Research paperCorrectedARR-2026-53CTRKDSP685PT51 · v4 · 2026-08-30

The Schwarzian Bridge in a Single Sigmoid Neuron: Oracle Gram Resolution and Spherical Saturation Phases

Lluis Eriksson

Abstract

A negative-Schwarzian sigmoid is not matrix-monotone of order two, and the same local invariant determines the quadratic onset of radial-tangential anisotropy in a single neuron. For Gaussian inputs and every h_p=sigma'^p, p>0, we prove closed saturation laws and sharp pointwise bilateral oracle-Gram complexity Theta_p(r[d+log(1/delta)]/epsilon^2), now with an autonomous Gaussian-product lower tail and explicit constants. For the weaker bottom-eigenspace target at fixed d and n, the extreme-saturation projector converges to the normal of the d-1 samples nearest the transition hyperplane; its exact angle law gives matching Theta(d^(3/2)/epsilon) complexity at fixed nondegenerate confidence. For spherical X=RU with f_R(rho) asymptotic to c rho^(a-1) at zero, an autonomous Abelian kernel lemma yields exact leading constants in three radial phases: r^(-(a+2)) for 0<a<1, r^(-3) log r for a=1, and r^(-3) for a>1. Deterministic replays, hostile proof, novelty, and reproducibility audits, and exact source provenance accompany the paper.

Deposit recordedFirst repository commit · full Git SHA recorded
PublishedGitHub release · tag ARR-2026-53CTRKDSP685PT51-v4
Canonical PDF downloads9GitHub release downloads only; direct ARR PDF reads are not measured
Page viewsNot measuredPage views are not measured until ARR connects a privacy-reviewed, no-cookie analytics source.
Definitions and rankings →
Not yet rated

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. ARR verification and screening are not peer review.

Revision statement

Major revision. Adds spectral lexicography with an exact limiting angle law, exact three-phase spherical constants, explicit confidence lower-tail constants, and new deterministic replays.

Version history

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

  • v5 · major revision · source snapshot available
  • v4 · major revision · source snapshot available · viewing
  • v3 · major revision · source snapshot available
  • v2 · major revision · source snapshot available
  • v1 · source snapshot available

AI assistance statement

OpenAI Codex assisted with public-repository and primary-literature triage, theorem exploration, proof development, adversarial proof/algebra/priority audits, numerical replay, drafting, LaTeX production, visual QA, and deposit preparation. The retained AI audits are not human peer review, formal verification, independent reproduction, or a guarantee of priority; the author remains responsible for every claim.

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

    Independent model assessments

    Read the scale and limits
    Not yet rated

    No eligible independent ARR-ASSESS-1.0 report is published for this exact version. Missing evidence is not scored as zero.

    No model reports are published for this version.

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

    Editorial disclosure

    Version v4 supersedes v3. Declared major revision: Adds spectral lexicography with an exact limiting angle law, exact three-phase spherical constants, explicit confidence lower-tail constants, and new deterministic replays. All assessments apply only when explicitly rerun and recorded for this exact version.