Skip to content

Adjacent coordinate refinement beyond PR93: conditional κ = 5.2790937858556e-5 - #94

Closed
maxime-fleury wants to merge 11 commits into
CrocSwap:mainfrom
maxime-fleury:research/aligned-coordinate-edge
Closed

maxime-fleury wants to merge 11 commits into
CrocSwap:mainfrom
maxime-fleury:research/aligned-coordinate-edge

Conversation

@maxime-fleury

Copy link
Copy Markdown
Contributor

Result: a small exact gain beyond #93

Conditional κ = 13197734464639/250000000000000000 = 5.2790937858556e-5, compared with #93's 5.2790906739644e-5.

This is deliberately presented as a small finite refinement, not an asymptotic breakthrough or a practical speedup.

Construction and attribution

Start from #91 at 264f202b52edc04d0c72ca9cb3138282ca68b9ad, with its original files and archives untouched. On the already selected coordinate labels apply:

  • h23: swap 5/6.
  • h25: swap 22/23 and 5/6.

The h23 5/6 and h25 22/23 choices match #93 at c42039369af319a4ccc249b11efcfedecc2cc72f; the additional disjoint h25 5/6 transposition supplies this increment. The literal XOR sequence, slots and chronology are unchanged. Frames, source triple indices, scatter rows and output targets are transported consistently with the credited PR74/PR91 relabel action.

Credits: Chafik Boukhalfa's #91 compiler, Thomas DiFiore's #74 coordinate construction, Rohan Arun's #93 and #78 refinements and #65/#67 arithmetic, and all inherited contributors, licenses and notices. Search and additive verification by Maxime Fleury with Codebuff assistance.

Verification completed locally

The new verifier:

--record followed by compare-mode replay passed. 11 controls passed, including invalid/incomplete relabels, terminal corruption, rank-mass corruption, exact baseline comparisons, source closure and next-grid rejection. git diff --cached --check passed. The added Linux workflow runs this interface on Python 3.11 and 3.14; remote CI remains pending.

python3 research/aligned-coordinate-edge/verify.py
(cd research/aligned-coordinate-edge && python3 -m unittest test_controls -v)

Dependency and limits

This branch stacks one additive commit / 16 files on the open #91 head; its diff against main includes #91's dependencies. No inherited tracked file is edited. The #93 profiles are arithmetic comparison inputs; its producer is not independently regenerated here.

The parent #91 producer is inherited, not recompiled by this verifier; the supplied physical words are fully replayed. #91's Lean companion proves its own values, not these new numbers. This increment is rational Python verification, not a new Lean proof. Analytic, all-size framed/residual compiler, routing, recovery, prime selection and fixed-tape hypotheses remain inherited. Full repository verification was not rerun locally. No global optimum or O(n sqrt(log n)) multiplication algorithm is established.

@Mister-Meeseeks

Copy link
Copy Markdown
Contributor

Thank you @maxime-fleury. This contribution is now recorded on main in the scoped reconciliation review and contributor credits.

Contribution: Adjacent-flag improvement with inherited formal artifacts.

Review and limits: Exact-head CI evidence. The older #91 Lean package is not a formal proof of this newer kappa.

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: 0c440e40c9d4a5a02ca5bd028a2ebb5ebbca55d7. Publication: b84ba67156a01f9c2cb34fe1e3cb1dcdeab00d9c.

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.

3 participants