Research paperWorking paperARR-2026-5XANPRA7E891Q9FZ · v1 · 2026-11-11

Canonical rectangle observations for matrix gradients: fixed weak spaces and checkerboard counterexamples

Lluis Eriksson

Citable version that has not completed AIRR screening. Not peer reviewed. Screening record

Abstract

We study finite positive matrix-weighted gradients on periodic lattices, with observations restricted to normalized rectangular means in a fixed internal basis. For a spatially constant coherent rank-q weak projector, the minimum actual restoring rank on the flat coexact domain has order q*n^D+g*n^(D-2), uniformly in q. A coefficient-only original-channel rectangle construction attains this order; its least quadratic coefficient has order c*min(g/q,n^2). The proof uses a stated Marcus-Spielman-Srivastava theorem augmented to control Gram matrices and cardinalities, with derived finite Neumann bounds. Spatial constancy is substantive: an exact positive two-channel checkerboard with identical local spectra has full output counts O(n log n) in dimension3 and O(n^2) in dimension4, with finite size-dependent penalties. A plane construction works at every fixed positive angle. Exact historical priority, practical usefulness, efficient implementation and conditioning remain unestablished. The first involved-AI assessment of this exact20-page PDF is4.50/10, overall5/10stars, recommendation major_revision, with all3 material objections retained. Proposed as a Working paper, not peer review or editorial acceptance. No interacting producer, continuum theorem or physical mass-gap conclusion follows.

Preview PDF on this page v1

Uses your browser's PDF viewer. Open the PDF directly if a preview is unavailable.

Reference this work

Cite this paper

v1 · 2026-11-11

Lluis Eriksson (2026). Canonical rectangle observations for matrix gradients: fixed weak spaces and checkerboard counterexamples. AIRR.SCIENCE, ARR-2026-5XANPRA7E891Q9FZ, v1. https://airr.science/papers/ARR-2026-5XANPRA7E891Q9FZ/versions/v1/ Research preprint; ARR-2026-5XANPRA7E891Q9FZ v1. Not peer reviewed by AIRR. Working paper; not admitted to the AIRR accepted collection.

Use this version link to cite the manuscript you read. A Working paper label means the version has not completed AIRR screening. AIRR screening is not journal peer review. Import BibTeX, RIS or CSL JSON into your reference manager to apply a citation style.

Version history

The paper identifier remains stable. Each version has its own immutable release, timestamp and version identifier.

  • v1 · source snapshot available · viewing

Verification record

Frontier-model screening
Did not pass
Source integrity
Pass
Bibliographic integrity
Partial
Reproducibility
Partial
Lean 4
Not assessed

Recorded under ERIKSSON-EXACT-ARTIFACT-20261031. AIRR verification and screening are not peer review.

AI assistance statement

Codex/OpenAI assisted drafting/consolidation and involved analytical source audits. First exact250B PDF AIreview: importance4.50/10,5/10overallstars; criteria correctness4,rigor4,novelty3,significance3,reproducibility4 outof5. Original recommendation major_revision; all3material objections remain open. Full original report, prompt, rawartifact binding, model/context limits and author response retained. Exposed Codex based on GPT-6; exact servedbackend/version/effort unknown. Shared account/context and prior involvement; not independent, blind or human peer review. All20pages individually inspected. Same-source freshColab clones produced byteequalPDFs; original wholeproduction failure preserved, followed by successful immutablePDF render-only continuation, own stop and separateinactivebackend check. This is not formal mathematical verification. Later249/251-253 absent from exactPDF and not used to repair/grade it. Previous232Bscore not inherited. Workingpaper only; no acceptance asserted.

Frontier-model screening

Status: fail. Any listed reports correspond to this exact version under ARR-SCREEN-1.0; no absent assessment is represented as a pass.

  • Codex (based on GPT-6; exact served backend/version and reasoning effort unknown)OpenAI · concerns · creation conflict declared
Longitudinal frontier-model record

Model assessments and review context

Read the scale and limits
4.50 / 10.00Very good
Distinct models
1
Range
4.50–4.50
Method
bounded capability-weighted mean · exact version only

Limited evidence · 0/1 benchmark matches. This is an evidence description, not a probability of correctness.

Only one distinct model; One provider; shared blind spots possible; Fresh blinded review context not documented for every model; Benchmark match incomplete; Adverse reports or unresolved objections require editorial disposition

Score calculation and evidence

Unweighted mean: 4.50. Policy: AIRR-RATING-1.0. Method and limits.

ModelOriginal scoreWeightBasis
OpenAI · Codex (based on GPT-6; exact served backend/version and reasoning effort unknown)4.501.000No evidenced exact benchmark configuration

Download the calculation snapshot

OpenAI · Codex (based on GPT-6; exact served backend/version and reasoning effort unknown) · unknown · review context: unknown4.50

Recommendation: Major Revision · Material objections: 3

The exact20-page PDF presents apparently correct finite theorems: sharp actual coexact rank order qV+gV/n^2 under a spatially constant coherent weak projector, a matching fixed-map coefficient order, and original-channel canonical checkerboard upper constructions showing spatial constancy is substantive. Scientific importance4.50/10 is a specialist AI opinion, not probability or physical Yang-Mills progress. No definite fatal mathematical error was found; exact priority and demonstrated practical/broader relevance still require substantive external judgment. Journal recommendation major_revision; transparent working-paper deposit is distinct from acceptance.

