{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreib7ljhljsq24hwlgyxpgehlg73xym3fau2mcyymtuivxcrmfvz44i",
"uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mn5egfdyhfv2"
},
"path": "/t/v7-2-update-pattern-f-gap-closed-hard-negatives-batch-2/176420#post_1",
"publishedAt": "2026-05-31T10:00:29.000Z",
"site": "https://discuss.huggingface.co",
"tags": [
"Objective Projection v7.2: Closing the Pattern F Gap and Extending the Hard Negatives",
"leventbulut.com"
],
"textContent": "A short follow-up on the dataset. Two things from v7.1 are now resolved.\n\nWhen I introduced Pattern F (Mundane Parallel Life, the sixth Atmosphere Contradiction sub-pattern) in v7.1, I noted there were zero pure-corpus examples of it — it existed only inside the hard negative target outputs. That gap is now closed: ten pure scenes (`pf_001`–`pf_010`, 5 TR + 5 EN) across ten emotion categories, in `pattern_F_pure_corpus_batch1.jsonl`. To be clear about scope — these scenes _apply_ the five-criterion signature; they don’t _validate_ it. Independent validation against scenes I didn’t write is still open, and I’d genuinely welcome a counterexample that breaks the typology.\n\nI’ve also added Hard Negatives Batch 2 — ten more pairs (TR+EN) extending the five violation types into five new categories: shame, determination, awe, remorse, jealousy. Same schema as Batch 1, same `load_bearing_elements` markers.\n\nTwo housekeeping notes: the Pattern F scenes carry experimental v8-alpha schema fields (flagged as such, manually annotated — `apply_rules.py` only catches atmosphere contradiction at ~9.8%, so hand-labelling was the honest option), and the DOI record in the README is now complete at the full chain plus the canonical Sₙ reference.\n\nLonger write-up here: [ Objective Projection v7.2: Closing the Pattern F Gap and Extending the Hard Negatives ]\nFull methodology archive is at **leventbulut.com**",
"title": "V7.2 update — Pattern F gap closed + Hard Negatives Batch 2"
}