Skip to content

WO39: stage-1 stats + spine-dogfooded stage-2 launch #302

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ronaldtse merged 1 commit into from
Oct 18, 2026
Merged

WO39: stage-1 stats + spine-dogfooded stage-2 launch #302

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions TODO.sota/39-ar-oracle-distill.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,3 +61,19 @@ silver itself.

≈ 20-28 GPU-h total (relable 10-14 + train 10-14). Under every
prior arm's budget; no new data, no new infra.

## Stage-1 COMPLETE (2026-10-18, resume9)

290,575 rows consolidated in minutes once the consolidation levenshtein
moved to the C wheel (the pure-python loop needed ~56h vs the 23h
timeout — resume6/8 died exactly there). Stats:

- arm CER vs silver: p047 0.72% · p040 0.74% · p029 0.82%
- disagreement 31.4% — the oracle's opportunity space
- **oracle CER 0.54%** (−25% vs best single arm); picks: p047 87%,
p040 11%, p029 2%

**Stage-2 LAUNCHED THROUGH THE SPINE** (job 6ad494b9): the ara recipe's
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.
Loading