Skip to content

docs(verdicts): run-009-yallamorph teacher negative on both surfaces - r7 stays canonical - #235

Merged
ronaldtse merged 1 commit into
mainfrom
docs/r9-negative-verdict
Oct 2, 2026
Merged

ronaldtse merged 1 commit into
mainfrom
docs/r9-negative-verdict

Conversation

@ronaldtse

Copy link
Copy Markdown
Contributor

Verdict for run-009-yallamorph (TODO.sota-2026/01) - the sweep-identified teacher-side lever, now measured.

NEGATIVE on both surfaces; r7 stays canonical.

  • SadeedDiac-25 windowed zero-skip DER: 2.4895 vs r7's 2.2864 (gate bar 2.389 - FAILED)
  • WikiNews-2024 multi-ref WER/DER: 17.4265/12.1093 vs r7's 17.3794/11.8273 (no OOD trade either)
  • The lineage's first negative teacher-side result.

Mechanism is data-backed: the CamelMorph paradigm corpus vocalizes 1.5-2.0x denser than benchmark text (fatha 43.0 vs 28.4, damma 11.9 vs 7.1, shadda 8.6 vs 4.3 marks per 100 letters; tanwin nearly absent) and its forms are isolated words, not running text. At 25% dose the aux stream injected convention drift rather than morphology.

The knowledge-injection template survives with a sharper edge: r6's aux worked as running text IN benchmark convention; delivery vehicle matters as much as the knowledge. Paradigm-table aux is closed - no dose tuning or convention-normalization retry recorded.

Changes: TODO 01 closed with both gates + mechanism table; RESULTS.md entry; paper.adoc residual paragraph updated from "in flight" to the measured outcome. Memory updated. GPU spend for the campaign ends here - r7 remains the canonical teacher and ara-diac-small-2.1 (4.5701) the frontier student.

@ronaldtse
ronaldtse merged commit 7285b0b into main Oct 2, 2026
11 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant