Dense-arm SSI swap runaway cap at 0.12 × fresh national (#447)#452
Merged
Conversation
…ation) Fixes populace#447. The #431 cap (0.10 x fresh national) was sized on sparse behavior: measured sparse deltas run 7.1-8.0% across attempts 9-16. The dense arm's #448-instrumented trajectory (af02c91, full pool) converges monotonically 11.97% -> 11.64% -> 11.38% with decelerating decrements (asymptote ~10.6%) - not a runaway, but structurally above 0.10 on the same #424-undercounted candidate universe, and unreachable by more passes alone. Per the maintainer's adjudication: the dense arm gates at 0.12 x fresh national; sparse keeps 0.10. The delta and the ratio used are recorded in the reconciliation payload, the pass trajectory rides both the success record and the terminal raise (#448), and both caps converge back to one number when #424 restores the candidate universe. Tests: the fail-closed reconcile case recalibrated above BOTH ratios (+1.1M ~ 14.9%, so the dense ratio cannot flip it); a new delta-level test pins the measured dense equilibrium class (11.42%) as in-bound on 0.12 and out-of-bound on 0.10, with >12% still refused on dense. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
MaxGhenis
added a commit
that referenced
this pull request
Jul 18, 2026
…455) The #452 dense cap keyed on dense_default_dataset - but that is an EXPORT-MODE flag the sparse rmloss100 script also passes, so sparse runs inherited the dense 0.12 bar: the 32605c9 sparse rerun gated its swap delta (710,076 = 10.9%) at 0.12 and exited reconciliation at pass 1, where the sparse 0.10 bar would have forced further passes (the certified attempt-16 run converged to 7.1% by pass 2). Caught by the recorded national_swap_sanity_cap_ratio in the rerun's manifest before anything shipped; the certified default (98bf731) predates #452 entirely and the dense arm's behavior is unchanged under either discriminator. The arm identity is the frozen-selection identity: reconcile now takes sparse_selection_arm (main passes selection_source_manifest presence) and keys the ratio on that. Constants document the hazard so the export-mode flag is never reused as an arm discriminator. Tests: the fresh-pair reconcile test is parametrized over both arms - the sparse-arm-in-dense-export-mode leg is the leak regression and fails without this fix; the dense-arm leg pins 0.12. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
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.
Fixes #447
Implements the maintainer-ratified adjudication (trajectory + rationale in the issue): the dense arm's SSI-reconcile runaway cap moves to 0.12 × fresh national; sparse keeps 0.10.
national_swap_sanity_cap_ratioalongside the delta, so every manifest states which bar it was gated on.Tests: fail-closed case recalibrated above both ratios (14.9%) so the dense ratio can't flip it; new delta-level test pins 11.42% as in-bound on dense / out-of-bound on sparse, with >12% still refused on dense. Full builder suite green.
🤖 Generated with Claude Code