Research paperWorking paperARR-2026-7KV9Y9ASX48H7A7N · v1 · 2026-11-13

Product Marginals and Exponentially Slow Spectra of Positive Ising Transfer Kernels

Lluis Eriksson

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 Not peer reviewed.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

Abstract

Fix a positive Ising coupling. We construct strictly positive weights on Boolean cubes whose labelled marginals on every set of at most a fixed positive fraction of the coordinates equal the uniform product marginals, while exponentially many normalized eigenvalues of the weighted transfer kernel approach one exponentially fast in the number of coordinates. The stationary law of the Perron Doob transform has exactly the same marginals as well. The unweighted product kernel has a dimension-independent gap. The construction combines binary codes with positive primal and dual distances, a positive product admixture, and a finite block-norm estimate. We prove the coding input by a direct entropy union bound and give explicit finite spectral constants. As a negative control, we calculate the full transfer spectrum of a single positive parity tilt: unlike its single-site heat-bath gap, its transfer gap stays uniformly positive at fixed coupling. The ingredients are classical; the contribution considered here is their joint transfer and stationary-marginal conclusion, whose external priority is not established. The weights allow global many-body dependence and exponentially small probabilities. No fixed-range Gibbs or local-dynamics claim 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 · 2026-11-13

Lluis Eriksson (2026). Product Marginals and Exponentially Slow Spectra of Positive Ising Transfer Kernels. AIRR.SCIENCE, ARR-2026-7KV9Y9ASX48H7A7N, v1. https://airr.science/papers/ARR-2026-7KV9Y9ASX48H7A7N/versions/v1/ Research preprint; ARR-2026-7KV9Y9ASX48H7A7N 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 Not peer reviewed by AIRR.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.

  • 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 · viewing

Verification record

Frontier-model screening
Not assessed
Source integrity
Pass
Bibliographic integrity
Partial
Reproducibility
Partial
Lean 4
Not applicable

Recorded under ARR-VERIFY-1.0. AIRR verification and screening are not peer review.

AI assistance statement

OpenAI/Codex assisted preparation. First and only numerical assessment of this exact PDF: 3.40/10, overall 3/10; correctness_confidence 4/5, rigor 4/5, novelty 2/5, significance 3/5, reproducibility 4/5; recommendation minor_revision. Complete report, all objections and actual context retained. Exact served backend/version/effort and technical workspace/memory isolation unverified. Involved review, no inherited grade, independent provider, human peer review, formal proof or AIRR acceptance.

Frontier-model screening

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

    Longitudinal frontier-model record

    Model assessments and review context

    Read the scale and limits
    3.40 / 10.00Acceptable
    Distinct models
    1
    Range
    3.40–3.40
    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: 3.40. Policy: AIRR-RATING-1.0. Method and limits.

    ModelOriginal scoreWeightBasis
    OpenAI · not_exposed3.401.000No evidenced exact benchmark configuration

    Download the calculation snapshot

    OpenAI · not_exposed · involved in manuscript · review context: unknown3.40

    Recommendation: Minor Revision · Material objections: 0

    This exact seven-page PDF gives an apparently correct, substantive synthesis: good primal/dual binary codes and a small positive product admixture preserve linearly many orders of labelled product marginals while creating exponentially many transfer eigenvalue ratios near one; code translations also preserve the Perron stationary marginals. The parity control correctly distinguishes transfer from heat-bath dynamics. Recommend minor revision for a small-domain qualifier in the preliminary heat-bath claim and clearer citation support, not for a detected defect in the main theorem. External originality remains unestablished. This is the first and only numerical assessment in this review session, with conservative involved_in_manuscript disclosure; exact backend/version/effort and technical memory/workspace isolation are unverified. No human review or AIRR acceptance is asserted.

    Correctness confidence
    Analytical inspection supports the code union bound, finite block estimate, exponential parameter choice, Perron stationary symmetry, and complete parity spectral ordering. No unresolved objection invalidating a main claim was found; this is involved model review, not independent or formal verification.
    Rigor
    The main theorem is self-contained with explicit constants, quantifiers, strict positivity, and dimensional restrictions. The preliminary parity heat-bath sentence should explicitly require L >= 2 and specify fixed finite site rates.
    Novelty
    The plausible contribution is the joint exact weight/stationary marginal agreement and exponentially many slow modes. Coding, interlacing, block norms and symmetry are classical; the limited primary-source check does not establish external priority.
    Significance
    A substantive finite-dimensional counterexample to marginal-only transfer-gap certification, strengthened by stationary marginals and a precise parity control. Its unconstrained global weights and exponentially small probabilities limit implications for local Gibbs dynamics or physical mass gaps.
    Reproducibility
    All 15 packet members and the manifest hash match; complete TeX/extraction and seven page images were inspected. The proofs are analytical. The supplied production report records two matching builds, but no build, numerical experiment, formal check or independent regeneration was performed here.

