Skip to content

Research: nonvacuous coherent permutation entropy profiles - #989

Closed
amaybaum wants to merge 4 commits into
mainfrom
research/page-coherent-entropy-20261028
Closed

amaybaum wants to merge 4 commits into
mainfrom
research/page-coherent-entropy-20261028

Conversation

@amaybaum

@amaybaum amaybaum commented Oct 28, 2026 •

Copy link
Copy Markdown
Owner

Active research for #970, separate from SC-36 #978 and Introduction #918. Stacked base/D: 20ca05fe9e7e460de29e08f7a21029042fa45b12 on held draft #982. Current checkpoint: ce978ce5db4165dd1e618e9f44bb67281ac152df, tree f19e8bb1fb89bf12b732f57a3a5084c82a2cdec4.

Accepted finite mathematical core

Independent review accepts the written core at immutable bdcbd32bb55a5464d72b61682721f0b4bfaf446c / tree 700fa72b23aaa37245d136804d0ad62bb5337a21. Its proof/control/results blobs remain unchanged. The exact review, manifest and independent checker are retained under independent-core/; review SHA256 b9a8be21925cb2280fc02ccd6e70195ff21241b62ae9099745aae8dd5da71ab1.

The coherent seed is |+>^k|0>^(n-k). One uniformly random GL(n,2) configuration permutation prepares a uniform k-subspace state. All nested cuts use that same state. Written results give the flat Schmidt spectrum, exact Gaussian-binomial entropy distribution, an independent incidence formula for mean purity, a mean deficit of at most two bits below min(k,n-k,r,n-r), and a simultaneous shared-state cut bound.

Author controls pass 18,837 predicates over 2,658 state/cut cases and 35 larger moments. The independent RREF/amplitude checker passes 14,924 predicates, the same 2,658 cases and 104 larger moments. This is written finite mathematics, not a Lean certificate.

Separate sequential-release companion

Immutable companion 2624d49cffc0794b300517719c879de6b57d395e adds an autonomous clocked-SWAP release window using one fixed coherent sample and a fixed radiation register. Its clock is definite, and the finite permutation recouples later and returns after 2(n+1) steps. Controls pass 4,350 predicates over 1,592 basis configurations and 418 coherent window cases. The coherent-clock counterexample gives radiation purity 3/4. Separate independent review accepts this finite autonomous toy at 2624d49, with its files unchanged at the current head. Report SHA256: 3d3eb16f578f490e8aeb23ea9ab7fab9e2e8b4355ec31c6a6af2297be4d69bec. The independent checker passes 5,726 predicates, 1,592 basis configurations, 418 uniform-state windows and 14 complex nonuniform windows; author-result replay matches.

Checks and scope

The new path-scoped ordinary pull_request-only workflow replays author, independent and release-window exact controls from temporary copies, comparing structured JSON against retained results. Local replay-driver smoke matched all three. The dedicated research replay run 40725910910 passed on this head. Normal verify run 40725910952 is queued. Computational replay and written-proof acceptance remain separate; neither is a Lean certificate or a physical Page-curve result.

This restricted positive-amplitude subspace ensemble is not Haar measure or the full stabilizer ensemble. The mirror Bell-pair tent, repeated-CNOT return, configuration and incoherent-mixture controls retain that distinction. Coherent preparation, ensemble measure, locality, physical scrambling, energy shell, decoupling and gravitational evaporation remain separate inputs. No GR A.8–A.9 restoration, manuscript/coverage promotion, merge, native F/E designation or special dispatch is authorized. The Introduction need not wait for unrelated physical reconstruction.

Evidence: verification/audits/manuscript/page-coherent-recovery-20261028/.

Research-only written proof candidate and bounded exact controls; independent acceptance and CI outstanding. No physical Page-curve or manuscript upgrade.
Separate research-only companion; core bdcbd32 theorem/proof/control files preserved unchanged. Definite clock, finite window and eventual recoupling explicit. No physical Page-curve or status upgrade.
Core acceptance is scoped to bdcbd32's finite mathematics; companion remains separate. Automatic pull_request-only exact replay workflow, no push launch, native dispatch/designation or merge. All three replay drivers matched retained JSON in bounded local smoke.

amaybaum commented Nov 4, 2026

Copy link
Copy Markdown
Owner Author

Retired without merging (PR-AUDIT-1, #1088). The branch and its commits are kept.

Owner decision (amaybaum, Claude Code session chat, 2026-11-04):

#989 — Entropy profiles | Close unmerged | Approved after handoff to #970 and retention of the #941/NEW-3 pin obligation. Preserve branch ce978ce. Do not merge now.

The governing distinction is that closing a PR transfers responsibility; it does not establish scientific results or make branch-only evidence durable on main.

Closing this PR is administrative. It resolves no scientific question, accepts or withdraws no result, and changes no status.


Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant