From b56e589e716a9958182958ab835b49c8decab949 Mon Sep 17 00:00:00 2001
From: Ronald Tse <ronald.tse@gmail.com>
Date: Mon, 19 Oct 2026 07:02:46 +0800
Subject: [PATCH] =?UTF-8?q?docs(wO39):=20FINAL=20=E2=80=94=20stage-2=20rej?=
 =?UTF-8?q?ected=20(identity=20collapse=208.32=20DER);=20ara=20canonical?=
 =?UTF-8?q?=20unchanged?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 TODO.sota/39-ar-oracle-distill.md | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/TODO.sota/39-ar-oracle-distill.md b/TODO.sota/39-ar-oracle-distill.md
index a300c37c..38a572b5 100644
--- a/TODO.sota/39-ar-oracle-distill.md
+++ b/TODO.sota/39-ar-oracle-distill.md
@@ -77,3 +77,22 @@ timeout — resume6/8 died exactly there). Stats:
 sft stage planned it (O_* envs, both mounts) and the Runner's budget
 guard cleared it — the first dogfooded launch; hand-typed job commands
 retire here. Gate: first-alt WER < 3.5607; stretch ≤ 2.70.
+
+## FINAL (2026-10-19): stage-2 REJECTED — ara canonical unchanged
+
+oepoch-1 dual-surface eval (the training itself converged, loss 0.002;
+eval initially crashed on missing pandas — rerun eval-only):
+
+- SadeedDiac-25: Total DER **8.32** (gate: hold r7's 2.29) — identity
+  COLLAPSE
+- WikiNews-2024 multiref: RAW 9.95 WER / 9.06 DER; first-alt
+  **3.97 WER** (gate < 3.5607) — MISS
+
+The stage-1 oracle was real (0.54% relabel CER) but continue-training
+on the relabeled 290K subset regressed both surfaces — domain-narrowed
+distillation destroyed general calibration. Same family of negative as
+fas stage-2. **ara canonical: ara-diac-2.0 (shipped, DER 2.29); the
+oracle-distillation lever is now data-closed for BOTH ara and fas.**
+
+Ops scars recorded: spine deps gap (fixed, #303), eval-only rerun
+pattern (eval_oepoch1.py), eval extras belong in recipe deps.