Research paperWorking paperARR-2026-0MKJW2Y51J8AQ96A · v2 · 2026-11-12

Sharp Logarithmic Loss for Endpoint Observations in a Discrete Coexact Model

Lluis Eriksson

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 Not peer reviewed.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

Abstract

We determine the asymptotic order of the minimum energy invisible to a fixed canonical partition in a finite coexact vector-field model. On an even three-dimensional torus with counting norm and checkerboard source weights H=cn^2, a=c/48, the full endpoint cell observations leave a nonzero kernel whose minimum Rayleigh quotient has order c/log n. Explicit finite upper and lower bounds are proved. The upper witness is real, lies in the actual image of the forward curl adjoint, and annihilates every original cell mean. Finite penalties on these observations therefore cannot restore a size-uniform target c. The proof combines the complete restricted Fourier spectrum, an original-row range theorem, finite Green estimates, complementary singleton optimization and a frequency-band removal localized to the lowest nonzero component sector. We also give an exact conditional-cost identity for remaining moving-partition rows. The best canonical partition, including a quadratic moving cut, remains unresolved. Green logarithms and Schur minimization are classical; external priority of the precise model-specific theorem is unestablished. No physical Yang-Mills gap or continuum conclusion is asserted.

Preview PDF on this page v2

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

Reference this work

Cite this paper

v2 · 2026-11-12

Lluis Eriksson (2026). Sharp Logarithmic Loss for Endpoint Observations in a Discrete Coexact Model. AIRR.SCIENCE, ARR-2026-0MKJW2Y51J8AQ96A, v2. https://airr.science/papers/ARR-2026-0MKJW2Y51J8AQ96A/versions/v2/ Research preprint; ARR-2026-0MKJW2Y51J8AQ96A v2. Not peer reviewed by AIRR.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.

  • v2 · major revision · source snapshot available · viewing
  • v1 · source snapshot available

Revision statement

Major revision. Proves sharp Theta(c/log n) endpoint-kernel order through optimized polarization, band removal and a full real endpoint witness; retains the complete earlier finite-model development.

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

AI-assisted preparation and involved analytical source checks. First and only numerical assessment of this exact PDF: Millennium comparison3.40/10, overall3/10; correctness4,rigor4,novelty3,significance2,reproducibility4(each/5); minor_revision, no unresolved material objections. Full original findings, all limitations and supplemental author response retained. Exact served backend/version/effort and technical workspace/memory isolation are unverified. No previous grade inherited; no independent provider, human review, formal proof or AIRR acceptance claimed.

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 · unknown (exact backend model identifier not exposed)3.401.000No evidenced exact benchmark configuration

    Download the calculation snapshot

    OpenAI · unknown (exact backend model identifier not exposed) · involved in manuscript · review context: unknown3.40

    Recommendation: Minor Revision · Material objections: 0

    This exact 19-page draft gives an internally coherent two-sided order c/log n theorem for the full observation kernel of the fixed endpoint cut at H=cn^2 and a=c/48. Section 8 controls optimized complementary singleton coefficients, removes coordinate bands while retaining the low sector, and constructs a nonzero real actual coexact field annihilating every original endpoint cell. The conditional residual-cost identity correctly leaves moving cuts unresolved. No unresolved material objection was identified. Minor revisions should clarify the spaces/scalings and the enormous sufficient-size threshold. The result is an acceptable substantive finite-model contribution, without independently established external priority or broad physical implications. This involved OpenAI/Codex assessment has no exact backend attestation, no verified memory/workspace isolation and no provider independence; it is local draft evidence, not acceptance or a publication receipt.

    Correctness confidence
    The full endpoint proof chain, actual coexact spectrum, normalization, complement optimization, band removal and real full-cell witness are internally coherent on detailed analytical inspection. No material invalidating objection was found; this is not formal or independent verification.
    Rigor
    Explicit finite estimates, range-aware pseudoinverses and controlled rotation/removal justify Theorem 2.1. The dense transition among physical, singleton and sector spaces would benefit from a concise domain/scaling dictionary.
    Novelty
    The model-specific endpoint theorem and its common-polarization, complement-optimization and sector-removal combination are plausible novelty candidates. Classical Green and Schur mechanisms are acknowledged; external priority was not independently checked.
    Significance
    A substantive obstruction for one fixed partition in a prescribed finite coexact family, with extremely conservative thresholds. Optimized or quadratic moving cuts remain open; no physical Yang-Mills or continuum result follows.
    Reproducibility
    All 27 manifest-listed files match their raw hashes and sizes; all 19 page images and the full extraction were inspected, with targeted exact TeX checks. Proofs are self-contained and production records are supplied, but no rebuild, numerical experiment or formal verification was executed in this review.

