fix(eval): retry empty checkpoint rows instead of resuming them as done - #65
Merged
Conversation
Exhausted API retries checkpoint as empty-string predictions. Resuming treated them as completed rows, silently scoring them as all-error — the 2026-08-31 GLM-5.3-Flash run first read 15.96 DER because 140 such rows (produced by the pre-#62 400-ing payload) were resumed as 'done'; the true number was 8.57. Empty rows are now dropped at load and re-fetched, with a startup line disclosing how many.
ronaldtse
added a commit
that referenced
this pull request
Sep 1, 2026
glm-4.7-flash currently sits behind sustained 429s (code 1305); hammering it with 8 concurrent workers burns the 5-retry budget into empty rows. The knob lets overloaded models run politely; empties remain self-healing on resume (#65).
ronaldtse
added a commit
that referenced
this pull request
Sep 1, 2026
…regression is family-wide Both 5.x variants measured: Flash 8.7978 (wrong haraqat 10.05%), full 5.3 9.8971 (missing 3.75%, wrong 8.59%) — vs GLM-5.2's 2.6911 whose wrong-rate (2.64%) matches our dedicated 580M teacher (2.62%) to within 0.02pp. Flash beats the full model significantly (bootstrap CI excludes zero); both decisively behind 5.2 (CI [-18,028, -14,685] error positions, n=1,200). Convention effect 0.015pp — negligible again. 41 exhausted-retry sentinels were caught by the #65 resume guard and re-fetched; final pass has zero empties.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Encodes the measurement lesson from the GLM-5.3-Flash run (#63): exhausted API retries checkpoint as empty strings, and resume treated them as completed rows — 15.96 DER on a contaminated first pass vs the true 8.57. Empty rows are now dropped at checkpoint load and re-fetched, with a startup line disclosing the count.