Skip to content

docs(WO81): first VRR verdict #342

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 22, 2026
Merged

docs(WO81): first VRR verdict #342

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
17 changes: 16 additions & 1 deletion TODO.sota/81-vrr-metric.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,21 @@
# WO81 — VRR: valid-reading-rate, the ambiguity-aware metric

Status: PLANNED · Plane B (B3 closure requirement, WO56 structural rule 1)
Status: IN_PROGRESS — metric SHIPPED (models#340: build_lexicon /
score_predictions_vrr / score_text_vrr, word-level); first verdict
computed 2026-10-22 (run-081-vrr, cpu job; tools/vrr_lexicon_verdict.py):

| run | student VRR | teacher VRR | words |
|---|---|---|---|
| run-006 (rung 2.0) | 0.7974 | 0.8101 | 51,808 |
| run-007 (rung 2.1) | 0.8004 | 0.8101 | 51,808 |

Lexicon: tashkeela-full train (declared source — classical register),
309,508 skeletons. Verdict artifact:
bucket Interscript/isx-training/run-081-vrr/vrr-verdict.json (release
asset rides E3). NEXT: gold-column ceiling (oracle VRR) to separate
model error from lexicon coverage; publish VRR into the models.yaml
metric blocks once the rung-2.0 DER provenance question (WO64/B9) is
resolved.

Single-gold DER systematically undercounts legitimate ambiguity: when a
skeleton has several dictionary-licensed readings, scoring against one
Expand Down
Loading