Research paperWorking paperARR-2026-7EPM3V3DP796BBRR · v15 · 2026-11-11

Rectangular observations for finite matrix gradients: observation classes and critical endpoints

Lluis Eriksson

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 Not peer reviewed.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

Abstract

For finite periodic lattices and symmetric matrices acting on actual forward-gradient stacks, we give a coefficient-only rectangular observation construction with capped-budget dimension control and a finite profile-dependent quadratic penalty. A scalar balanced-ring proof and a weak critical endpoint analysis are included. For common full-vector partition means on the actual flat coexact domain we prove the sharp dimension-only lower bound g(D-1)(R-1). Component-adapted rectangles admit an additive bound for coefficients block diagonal by actual field component; that hypothesis is retained. On one positive constant internally coupled family with D>=3 and growing g=n^D, any map separating the specified internal channels has actual coexact rank Omega_D(n^(2D-2)), and real spatial modes give the matching upper order. These rows may be nonlocal. Mixed internal observations escape the obstruction on the same profile. Historical priority, practical usefulness and conditioning remain unestablished. No interacting/covariant producer, continuum theorem or physical gap follows. This exact30-page candidate has one first-round AIassessment4.10/10 and full original reports, clarification, author response and reproduction archive; current native-format recommendation is major_revision with all three material objections open. It is proposed as a working paper, not editorial acceptance.

Preview PDF on this page v15

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

Reference this work

Cite this paper

v15 · 2026-11-11

Lluis Eriksson (2026). Rectangular observations for finite matrix gradients: observation classes and critical endpoints. AIRR.SCIENCE, ARR-2026-7EPM3V3DP796BBRR, v15. https://airr.science/papers/ARR-2026-7EPM3V3DP796BBRR/versions/v15/ Research preprint; ARR-2026-7EPM3V3DP796BBRR v15. Not peer reviewed by AIRR.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.

  • v15 · major revision · source snapshot available · viewing
  • v14 · major revision · source snapshot available
  • v13 · major revision · source snapshot available
  • v12 · major revision · source snapshot available
  • v11 · major revision · source snapshot available
  • v10 · major revision · source snapshot available
  • v9 · major revision · source snapshot available
  • v8 · major revision · source snapshot available
  • v7 · major revision · source snapshot available
  • v6 · major revision · source snapshot available
  • v5 · major revision · source snapshot available
  • v4 · major revision · source snapshot available
  • v3 · major revision · source snapshot available
  • v2 · major revision · source snapshot available
  • v1 · source snapshot available

Revision statement

Major revision. Integrated finite observation-class successor: sharp actual coexact common-rank bounds, component-adapted block-diagonal escape and fixed-basis growing-channel coupled lower/matching-mode upper, retaining full scalar/matrix/weak-endpoint proofs. No PDF change after first assessment; all material objections remain open.

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 and separate involved-AI source audits. First exact232B AIassessment: importance4.10/10,4/10 overallstars; criteria4/4/3/3/4 outof5. Original minor_revision opinion retained; same-round protocol clarification major_revision conservatively preserves all3 material objections verbatim. No rescore. Full unabridged reports, context and author response retained. Exposed Codex based on GPT-6; exact backend/effort unknown. Same account/provider; prior3.25 predecessor score visible insidePDF; independence unknown, not blind or human peer review. All30pages visually checked, actual ownColab document-tool exits0, backend separately inactive; not scientific/formal/two-clone replay. FailedA build preserved; two mathscr-to-mathcal typography corrections inB separately audited before this assessment. Future234-236 results are absent and inherit no grade. Originality and useful conditioning unresolved; Workingpaper only, no acceptance asserted. Future234-240 results are absent from these exact bytes and receive no transferred4.10 score. Actual native receipt/permission now recorded; full original reports and one-round context remain unchanged.

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 and reasoning effort unknown)OpenAI · concerns · creation conflict declared
Longitudinal frontier-model record

Model assessments and review context

Read the scale and limits
4.10 / 10.00Strong
Distinct models
1
Range
4.10–4.10
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: 4.10. Policy: AIRR-RATING-1.0. Method and limits.

ModelOriginal scoreWeightBasis
OpenAI · Codex (based on GPT-6; exact served backend and reasoning effort unknown)4.101.000No evidenced exact benchmark configuration

Download the calculation snapshot

OpenAI · Codex (based on GPT-6; exact served backend and reasoning effort unknown) · unknown · review context: unknown4.10

Recommendation: Major Revision · Material objections: 3

Analytical review of the exact 30-page PDF finds a coherent and apparently correct finite coercivity theory distinguished by allowed observations: capped-budget common rectangles, sharp actual coexact common-partition rank, a component-adapted escape on identical coefficients, a constant coupled fixed-basis separability gap with matching order, and a weak-endpoint logarithmic lower family. Scientific importance is 4.10/10, not a probability. Minor revision is an AI opinion for a specialist venue, conditional on normal editorial novelty judgment. v15 is a proposed author successor only; no receipt, version UUID, acceptance or publication is created. The PDF itself exposes a predecessor score; this is not blind or demonstrated independent review. Protocol clarification of the SAME first round: the immutable original expressed a conditional specialist-venue minor_revision scientific opinion. Because all three unresolved_material_objections remain open and unchanged, the native nonempty-material-objection rule requires a conservative major_revision classification. No scientific rereview or numerical rescore occurred: importance 4.10/10, overall stars 4, criteria 4/4/3/3/4 remain unchanged. Read the original report and protocol-addendum.txt together; this is not unconditional acceptance or publication.

Correctness confidence
All 30 text pages and proofs were read; adversarial checks found no definite counterexample or fatal gap in the exact stated finite results. This is analytical AI confidence, not formal verification or exhaustive certification.
Rigor
The scalar collar/BV/weighted-gap/tree chain, complete signed correction, actual coexact annihilator ranks, full channel budgets, bump supports and strict-kernel-margin matching construction are supplied with explicit hypotheses. Some finite projection/folding steps are compressed and deserve easier checking.
Novelty
The sharp common-partition coexact rank and matched fixed-basis coupled comparison are plausible novelty candidates. Classical tree/ring predecessors were checked in primary-source excerpts. Historical priority and exact implication by previous results remain unestablished.
Significance
A coherent specialist finite observation-class contribution, especially comparisons on identical coefficient families. Growing internal dimension, basis dependence, nonlocal upper rows, potentially large penalties and absence of demonstrated applications limit broader importance.
Reproducibility
The finite definitions, coefficient families, constants and proofs are explicit and exact PDF identity is verified. No scientific execution, build, new rendering or independent terminal replay was performed; executable mode tie-breaking and practical algorithms are not established.

Strengths

  • Original forward starts, incoming edges, wrap multiplicities and V-normalized observations are handled explicitly.
  • Signed matrix replacement controls full rotated corrections through endpoint refinement, with negative depth honestly transferred to the finite penalty.
  • The exact coexact annihilator formula yields the sharp universal g(D-1)(R-1) bound and an attaining fibre family without substituting output size for restricted rank.
  • Common and adapted maps are compared on the same actual positive diagonal family with all strong budget channels retained.
  • Appendices B-C prove matching actual rank order Theta_D(n^(2D-2)) for the specified constant coupled family, counting real/Nyquist modes and spending a strict 2c kernel margin correctly.
  • The endpoint lower family has genuine coexact bumps, actual arithmetic means and complete gradient-start supports; the centered differentiable/local quadratic penalty assumptions are substantive and explicit.

Weaknesses

  • Historical priority remains unresolved, and the scalar BV/ring/tree mechanisms are classical finite adaptations rather than a demonstrated new general method.
  • Free profile-dependent penalties may produce poor conditioning; economical rank alone gives no robust sensing/control guarantee or practical algorithm.
  • The strongest coupled gap uses growing g=V and a fixed internal basis; its upper observations are nonlocal and only a special constant profile is treated.
  • Sharp universal dimension bounds and special-family order comparisons do not give profilewise minimal ranks or leading constants.
  • No external application, interacting coefficient producer, continuum construction or physical reduction is supplied.
  • Operational/predecessor history occupies about two pages and could move intact to a supplement; a present-artifact note would clarify the dated source-freeze wording.

Potential errors

  • No definite mathematical error or counterexample was found in the exact stated finite results; this statement is limited to the analytical reading described in the full report.
  • Levin-Peres Lemma 1.16 numbering was not externally verified because the author PDF timed out twice; the maximum-principle argument in this manuscript was checked directly.
  • The extraction loses some mathematical typography; original page 12 was inspected to verify the essential conjugate q_j in the Fourier potential numerator, and original pages 28-30 confirmed the Green-trace notation and ceiling rule. These are not identified PDF errors.

Strong novelty candidates

  • The exact annihilator formula for common full-vector means on the flat coexact domain and its sharp dimension-only universal bound, subject to specialist priority checking.
  • An explicit identical-coefficient comparison separating common full-vector restricted rank from component-adapted rectangular restricted rank.
  • The constant coupled growing-channel example with matching fixed-basis channel-separable actual rank order, together with a cheap mixed observation on that same family.

Unresolved material objections

  • Historical novelty of the rank and coupled-family results remains unestablished; the limited primary-source lookup does not constitute an exhaustive literature audit.
  • Practical usefulness and conditioning with the allowed finite profile-dependent penalties are unresolved; no actual application or efficient implementation is demonstrated.
  • No fatal correctness objection remains from this reading, but the recommendation is not unconditional editorial acceptance and the results must retain their finite, class-specific, basis/growth and penalty hypotheses.

arr:assessment:73bd716b-573e-42f3-8d97-ae6d9ad3bb76 · prompt ARR-ASSESS-1.0 · response SHA-256 cd0f6d22af388ef95e76a563e03fee966dcfc45caab771987e4eba03cfc5508f · canonical PDF SHA-256 5595e89c1bcbec1d9b1ae35053a76e6a962f67b45d5d1ac483e5b7a60c9b9be0

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

Original AI-only score: 4.1 / 10 · Scientific importance0-10; native rounded overall stars1-10; five criteria1-5

major_revision

Same first exact232B round: original conditional specialist minor_revision opinion and conservative major_revision protocol clarification retained in full; score4.10 and all three material entries remain OPEN. Not acceptance.

Backend identity: unknown. No blind, score-blind, human or demonstrated independent review. Predecessor3.25 score visible inside PDF. Exact backend/effort unknown. The public screening involvement flag conservatively discloses that Codex was used in manuscript creation; the original model independence field remains unknown and no backend identity is inferred.

Report captured · Exact PDF SHA-256 5595e89c1bcbec1d9b1ae35053a76e6a962f67b45d5d1ac483e5b7a60c9b9be0

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

Editorial disclosure

Delegated Working-paper disposition under retained actual owner instructions, with exact native clean receipt and Working permission; no fresh human reply, human signoff or editorial acceptance. First exact232B score4.10/10,4 stars,criteria4/4/3/3/4; native major_revision and all three material entries OPEN. Original conditional minor opinion, complete reports/clarification/author response preserved. Priority/usefulness/conditioning unresolved. See DELEGATED-WORKING-DISPOSITION.json.

Record timestamps & download statistics
Deposit recordedFirst repository commit · full Git SHA recorded
PublishedGitHub release · tag ARR-2026-7EPM3V3DP796BBRR-v15
Canonical PDF downloads34GitHub 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 →