fix(compare): regression detection for non-alphabetically-first baselines#383
Merged
Conversation
…ines determineMatrixExitCode only checked pairs where the designated baseline appeared as .baseline (first in sorted order). Targets sorting before the baseline name were never checked for regression — e.g. --baseline gpt-4.1 missed regressions from gemini-3-flash-preview. Also fixes: - --candidate without --baseline now errors instead of silently entering matrix mode - --targets filtering to zero results shows available targets - --baseline excluded by --targets shows clear error - Eval YAML uses correct schema keys (id, criteria) - Move maxLabelLen computation out of pairwise loop Refs: #381
Deploying agentv with
|
| Latest commit: |
09325d0
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://3e0040ea.agentv.pages.dev |
| Branch Preview URL: | https://fix-compare-bugs.agentv.pages.dev |
Show actual matrix and pairwise output inline so users can see what to expect without running the command. Add Quick Start section pointing to the included fixture, exit code reference table, and pairwise mode example.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Fixes 5 bugs found during code review of #382:
determineMatrixExitCodemissed regressions for targets sorting alphabetically before the baseline. Pairwise pairs are generated in sorted order, so--baseline gpt-4.1only checkedgpt-4.1 → gpt-5-minibut missedgemini-3-flash-preview → gpt-4.1(where gpt-4.1 is the candidate, not the baseline in the pair). Now checks both directions.--candidatewithout--baselinesilently entered matrix mode. Now errors with a clear message.--targetsfiltering to zero results gave "No results found" with no hint about available targets. Now shows available target names.--baselinecombined with--targetsthat excludes the baseline silently ignored regression check. Now errors early.test_id,expected,scoresinstead ofid,criteria). Fixed to match agentv's actual YAML parser.maxLabelLencomputation out of the pairwise summary loop (was O(n²), now O(n)).Test plan
agentv compare combined.jsonl --baseline gpt-4.1correctly exits 1 (gemini regression detected)--candidate gpt-5-mini(no baseline) → clear error--targets nonexistent→ lists available targets--baseline gpt-4.1 --targets gemini gpt-5-mini→ baseline excluded errorRefs: #381