Repository navigation
Parse ARC answer letters from the raw reply, not out.upper() - #6
Merged
Merged
Conversation
mc_generative searched out.upper() for any standalone A to E or 1 to 4, so the article in "It's a B" or "I think a good guess" was read as the letter A. The parser now lives in task.parse_mc_letter, reads the raw text case-sensitively, and only accepts a label that opens the reply (B, B. Paris, (C), Answer: D) or follows "answer is". Anything else counts as unparsed. Parametrized test added; the old regex returns A on three of its cases. The raw ARC replies were never saved, so the published 88.7% generative accuracy cannot be rescored without rerunning the model; notes/METHODS.md now says so. Co-authored-by: aghasalim <75713741+aghasalim@users.noreply.github.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
mc_generative searched out.upper() for any standalone A to E or 1 to 4, so the article in "It's a B" or "I think a good guess" was read as the letter A. The parser now lives in task.parse_mc_letter, reads the raw text case-sensitively, and only accepts a label that opens the reply (B, B. Paris, (C), Answer: D) or follows "answer is". Anything else counts as unparsed. Parametrized test added; the old regex returns A on three of its cases. The raw ARC replies were never saved, so the published 88.7% generative accuracy cannot be rescored without rerunning the model; notes/METHODS.md now says so.
Co-authored-by: aghasalim 75713741+aghasalim@users.noreply.github.com