data/original gets a Test split: 8 public-domain excerpts with full provenance (ModelWritten stays out of Test)

weksa commented

2026-11-10 18:51:14 +00:00

Member

Cut cut-corpus-test-holdout.r1 (report h1), Soul verdict s1 (no blockers): all 8 excerpts verified verbatim at their Gutenberg pages, authors/translators dead >100 years, 8 pd-* speaker entities added. Test 8, Train 14, Validation 14. Caveat: public-domain text is likely memorised by models; this Test split guards against corpus leakage into the loop, not against pretraining.

🤖 Generated with Claude Code

Cut cut-corpus-test-holdout.r1 (report h1), Soul verdict s1 (no blockers): all 8 excerpts verified verbatim at their Gutenberg pages, authors/translators dead >100 years, 8 pd-* speaker entities added. Test 8, Train 14, Validation 14. Caveat: public-domain text is likely memorised by models; this Test split guards against corpus leakage into the loop, not against pretraining. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

weksa added 1 commit

2026-11-10 18:51:14 +00:00

data/original: 8 public-domain Test items with full provenance; Test held out of rounds and recall e7aef81309

Eight one-sentence English excerpts from Dhammapada (Muller), Walden, Emerson,
Bacon, Aesop (Townsend), Mill, Plato (Jowett), Darwin, each taken from the
exact Project Gutenberg page named in edition_url and checked verbatim against
the fetched text. Groups hash into Test (buckets 83-97); split.rs untouched, the
ModelWritten-never-in-Test rule stands. Items 1..28 are byte-identical (digest
pinned). Tests: counts Train/Validation 14/14, Test 8 all PublicDomain; no Test
text in any round-1 or recall prompt or RoundRecord over data/original; a
ModelWritten item in a Test group is read in no split.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

weksa merged commit 2d65239cbc into main

2026-11-10 18:51:30 +00:00

weksa referenced this pull request from a commit

2026-11-10 18:51:31 +00:00

Merge pull request 'data/original gets a Test split: 8 public-domain excerpts with full provenance (ModelWritten stays out of Test)' (#45) from weksa-lowering/corpus-test-holdout into main

Rows
Columns

data/original gets a Test split: 8 public-domain excerpts with full provenance (ModelWritten stays out of Test) #45