Skip to content

WO47: fas release-candidate comparison — run-077/best wins #297

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

WO47: fas release-candidate comparison — run-077/best wins #297

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
15 changes: 15 additions & 0 deletions TODO.sota/47-fas-diac-release.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,3 +115,18 @@ pending its dual-metric re-score (run-077-metrics) to compare like
with like against run-079 (0.64% row / 0.81% corpus). Gate value is an
owner call: accept the verified ceiling and re-anchor the recipe gate,
or commission a corpus-rebuild arm before shipping.

### Release candidate selected (2026-10-18): run-077/best

Like-with-like dual re-scores on the docs split (9,288):

| checkpoint | corpus CER | row-mean CER | word acc | size class |
|---|---|---|---|---|
| **run-077/best** (byt5-small, 6ep) | **0.777%** | **0.603%** | **82.33%** | small |
| run-079/best (byt5-base, 3ep) | 0.811% | 0.639% | 81.66% | base |

The small model wins both metrics — base adds nothing here. Gate call
remains owner-gated: accept 0.603% row-mean (0.777% corpus) as the
canonical fas-diac number and re-anchor the recipe gate, or commission
the corpus-rebuild arm (recover the ~18% of rows the derivation dropped
vs the raw 528,837) before shipping.
Loading