Repository navigation
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #1442.
TruthfulQA MC2 stopped counting gold answers at the first gap in their indexes. With labels
[1, 0, 1]and answer probabilities[0.2, 0.3, 0.5], it returned 0.2 instead of 0.7. Reordering the same answers could change the score.Partition the MC2 log probabilities by the actual gold indexes before calling the existing scoring helper. The helper calculation and MC1 behavior are unchanged.
Eight new tests use the real task adapter. They cover contiguous, gapped, false-first and single-nonleading-gold labels, including permutations of the same answers and both MC1 outcomes. Six fail on main; all eight pass with this change.
Tests: the new tests plus base metrics and normalization tests: 19 passed, 3 skipped, 2 deselected. The skips are existing placeholders. The two deselected dynamic metric tests require the optional multilingual extra, which is not installed locally. Repository-wide
make qualityand changed-file pre-commit checks pass. This fixes order dependence for valid labeled inputs; no change in a full benchmark score was measured.Codex assisted with the fix, regression tests and review.