Increase output token limit for EquivalenceEvaluator #6835
Merged
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
EquivalenceEvaluator was specifying MaxOutputTokens = 1 since its prompt instructs the LLM to produce a response (score) that is a single digit (between 1 and 5). Turns out that while this works for most models (including the OpenAI models that were used to test the prompt), some models require more than one token for this. For example, looks like Claude requires two tokens for this - see dotnet#6814). This PR bumps the MaxOutputTokens to 5 to address the above issue. Fixes dotnet#6814
There was a problem hiding this comment.
Pull Request Overview
Increases the maximum output token limit for EquivalenceEvaluator from 1 to 5 tokens to accommodate different LLM tokenization behaviors while maintaining the single-digit score output requirement.
- Bumps MaxOutputTokens from 1 to 5 in ChatOptions configuration
- Adds explanatory comment linking to the GitHub issue for context
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
github-actions
bot
added
the
area-ai-eval
peterwald
approved these changes
Sep 23, 2025
This was referenced Oct 14, 2025
Open
Merged
This was referenced Oct 15, 2025
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
EquivalenceEvaluator was specifying MaxOutputTokens = 1 since its prompt instructs the LLM to produce a response (score) that is a single digit (between 1 and 5).
Turns out that while this works for most models (including the OpenAI models that were used to test the prompt), some models require more than one token for this. For example, looks like Claude requires two tokens for this - see #6814).
This PR bumps the MaxOutputTokens to 5 to address the above issue.
Fixes #6814
Microsoft Reviewers: Open in CodeFlow