fix(generative): use last-match for pairwise verdict extraction - #271
Closed
AUTHENSOR wants to merge 1 commit into
Closed
fix(generative): use last-match for pairwise verdict extraction#271AUTHENSOR wants to merge 1 commit into
AUTHENSOR wants to merge 1 commit into
Conversation
process_judgement used substring 'in' checks ([[A]] before [[B]] before [[C]]/[[D]]) which matched the first verdict token anywhere in the judge response. Switch to re.findall + last match so reasoning text does not override the judge's final verdict.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
process_judgementin bothgenerative.py:493andgenerative_v2.py:175used substringinchecks (if "[[A]]" in judgmentchecked before[[B]],[[C]],[[D]]). This means the first verdict token found anywhere in the judge response wins, regardless of whether it appears in reasoning text or the final verdict.Fix
Replace the substring chain with
re.findall(r"\\[\\[([ABCD])\\]\\]", judgment.upper())[-1](last match). This binds the verdict to the judge's concluding token rather than to any token echoed in reasoning.The comment at
generative_v2.py:187notes this code was adapted from FastChat'srun_judge_pair, which had the same issue and was fixed with the same approach (lm-sys/FastChat#3921).Verification
python3 -m py_compile: compiles OKblack --check --line-length 119: clean