Repository navigation
test(retain): judge agent-experience classification instead of counting fact types - #5525
Merged
Merged
Conversation
…ng fact types Since #2440 the extraction prompt classifies the user's requests, approvals and preferences as 'world' even when stated during an agent interaction; only the agent's own actions are 'assistant' -> 'experience'. The user-interaction test still required experience > world by count, so a correct 2/2 split (user request + approval as world, agent actions as experience) failed Core LLM tests consistently. Replace the count asserts in the code-change and user-interaction tests with the LLM judge: the agent's own actions must be 'experience', and the user's statements must be attributed to the user, never to the agent.
This was referenced Oct 14, 2026
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Core LLM tests fail consistently on
test_user_interaction_classified_as_experience: it asserted more 'experience' than 'world' facts, but since #2440 the extraction prompt (correctly) classifies the user's request and approval as 'world', so a right answer (2 user facts world, 2 agent actions experience) fails.Rewrites the two count-based tests in
test_fact_extraction_agent_experience.pyto useassert_meets_criteriawith a[fact_type] factsummary, per CLAUDE.md:No production prompt change. Ran locally against Gemini twice: 4/4 pass.