Repository navigation
Prove sharp source-ordered native port error amplification - #5021
Merged
Mister-Meeseeks merged 3 commits intoOct 23, 2026
Merged
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Answers Nick Petalas's PR #4828 Open next question2 for Dugongue's actual #4743 full-input ports: the sharp Euclidean endpoint constant does not determine how initial quantization and each ordered paid-call defect reach the outputs. This package preserves the literal monomial phases and all two/three forward/reverse calls, including the third forward defect's post-call monomial.
For every positive tensor width f, let L=2^(f/2). Under explicit independent complex-coordinate defects of size rho on each completed call's actual input and exact free grid arithmetic, the sharp complete error factors are:
The all-width proof uses exact flat base-row norms and coherent row-zero overlaps preserved under individual tensor products. Independent defect phases attain the sums. Sufficient fractional accuracy requests become P+f+3, P+ceil(3f/2)+3, P+ceil(f/2)+2 and P+f+2 respectively. These are external completed-call accuracy contracts, not measured word/time savings from a new Gaussian implementation, and their abstract sharp defects need not be actual errors generated by one quantizer.
Validation: registered warning-strict CPython3.12 normal/optimized replay;12width1/2/3 cases,54exact perturbed dirty schedules and144independent impulse columns; six phase/order/cap controls;13-file source-history/committed-byte preflight; official export and Git-free cold replays; source/bool-int receipt corruption rejection. Hosted CI is unverified when opening this PR.
Gaussian internal-prefix/error guarantees and costs, physical address/workspace execution, recursive substitution and complete supplier/outer admission remain prerequisites. Paid counts, selected frontier andκ are unchanged. The package gives a scoped proof with finite source-bound corroboration and a concrete implementation next question.
Canonical record: #1166 (comment). Base
bee95ce88f9f19802f0613f92838895cd0e14a1f, priority7. Initial overlap covered310openPRs,4096Discussion records and6690base notes with reported complete pagination/inventory; arbitrary diffs and unpublished work remain outside coverage. Submission refresh is recorded separately.Credit Nick Petalas #4828, Dugongue #4743, Yiyang/Lyy-iiis #4710, lee-phys #4688, Jacob Alan Sussman #4628, eumemic, Swapnil Jain and all retained notices. Prepared by JoyboyBrian with substantial OpenAI Codex assistance, Apache-2.0; no independent human or formal-kernel acceptance claimed.
Submission refresh (2026-10-23 06:22–06:24 UTC): canonical main
f259dabc7d70e0db8d6d36bfda888f008bbfecc1was fetched; guidance, selected record and scientific input bytes remain unchanged. Pinned scientific basebee95ce88f9f19802f0613f92838895cd0e14a1fis retained; new package paths are disjoint from upstream changes. Complete remote coverage: 304 open PRs,4,114 Discussion records; local coverage: all6,690 notes at the scientific base. Coverage is advisory, not novelty proof.