    Strengths

    • Theorem 3.2 accounts for both transverse polarizations, the exceptional zero-frequency pair and real multiplicities on the actual curl-adjoint image.
    • Theorems 4.1 and 5.1 prove feasibility and exact shifted cancellation costs without assuming independent observation rows.
    • Theorem 2.1 is supported by a full-cell real witness; the proof does not substitute a large direct dual response for an optimized cancellation argument.
    • Band removal and sector localization preserve U, retain the low mode, and annihilate the long endpoint cells as well as singleton samples.
    • The manuscript openly limits its result to a fixed cut and discloses AI involvement, author/platform ownership and unestablished external priority.

    Weaknesses

    • The explicit upper-bound constants force fantastically large sufficient sizes and provide no practical small-size witness or sharp leading asymptotic constant.
    • The physical, singleton, useful and sector spaces and their n-dependent scalings are densely interwoven; a compact dictionary would materially improve auditability.
    • External literature comparisons are bounded, and this local-only review did not independently verify cited theorem numbering or establish priority.
    • The best canonical partition and the quadratic moving cut remain unresolved, and no continuum or physical model connection is established.
    • Reported duplicate production builds were inspected as records, not independently rerun; no numerical witness or formal proof was checked.

    Potential errors

    None reported.

    Strong novelty candidates

    • The sharp c/log n order for the full original endpoint observation kernel in this precise actual-coexact checkerboard model.
    • The combination of a common-polarization response floor, optimization over all complementary singleton coefficients and bounded band-removal loss in the lowest nonconstant sector to obtain a real full-cell witness.

    Unresolved material objections

    None reported.

    arr:assessment:e3edb1f7-f08e-4b63-abd4-de65121dc573 · prompt ARR-ASSESS-1.0 · response SHA-256 4f8b365c07abe79b71ff0e7e353a2050c341616add6706f9cba2d479f41a26b5 · canonical PDF SHA-256 03400cf6c9ecb9d12f7ca3ac0a881683d656e1216466402789262a7bec9cdc62

    OpenAI · Codex/OpenAI; exact backend not attested

    Original AI-only score: 3.4 / 10 · Original first exact-PDF assessment3.40/10; overall3/10; criteria4,4,3,2,4(each/5). Same assessment as the registry, not another review.

    minor_revision

    Internally coherent finite endpoint obstruction, with conservative constants, dense normalization, unestablished priority and optimized moving cuts still open. No independent/human review or acceptance.

    Backend identity: unknown. AI-assisted preparation and involved analytical source checks. First and only numerical assessment of this exact PDF: Millennium comparison3.40/10, overall3/10; correctness4,rigor4,novelty3,significance2,reproducibility4(each/5); minor_revision, no unresolved material objections. Full original findings, all limitations and supplemental author response retained. Exact served backend/version/effort and technical workspace/memory isolation are unverified. No previous grade inherited; no independent provider, human review, formal proof or AIRR acceptance claimed.

    Report captured · Exact PDF SHA-256 03400cf6c9ecb9d12f7ca3ac0a881683d656e1216466402789262a7bec9cdc62

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

    Editorial disclosure

    Exact-hash native linked-revision and Working-paper permission recorded through the existing revocable sole-author route. The first exact-PDF assessment recommends minor_revision; all requests remain visible with a supplemental author response. No new manual assent, human signoff, independent review or accepted-collection admission. Public release is a separate checked operation.

    Record timestamps & download statistics
    Deposit recordedFirst repository commit · full Git SHA recorded
    PublishedGitHub release · tag ARR-2026-0MKJW2Y51J8AQ96A-v2
    Canonical PDF downloads5GitHub 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 →