Conversation
The LLM judge was penalizing answers that included additional correct context beyond the reference answer, treating them as 'unverified specifics'. This caused Lore to score lower when it correctly recalled MORE than the reference answer contained. Added guidance: reference answer is a MINIMUM, extra correct details should not be penalized, only contradictions or fabrications. PR-2 at 400K: 4.68 → 4.80 (route handler: 3.0 → 4.0)
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
The LLM-as-judge was penalizing Lore answers that included additional correct context beyond the reference answer, treating extra information as 'unverified specifics'.
Example
Route handler error handling question:
The extra information was correct (from other preferences in the same scenario) but the judge treated it as potentially fabricated.
Fix
Added guidance to the judge system prompt:
Results
PR-2 at 400K tokens (Lore only):
Files Changed
packages/core/eval/judge.ts— judge system prompt