Repository navigation
Conversation
…78e-5 Under PR104's opposite-bank rule every residual is one child, so deferred readouts pay on both networks. Bit: Swapnil Jain's round-seven deferred word, unchanged (stopped saving 1.2402e-4). Complex: PR104's h=24 network with lifted binary frames and deferred readouts on 8,893 of 44,918 roles (8.58628e-5). Replay with arbitrary scratch, exact F2 frame checks, exact moments and the 47-constraint assembly give kappa = 429239/5000000000. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…kappa 9.8699350338e-5
…085428e-4 Run the PR110/PR116 deferral compiler unchanged on PR117's replayed 28,705-role h24 complex DAG (3,020 deferred roles). Complex saving 108566486/10^12 still binds below the round-seven bit word; kappa = 6783925373/(6.25*10^13), 3.875% above PR117. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011S35E2acyQDaVRjPRw4TDy
|
Thank you @rohanarun. This contribution is now recorded on main in the scoped reconciliation review and contributor credits. Contribution: Repair the 55/42 coefficient by splitting it into 1+13/42. Review and limits: Exact-head CI evidence; both scalar additions are charged. This is not a free precision improvement. The complete original source, proof/research notes, notices and relevant replay/CI evidence are retained through the pinned source archive. The review distinguishes fresh runs, exact-head historical CI and prior integration evidence. Closing this historical PR as reconciled; its obsolete headline does not replace the current selected bound, and draft/retracted claims retain their explicit qualifications. Reviewed head: |
5 participants
Audit correction: the inherited unit-coefficient check failed on 55/42. This update implements and verifies an exact split into at most two unit-magnitude shears at the same frame. The independent audit now passes, charging 11,862,944 unit shears for 11,862,912 nonzero old readout coefficients (32 extra additions). The conservative scalar guard is doubled and recertified; the complete child profile and κ are unchanged. Eight controls reject missing, oversized and incorrectly signed pieces. The original failure is preserved in the validation evidence. Combined coverage is now complete; the original failed invocation remains disclosed below.
Verification completed: Combined verification passed. The original full invocation on
ec862a5af51537495c745d9f2323e8a5736ee265passedmake -j1 verify, producer, bit and certificate commands, then failed the inherited unit-coefficient audit; its sixth planned equality command did not run. The correctedverify.py --fullpassed on6ec868e8568ca9a74b7245494de1b187c7be78fein 303.50 seconds, including exact regenerated artifacts, pins, eight controls and the paid unit-shear reflected audit. The broad sources and physical producer/DAG/profiles are byte-identical across these commits, and both runs had no source drift. All 42 corrected-head GitHub checks passed. This is combined coverage, not a monolithic full-make run on the corrected commit. Seeresearch/deferred-replayed-117/validation.json.Conditional κ = 6783925373/(6.25·10^13) = 1.085428059680×10⁻⁴, +3.875% over #117 (1044939/10^10), +9.97% over #116.
#117 found an h=24 all-disjoint complex DAG with only 28,705 roles, but charged it with #104's plain profile; its complex saving 26129/(2.5·10^8) binds. #110/#114/#116 showed deferred readouts pay on the complex network under #104's one-child rule. This PR runs the same deferral algorithm, with the producer adapter, expected graph counts and pair-star root ordering updated on #117's replayed DAG.
Bit side unchanged: Swapnil Jain's round-seven deferred word, one-child rule (same as #117 and #110/#116).
Checks:
complex_deferred.py(DAG replay via #117's unchangedreplayed.py, maximum carrier matching, lifted frames, deferral, arbitrary-scratch replay, exact F2 checks; pins 91,770/8,120/71,185/28,705) andcertificate.py(#116's exact certificate: moments, next-grid rejections, expanded scalar charge, 47 constraints, 7 margins; requires κ above #117 and #116).Credit: eumemic (#117, #114), Avi Eisenberg (#110), Swapnil Jain, icekylinx (#104/#115), #116 and all predecessors. Composition by Rohan Arun with Anthropic Claude assistance.
Independent local cross-check: a separately adapted PR116 deferral run on pinned PR117 head
cbb05ce504d571546d9b7794c186a613c659c3bfpassed and reproduced both the complete complex profile and exact certificate in this branch exactly. The corrected independent reflected-word audit passed, including exact dirty cancellation, all frame incidences and the complete child histogram. The original broad run and corrected package replay are documented in the combined validation receipt. All inherited analytic, fixed-tape, stopped-product and residual-to-child interfaces remain conditional assumptions.