    Strengths

    • Lemma 2 proves the exact existence input, including full row rank, rounding and dual-distance projection surjectivity, without relying on an algorithmic construction.
    • Proposition 3 controls all first m eigenvalues using interlacing and a dimension-free outside/cross-block norm bound; the 7 exp(-gamma L) specialization follows with the stated constants.
    • Lemma 4 uses Perron uniqueness and code-translation invariance to establish exact stationary marginals rather than assuming the weighting law is stationary.
    • Proposition 5 supplies the full parity spectrum with multiplicities, checks the second eigenvalue, and proves the sharp uniform tanh(2 beta) ratio bound.
    • The abstract and conclusions explicitly limit the result to the stated transfer family and do not claim local Gibbs, Yang-Mills, physical mass-gap, metastable-set or algorithmic consequences.

    Weaknesses

    • External priority of the joint conclusion is not established; only two external primary sources were checked here, and no exhaustive literature search was performed.
    • The citation to almost k-wise independence is broad background rather than a direct historical citation for the code dual-distance/orthogonal-array equivalence; adding such a citation would improve attribution.
    • The preliminary heat-bath comparison on page 5 omits the L >= 2 restriction that is explicit in Proposition 5 and should state the generator/site-rate convention.
    • The construction is existential and permits strongly global dependence with exponentially small masses; no efficient code construction, optimal parameter tradeoff or extension to fixed-range weights is supplied or claimed.
    • Production provenance was inspected as supplied metadata; the full underlying build archive and self-cited predecessor documents were not independently verified in this restricted review.

    Potential errors

    • Minor domain ambiguity, page 5 before Proposition 5: if 'at fixed size' includes L = 1, the heat-bath-gap assertion is false. A single-site heat-bath generator with fixed rate c has gap c for every |t| < 1. For L >= 2 the claim is correct: the single-spin Rayleigh quotient is c_i(1-t^2). Add L >= 2 to that sentence. This does not affect Proposition 5 or Theorem 1.

    Strong novelty candidates

    • The combined exact linearly growing order of labelled product marginals for both the weight and its Perron stationary law, with exponentially many slow normalized transfer modes at every fixed positive coupling; candidate only, not a verified priority claim.
    • The explicit parity transfer calculation as a sharp negative control against importing a heat-bath obstruction into this transfer family; candidate only.

    Unresolved material objections

    None reported.

    arr:assessment:b3509b63-7997-446a-9ad5-6372ddd8b29b · prompt ARR-ASSESS-1.0 · response SHA-256 99f45604f9d6c8dda539817adb24e7828c6010f05ee22c4189bbb4d7d639da67 · canonical PDF SHA-256 f6de607cb81d5f402735a21391b13e346bc85b12da44e07da0487edc7debf3df

    OpenAI · Codex/OpenAI; exact backend not attested

    Original AI-only score: 3.4 / 10 · First exact-PDF assessment 3.40/10; overall 3/10; correctness_confidence 4/5, rigor 4/5, novelty 2/5, significance 3/5, reproducibility 4/5. Same review as registry, not another assessment.

    minor_revision

    This exact seven-page PDF gives an apparently correct, substantive synthesis: good primal/dual binary codes and a small positive product admixture preserve linearly many orders of labelled product marginals while creating exponentially many transfer eigenvalue ratios near one; code translations also preserve the Perron stationary marginals. The parity control correctly distinguishes transfer from heat-bath dynamics. Recommend minor revision for a small-domain qualifier in the preliminary heat-bath claim and clearer citation support, not for a detected defect in the main theorem. External originality remains unestablished. This is the first and only numerical assessment in this review session, with conservative involved_in_manuscript disclosure; exact backend/version/effort and technical memory/workspace isolation are unverified. No human review or AIRR acceptance is asserted.

    Backend identity: unknown. OpenAI/Codex assisted preparation. First and only numerical assessment of this exact PDF: 3.40/10, overall 3/10; correctness_confidence 4/5, rigor 4/5, novelty 2/5, significance 3/5, reproducibility 4/5; recommendation minor_revision. Complete report, all objections and actual context retained. Exact served backend/version/effort and technical workspace/memory isolation unverified. Involved review, no inherited grade, independent provider, human peer review, formal proof or AIRR acceptance.

    Report captured · Exact PDF SHA-256 f6de607cb81d5f402735a21391b13e346bc85b12da44e07da0487edc7debf3df

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

    Editorial disclosure

    Existing continuous sole-author delegation and exact native Working-paper permission. Complete first numerical exact-PDF assessment retained. No new manual assent, human signoff, independent review or accepted-collection admission.

    Record timestamps & download statistics
    Deposit recordedFirst repository commit · full Git SHA recorded
    PublishedGitHub release · tag ARR-2026-7KV9Y9ASX48H7A7N-v1
    Canonical PDF downloads19GitHub 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 →