Add counterbalanced within-match A/B (#374) - #498
Merged
Merged
Conversation
This was referenced Sep 21, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #374.\n\nAdds �rgus-mcp within-ab: four real bots share each match, two per arm, and a second tape swaps masks 5/10 across personality slots. Candidate code paths explicitly opt in through self.ar_abarm; without a gate the two arms are identical and measure the instrument's null error.\n\nReports retain per-arm movement/economy/combat means, paired candidate-minus-control deltas, and a two-arm Bradley-Terry log-strength interval from cross-arm kills. The output states that nav and other world-level changes are invalid on this instrument.\n\nLive null validation ran counterbalanced pairs on dm4 and dm2. dm4 cross-arm kills were 2-2 (log strength 0); dm2's 2-0 observation remained inconclusive because its 95% interval crossed zero. Ordinary dm2/dm4 matrix probes remained in band.\n\nVerified with modern FTEQCC, 258 Rust tests, 87 Python tests, strict clippy, and ship/EOL/tape/nav invariants.