Repository navigation
fix(ragas): join list contexts in Faithfulness before judging - #246
Open
Yuriy H (jurayh) wants to merge 1 commit into
Open
Yuriy H (jurayh) wants to merge 1 commit into
Yuriy H (jurayh) wants to merge 1 commit into
Conversation
Faithfulness passed the context to the judge unjoined, so a list context reached the prompt as a Python list repr (['Paris is in France.', 'It is the capital.']) instead of the documents themselves. Every other ragas scorer joins list contexts with newlines, and the TypeScript scorer flattens them. Join in both the sync and async paths and add regression tests that capture the context the judge receives.
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Faithfulnessis the only ragas scorer that does not join a list context before sending it to the judge. The prompt ends up containing the Python list repr:context: ['Paris is in France.', 'It is the capital.']
statements: ['Paris is the capital of France.']
instead of the documents themselves. Every other scorer in the same file (
ContextRelevancy,ContextRecall,ContextPrecision,ContextEntityRecall) joins list contexts with newlines first, and the TypeScript scorer flattens them. The judge still gets the gist, so nothing looks broken, but it scores against a degraded prompt.This joins list contexts in both the sync and async paths, matching the sibling scorers. Regression tests capture the context the judge receives for list and string inputs. String contexts are unchanged.
Found while auditing the ragas scorers. Related to #245, which fixed the same kind of silent degradation in
AnswerCorrectness.