From 311e3afd8fba91fd9eaef60cbecd752c6c9863b4 Mon Sep 17 00:00:00 2001 From: Ronald Tse Date: Thu, 5 Nov 2026 14:30:57 +0800 Subject: [PATCH] docs(WO88): khm successor recommendation khm-latn-2.0 (decode evidence; fix the phonemic analogy) --- TODO.sota/88-secryst-migration.md | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/TODO.sota/88-secryst-migration.md b/TODO.sota/88-secryst-migration.md index 3a4e443..79b8c4a 100644 --- a/TODO.sota/88-secryst-migration.md +++ b/TODO.sota/88-secryst-migration.md @@ -67,8 +67,14 @@ The multi-task IPA+Latin reinforcement beats the legacy artifact by the framework-native model replaces the legacy artifact. Next: export the multi-task model → fixtures → gates → publish as the -khm-latn successor (owner: id + version — the convention is -khm-phonemic-1.0 under the MECE taxonomy, since the output is IPA). +khm-latn successor. RECOMMENDATION (2026-11-05, decode evidence per the +WO87 naming law): **khm-latn-2.0** — the like-for-like eval (24.50% vs +legacy 27.57%) was measured on the task output, so the target +tag is latn, not phonemic; major bump 1.0 -> 2.0 follows the +ara-diac-2.0 successor precedent. (The earlier khm-phonemic-1.0 +parenthetical here was analogy, the thing the naming law forbids.) +Publish runs the export -> fixtures -> gates chain first; owner +confirms the id before any release. ## Packaging defect (found 2026-11-03): tha-phonemic-1.0.zip