
• Decontamination: training rows that shared a word 13-gram with a held-out evaluation item of this or any sibling family were removed, and rows overlapping ALEA's evaluation sets were removed or logged according to ALEA's decontamination policy.
• Sampled records were reviewed blind by a separate reviewer who recomputed or re-derived the answers; defect patterns found in review were fixed in the generator, or the affected rows dropped, before the final build.

Columns

column	content
id	row id from the generator (not always unique: a few families reuse an id across variants, and a handful of exact duplicate documents were trained twice and appear twice)
text	the document exactly as trained; chat documents use ALEA's chat template, whose control tokens appear as text (`<
task	task kind
phase	stable (trained in R4's stable pretraining mixture)
generator	what wrote the row (deterministic = code only)
group_id	source passage / record group; all variants of one group share it
prompt_chars	length in characters of the request part of text (0 for plain documents; null where not recorded)
n_tokens	tokens under alea-v2b-sw32k, including the document start and end tokens
text_sha256	SHA-256 of text
meta	JSON: the generator's structured fields (source, presentation, answer form, gold values, checks)

Known limitations

• Documents whose first page does not reconcile get no citation question.
• The documents are generated and more repetitive than natural text.

License

CC-BY-4.0 for this compilation and the generated text. Public-domain and public-record source text keeps its status; Wikidata is CC0; OpenGloss and SHELF are CC-BY-4.0.

Published by the ALEA Institute.