Correctness confidence
All20textpages and all20original pagePNGs read; direct adversarial checks of the actual coexact domain, MSS augmentation/prefix induction, normalization, single energy budget, rank/penalty traces, strong graph transport, endpoint/fallback and mesh constants found no definite fatal mathematical error. Analytical AI confidence only, not formal certification.
Rigor
Explicit finite proofs with actual original source labels, period2, restricted rank, singleton priority, zero rows, constants and finite completion. The non-elementary MSS input is stated and directly matched to its primary version. No formal proof or efficient selection algorithm is supplied.
Novelty
Precise matched q-uniform actual rank/fixed-weight penalty orders and the identical-local-spectrum observation-class comparison are plausible specialist novelty candidates. Main mechanisms are classical; exact predecessor implication and historical priority remain unresolved after limited primary checks.
Significance
A coherent finite observation-class contribution, with a substantial hypothesis boundary from spatially constant to checkerboard weak spaces. Fixed-basis/internal-separability assumptions, potentially growing internal dimension, large constants/penalties, existential selection and absent applications limit broader importance.
Reproducibility
Complete finite definitions/proofs, exact PDF/source binding and all20rendered pages inspected. Parent supplied matching-clone and actual Linux render closure evidence; I performed no independent execution or proof-assistant verification. Lexicographic existence is not a demonstrated practical implementation.

Strengths

  • The actual flat curl-adjoint domain, original forward source weighting and period-two multiplicities are retained.
  • The MSS added signed count coordinate controls both Gram and cardinality in one partition, including zero rows; prefix whitening is not circular.
  • Disjoint channel groups spend strong energy once, with correct normalized observation and counting coefficient conventions.
  • Actual-rank lower bounds and the fixed-map penalty trace bound cancel weak dimension correctly and avoid coexact invariance of arbitrary averaging.
  • Checkerboard strong cycles/graphs, point estimates with constant-mode mass, both layer norms, original singleton isolation and complete row counts are derived.
  • Small-input, threshold-equality, singleton and fixed-positive-angle cases are explicit; large finite penalties and unresolved optimality are stated honestly.

Weaknesses

  • Exact historical novelty and implication by nearby operator-adapted/coarse-space literature remain unestablished.
  • Very large constants, existential selection, nonlocal observations and potentially large penalties limit demonstrated practical relevance.
  • The lower theorem requires a spatially constant weak projector, coherence and original internal channel separability; it is not basis-free.
  • Checkerboard upper bounds do not determine minimum actual rank, logarithmic necessity, uniform penalty or general varying-angle orders.
  • No application, efficient implementation, interacting producer, continuum result or physical mass-gap implication is supplied.
  • Source/review history is lengthy; an intact supplement and present-artifact sidecar would improve navigation without editing the already assessed PDF.

Potential errors

  • No definite mathematical error or counterexample was found within the exact stated finite hypotheses; this is limited analytical reading rather than exhaustive/formal verification.
  • The quoted SOURCE-NOT-BUILT phrase onpage19 is source-stage explanatory history with the actual full source printed above, not an unresolved actual build value. A current sidecar should clarify subsequent compilation/render/assessment.
  • The manuscript MSS theorem was matched to the primary statement, but the full external MSS proof and all cited predecessor proofs were not re-audited.

Strong novelty candidates

  • Matched actual coexact restoring rank order qV+gV/n^2 uniformly in q for spatially constant coherent weak projectors under original channel-separable observations, subject to exact priority scrutiny.
  • Matched least normalized coefficient order c min(g/q,n^2) for the particular economical fixed-weight canonical map, subject to priority scrutiny.
  • Same-local-spectrum positive checkerboards with explicitly realized original canonical rectangle observations and subvolume full counts, showing the spatial-constancy hypothesis cannot simply be removed from a per-profile lower.

Unresolved material objections

  • M1: Historical priority and exact implication by predecessors remain unestablished; limited primary readings are not an originality certificate.
  • M2: Practical/wider relevance, conditioning with the admitted large penalties, applications and efficient implementation remain unestablished.
  • M3: Retain finite observation-class, basis/separability/coherence and constant-projector hypotheses and the fixed-map penalty scope; no unconditional editorial acceptance or physical/continuum conclusion follows.

arr:assessment:5e80d50e-3aed-41b7-91e9-5a1298ecf517 · prompt ARR-ASSESS-1.0 · response SHA-256 39fe0ac2a5a5098ea58e43a1a7bd53421cd29735292234a2a05adade29f3e29c · canonical PDF SHA-256 381da284637d6abcd2922ff649550a45243c2000fe16cc5efc65336f5765df2d

OpenAI · Codex (based on GPT-6; exact served backend/version and reasoning effort unknown)

Original AI-only score: 4.5 / 10 · Scientific importance0-10; overall5of10, criteria4/4/3/3/4 of5

major_revision

First exact250B PDF round, no rescore. Full original opinion plus format-only native clarification retained. All3material objections OPEN. Not acceptance.

Backend identity: unknown. Involved shared-account inherited context; predecessor score visible, not blind or demonstrated independent/human review. Exact served backend/effortunknown. Later249/251-253 excluded fromPDF/proof repair/score.

Report captured · Exact PDF SHA-256 381da284637d6abcd2922ff649550a45243c2000fe16cc5efc65336f5765df2d

A model assessment is not peer review or a correctness certificate. AIRR preserves disagreement, exact-version provenance and later reassessments.

Editorial disclosure

Exact-hash clean native receipt and actual CC-BY/CC0 Working permission verified. First involved-AI importance4.50/10,5stars,criteria4/4/3/3/4; major_revision and all3material objections OPEN. No new human assent, human signoff or acceptance; see DELEGATED-WORKING-DISPOSITION.json.

Record timestamps & download statistics
Deposit recordedFirst repository commit · full Git SHA recorded
PublishedGitHub release · tag ARR-2026-5XANPRA7E891Q9FZ-v1
Canonical PDF downloads2GitHub release downloads only; direct AIRR PDF reads are not included
Page viewsNot availablePage-view counts are not available in this public snapshot. Optional operator statistics are separate.
Definitions and rankings →