Skip to content

fix(examples): data-creation loop must not force-accept a rejected example#389

Merged
drewstone merged 1 commit into
mainfrom
fix/data-creation-false-accept
Jun 25, 2026
Merged

fix(examples): data-creation loop must not force-accept a rejected example#389
drewstone merged 1 commit into
mainfrom
fix/data-creation-false-accept

Conversation

@drewstone

Copy link
Copy Markdown
Contributor

The agentic-data-creation example (#388) pushed result.winner to the accepted set unconditionally. defaultSelectWinner returns the best-SCORING iteration even when NONE passed the discriminative accept rule, so on a real null the loop falsely reported the least-bad reject as 'accepted'. It hid offline (scripted fixtures always pass) — surfaced by the live two-tier run that legitimately accepted 0/3. Fix: gate the push on the winner's own decision.accept; regression test added (force-reject → 0 accepted, no false-accept). Note: typecheck:examples needs pnpm run build first (the example self-imports the built /loops subpath).

…ample

createDataCreationLoop pushed result.winner unconditionally, but defaultSelectWinner
falls back to the best-SCORING iteration even when none passed the accept rule — so on a
real null (0 discriminating examples) it falsely reported the least-bad reject as
accepted. Gate the push on the winner's own decision.accept; add a regression test
proving the null case yields zero accepted. Surfaced by the live two-tier run (0/3).

@tangletools tangletools left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

✅ Auto-approved PR — 42f9c2ef

Blanket team auto-approval is enabled for this reviewer service.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

tangletools · auto-approval · reason: blanket_auto_approve · 2026-06-25T20:57:11Z

@drewstone drewstone merged commit 61f8346 into main Jun 25, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants