Pith. sign in

REVIEW 4 major objections 1 minor 3 cited by

Three distinct edge statistics govern non-Hermitian random matrices, with universal cubic repulsion at short range.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-13 16:18 UTC pith:KP4NQ35D

load-bearing objection Wrong full text was cached for 2603.28457; only the abstract is usable, so the edge-universality claims cannot be checked and this is not ready for a real referee read yet. the 4 major comments →

arxiv 2603.28457 v2 pith:KP4NQ35D submitted 2026-03-30 math-ph cond-mat.stat-mechmath.MPmath.PR

Three non-Hermitian random matrix universality classes of complex edge statistics: Spacing ratios and distributions

Gernot Akemann , Georg Angermann , Noah Ayg\"un , Adam Mielke , Patricia P\"a{\ss}ler , Christoph Raitzig , Tobias Winkler This is my paper
classification math-ph cond-mat.stat-mechmath.MPmath.PR MSC 60B2015B5282B44
keywords non-Hermitian random matricesedge statisticscomplex spacing ratiosnearest-neighbour spacing distributionsGinibre ensembleuniversality classes2D Coulomb gascubic repulsion
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Non-Hermitian random matrices are conjectured to fall into three generic local bulk statistics; that conjecture has recently been extended to the spectral edge. This paper characterises those three edge classes through complex spacing ratios and nearest-neighbour spacing distributions, using the simplest Gaussian representatives: complex Ginibre, complex symmetric, and complex self-dual matrices. Analytically it simplifies finite-N expressions for the complex spacing ratio in the Ginibre class and supplies a parameter-dependent N=3 surmise that interpolates toward the Hermitian GUE. Numerically it maps bulk versus edge repulsion for all three ensembles against a two-dimensional Poisson process, interpreting the differences via an effective two-dimensional Coulomb gas at different inverse temperatures. The work finds that complex spacing ratios do not fully unfold local edge statistics, yet verifies that short-distance nearest-neighbour spacings show the same universal cubic repulsion in bulk and edge for every class.

Core claim

There exist three generic local edge statistics among non-Hermitian random-matrix symmetry classes, faithfully represented by the complex Ginibre, complex symmetric and complex self-dual Gaussian ensembles; these classes are distinguished by their complex spacing-ratio and nearest-neighbour distributions, share universal cubic repulsion at small argument, and are only incompletely unfolded by complex spacing ratios at the edge.

What carries the argument

Complex spacing ratios together with nearest-neighbour spacing distributions, analysed via a conditional point process (analytic, class A) and an effective two-dimensional Coulomb-gas picture at different inverse temperatures β (numeric, all three classes).

Load-bearing premise

The three chosen Gaussian ensembles are faithful representatives of the three conjectured edge universality classes for all non-Hermitian symmetry classes, so that results for them extend to the generic local edge statistics.

What would settle it

Compute complex spacing-ratio histograms and nearest-neighbour spacing distributions at the edge for a non-Gaussian matrix ensemble belonging to one of the three symmetry classes (or a different representative of the same class) and check whether they match the Ginibre, complex-symmetric or complex-self-dual edge statistics reported here; a clear mismatch would refute the claimed universality.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Local edge statistics of non-Hermitian matrices can be classified by the same three symmetry classes already used for the bulk.
  • An effective 2D Coulomb-gas description at different β accounts for the observed variation in edge repulsion across the three classes.
  • Complex spacing ratios, while convenient, leave residual global density effects at the edge and therefore do not fully unfold local edge statistics.
  • Short-distance cubic repulsion of nearest-neighbour spacings holds universally in both bulk and edge for all three classes.
  • A simple N=3 elliptic-Ginibre surmise already captures the crossover of the complex spacing ratio toward the Hermitian GUE limit.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the three edge classes are truly universal, the same cubic short-distance law and Coulomb-gas β hierarchy should appear in non-Gaussian and sparse non-Hermitian ensembles used in open quantum systems and neural-network Jacobians.
  • Incomplete unfolding by complex spacing ratios at the edge suggests that edge-specific normalisations (for example local mean density or radial scaling) may be needed before ratio statistics can be compared across ensembles.
  • The analytic conditional-point-process simplification for finite-N Ginibre ratios could be extended to the other two classes to test whether their edge statistics also admit closed finite-N expressions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 1 minor

Summary. From the supplied abstract, the manuscript claims that the three conjectured non-Hermitian bulk universality classes extend to three generic local edge statistics, represented by the Gaussian complex Ginibre (A), complex-symmetric (AI†) and complex self-dual (AII†) ensembles. It reports (i) finite-N analytic results for complex spacing ratios in class A via a conditional point process, explaining why an earlier bulk approximation becomes controlled at large N, and a parameter-dependent N=3 surmise for the elliptic Ginibre ensemble that interpolates to the GUE; (ii) numerical comparisons of complex spacing ratios, their moments, and nearest-neighbour spacing distributions for all three ensembles against 2D Poisson, both in bulk and at the edge, interpreted via an effective 2D Coulomb gas at different β; (iii) indications that complex spacing ratios do not fully unfold edge statistics; and (iv) universal cubic small-argument repulsion of NN spacings in bulk and edge for all three classes. The body text supplied under this paper_id is, however, an unrelated systems paper on hierarchical sparse attention (HISA), so none of the analytic or numerical claims can be checked against derivations, figures, or tables.

Significance. If the claims hold, the work would complete the bulk-to-edge extension of the three non-Hermitian local statistics and supply practical diagnostics (complex spacing ratios, NN distributions, Coulomb-gas β picture) used across non-Hermitian RMT and open quantum systems. The announced finite-N conditional point process, the controlled large-N justification of a prior approximation, and a parameter-dependent N=3 surmise would be concrete technical contributions. Those strengths cannot be credited from the present file, because the manuscript body does not contain the RMT analysis.

major comments (4)
  1. Manuscript identity mismatch: the title, abstract and paper_id (2603.28457, math-ph) describe non-Hermitian edge statistics, but the full text is the unrelated HISA sparse-attention paper (arXiv:2603.28458, cs.LG). Every load-bearing claim in the abstract—conditional point process, N=3 elliptic surmise, bulk/edge numerics for A/AI†/AII†, Coulomb-gas β interpretation, incomplete edge unfolding, cubic repulsion—is therefore uncheckable. The correct full text must be supplied before any scientific assessment is possible.
  2. Abstract, choice of ensembles: the central universality claim rests on treating complex Ginibre, complex-symmetric and complex self-dual Gaussians as faithful representatives of the three edge classes. With no body text, there is no derivation, finite-N scaling, or comparison to other members of the same symmetry classes that would justify this representativeness. This premise remains an unchecked assumption.
  3. Abstract, analytic part (class A): the claimed simplification of finite-N complex spacing ratios via a conditional point process, and the argument that an earlier uncontrolled bulk approximation becomes controlled at large N, cannot be verified without equations, error estimates or comparison to exact formulae. Same for the parameter-dependent N=3 elliptic-Ginibre surmise and its GUE limit.
  4. Abstract, numerical part: comparisons of spacing-ratio moments and NN distributions to 2D Poisson in bulk and edge, the effective-β Coulomb-gas reading, and the “indications” of incomplete edge unfolding all require figures, sample sizes, unfolding procedure and error controls that are absent from the supplied file.
minor comments (1)
  1. Until the correct PDF is attached, presentation issues (notation for AI†/AII†, definition of the complex spacing ratio, edge vs bulk unfolding protocol) cannot be reviewed.

Circularity Check

0 steps flagged

No circularity detectable: abstract describes a standard RMT programme of ensemble definition, finite-N analytics, and numerical comparison; full RMT manuscript is not present in the supplied cache.

full rationale

The supplied CACHEABLE body is the unrelated HISA sparse-attention paper (arXiv:2603.28458), not the claimed non-Hermitian RMT work (arXiv:2603.28457). Only the RMT abstract is available. From that abstract the derivation chain is the ordinary RMT pipeline: choose the three simplest Gaussian representatives (Ginibre A, complex-symmetric AI†, complex self-dual AII†) of the conjectured edge classes; introduce a conditional point process to simplify finite-N complex spacing ratios in class A; give an elliptic-Ginibre N=3 surmise; numerically compare spacing-ratio moments and NN distributions of all three ensembles against 2D Poisson in bulk and edge; interpret repulsion via an effective 2D Coulomb gas at different β; and check small-argument cubic repulsion. None of these steps is self-definitional, a fitted input re-labelled as prediction, a load-bearing self-citation uniqueness theorem, an ansatz smuggled via self-citation, or a renaming of a known result. The three-class premise is an external conjecture the paper adopts as background, not a circular reduction. With no equations or self-citations of the RMT paper visible, no circular step can be exhibited. Score 0; steps empty.

Axiom & Free-Parameter Ledger

1 free parameters · 4 axioms · 2 invented entities

Abstract-only review of a mathematical-physics paper. Load-bearing background is the three-class bulk conjecture extended to the edge, the identification of three Gaussian ensembles as representatives, and the 2D Coulomb-gas effective description. No free parameters are fitted in the abstract; the elliptic deformation parameter is a model parameter, not a data fit. Invented entities are methodological (conditional point process, N=3 surmise) rather than new physical objects.

free parameters (1)
  • elliptic deformation parameter (Ginibre)
    Appears in the parameter-dependent N=3 surmise that interpolates to GUE; treated as a model parameter of the elliptic Ginibre ensemble rather than a fit to external data.
axioms (4)
  • domain assumption There are three generic local bulk statistics among non-Hermitian RMT symmetry classes, recently extended to three generic local edge statistics.
    Stated as the starting conjecture in the abstract; the paper studies representatives of those classes rather than proving the classification.
  • domain assumption Complex Ginibre, complex symmetric, and complex self-dual Gaussian ensembles represent classes A, AI†, AII† for local edge statistics.
    Abstract calls them 'the three simplest representatives'; universality of edge statistics rests on this identification.
  • domain assumption Local repulsion can be interpreted via an effective 2D Coulomb gas at different inverse temperatures β.
    Used in the abstract to explain varying degrees of edge repulsion across the three classes.
  • standard math Standard definitions of complex spacing ratios and nearest-neighbour spacing distributions for point processes in the complex plane.
    Background RMT/point-process machinery assumed throughout the abstract.
invented entities (2)
  • Conditional point process for complex spacing ratios in class A no independent evidence
    purpose: Simplify finite-N expressions for the complex spacing ratio and explain convergence of an earlier uncontrolled approximation in the bulk large-N limit.
    Methodological construction introduced in the analytic part; independent evidence would be the simplified formulas and large-N limit, not available in the abstract alone.
  • Parameter-dependent N=3 surmise for elliptic Ginibre complex spacing ratio no independent evidence
    purpose: Provide a simple approximate distribution interpolating between elliptic Ginibre and GUE.
    Analogous to Wigner surmise; accuracy claimed for GUE limit but not independently verified here.

pith-pipeline@v1.1.0-grok45 · 16537 in / 2865 out tokens · 32224 ms · 2026-07-13T16:18:40.865366+00:00 · methodology

0 comments
read the original abstract

The conjectured three generic local bulk statistics amongst all non-Hermitian random matrix symmetry classes have recently been extended to three generic local edge statistics. We study analytically and numerically complex spacing ratios and nearest-neighbour (NN) spacing distributions that characterise such local statistics. We choose the three simplest representatives of these universality classes, given by the Gaussian ensembles of complex Ginibre, complex symmetric and complex self-dual matrices, denoted by class A, AI$^\dag$ and AII$^\dag$. In the first part, we analytically study the complex spacing ratio in class A, at finite matrix size $N$. Introducing a conditional point process, we simplify existing expressions and show why an uncontrolled approximation introduced earlier converges well in the large-$N$ limit in the bulk. When specifying to the elliptic Ginibre ensemble, we present a parameter-dependent $N=3$ surmise for the complex spacing ratio, interpolating to that of the Gaussian unitary ensemble (GUE), where such a surmise is very accurate. In the second numerical part, we compare complex spacing ratios, its moments, and NN spacing distributions for all three ensembles with that of uncorrelated points, the two-dimensional (2D) Poisson process, both in the bulk and at the edge. The varying degree of repulsion within these different edge universality classes can be well understood in terms of an effective 2D Coulomb gas description, at different values of inverse temperature $\beta$. We find indications that the complex spacing ratio does not fully unfold the local statistics at the edge. Finally we verify that for small argument, in all three symmetry classes the NN spacing distributions in the bulk and at the edge are consistent with the universal cubic repulsion.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Interpolating non-Hermitian universality classes A and AI$^\dagger$: eigenvalue density and transition regime

    math-ph 2026-06 unverdicted novelty 7.0

    Finite-N derivation of eigenvalue density in interpolating non-Hermitian ensemble reveals transitional edge regime at σ = 1 - κ N^{-1/2} conjectured to be universal.

  2. What We Talk About When We Talk About Dissipative Quantum Chaos

    quant-ph 2026-05 unverdicted novelty 2.0

    The paper reviews spectral properties of operators for open quantum evolution and recent theoretical and experimental work on distinguishing chaotic from integrable dissipative quantum systems.

  3. Quantum chaotic systems: a random-matrix approach

    quant-ph 2026-04 unverdicted

    Review of random matrix theory application to quantum chaos, covering symmetry classes, eigenvalue statistics, unfolding, and correlation functions.

Reference graph

Works this paper leans on

16 extracted references · 12 linked inside Pith · cited by 3 Pith papers

  1. [1]

    Yushi Bai, Xin Lv, Jiajie Zhang, Hongchang Lyu, Jiankai Tang, Zhidian Huang, Zhengxiao Du, Xiao Liu, Aohan Zeng, Lei Hou, et al

    URL https://www.anthropic.com/news/ claude-sonnet-4-6. Yushi Bai, Xin Lv, Jiajie Zhang, Hongchang Lyu, Jiankai Tang, Zhidian Huang, Zhengxiao Du, Xiao Liu, Aohan Zeng, Lei Hou, et al. Longbench: A bilingual, multitask benchmark for long context understanding.arXiv preprint arXiv:2308.14508,

    Pith/arXiv arXiv

  2. [2]

    Indexcache: Accelerating sparse attention via cross-layer index reuse.arXiv preprint arXiv:2603.12201,

    Yushi Bai, Qian Dong, Ting Jiang, Xin Lv, Zhengxiao Du, Aohan Zeng, Jie Tang, and Juanzi Li. Indexcache: Accelerating sparse attention via cross-layer index reuse.arXiv preprint arXiv:2603.12201,

    arXiv

  3. [3]

    Flashattention-2: Faster attention with better parallelism and work partitioning

    Tri Dao. Flashattention-2: Faster attention with better parallelism and work partitioning. arXiv preprint arXiv:2307.08691,

    Pith/arXiv arXiv

  4. [4]

    DeepSeek-V3 technical report.arXiv preprint arXiv:2412.19437,

    DeepSeek-AI. DeepSeek-V3 technical report.arXiv preprint arXiv:2412.19437,

    Pith/arXiv arXiv

  5. [5]

    Deepseek-v3.2: Pushing the frontier of open large language models.arXiv preprint arXiv:2512.02556,

    DeepSeek-AI. Deepseek-v3.2: Pushing the frontier of open large language models.arXiv preprint arXiv:2512.02556,

    Pith/arXiv arXiv

  6. [6]

    Lazyllm: Dynamic token pruning for efficient long context LLM inference.arXiv preprint arXiv:2407.14057,

    Qichen Fu, Minsik Cho, Thomas Merth, Sachin Mehta, Mohammad Rastegari, and Mahyar Najibi. Lazyllm: Dynamic token pruning for efficient long context LLM inference.arXiv preprint arXiv:2407.14057,

    Pith/arXiv arXiv

  7. [7]

    Glm-5: from vibe coding to agentic engineering.arXiv preprint arXiv:2602.15763,

    GLM-5-Team. Glm-5: from vibe coding to agentic engineering.arXiv preprint arXiv:2602.15763,

    Pith/arXiv arXiv

  8. [9]

    Greg Kamradt

    URLhttps://www.arxiv.org/abs/2407.02490. Greg Kamradt. Needle in a haystack — pressure testing llms

    Pith/arXiv arXiv

  9. [10]

    Moba: Mixture of block attention for long-context llms.arXiv preprint arXiv:2502.13189,

    Enzhe Lu, Zhejun Jiang, Jingyuan Liu, Yulun Du, Tao Jiang, Chao Hong, Shaowei Liu, Weiran He, Enming Yuan, Yuzhi Wang, et al. Moba: Mixture of block attention for long-context llms.arXiv preprint arXiv:2502.13189,

    Pith/arXiv arXiv

  10. [11]

    MiniMax, Aonian Li, Bangwei Gong, Bo Yang, Boji Shan, Chang Liu, et al

    URL https://ai.meta.com/blog/ llama-4-multimodal-intelligence/. MiniMax, Aonian Li, Bangwei Gong, Bo Yang, Boji Shan, Chang Liu, et al. MiniMax-01: Scaling foundation models with lightning attention.arXiv preprint arXiv:2501.08313,

    Pith/arXiv arXiv

  11. [12]

    Kimi K2: Open agentic intelligence.arXiv preprint arXiv:2507.20534,

    Moonshot AI. Kimi K2: Open agentic intelligence.arXiv preprint arXiv:2507.20534,

    Pith/arXiv arXiv

  12. [13]

    Under review

    10 Preprint. Under review. Wentao Ni, Kangqi Zhang, Zhongming Yu, Oren Nelson, Mingu Lee, Hong Cai, Fatih Porikli, Jongryool Kim, Zhijian Liu, and Jishen Zhao. Double-p: Hierarchical top-p sparse attention for long-context LLMs.arXiv preprint arXiv:2602.05191,

    arXiv

  13. [14]

    Matanel Oren, Michael Hassid, Nir Rosenfeld, Yossi Adi, and Roy Schwartz

    URL https://openai.com/index/ gpt-5-4-thinking-system-card. Matanel Oren, Michael Hassid, Nir Rosenfeld, Yossi Adi, and Roy Schwartz. Transformers are multi-state RNNs. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing,

    2024

  14. [15]

    io/blog/qwen3.5

    URL https://qwenlm.github. io/blog/qwen3.5. Lei Wang, Yu Cheng, Yining Shi, Zhengju Tang, Zhiwen Mo, Wenhao Xie, Lingxiao Ma, Yuqing Xia, Jilong Xue, Fan Yang, and Zhi Yang. Tilelang: A composable tile-based programming model for ai systems.arXiv preprint arXiv:2504.17577,

    Pith/arXiv arXiv

  15. [16]

    Hierarchi- cal attention networks for document classification

    Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard Hovy. Hierarchi- cal attention networks for document classification. InProceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp. 1480–1489,

    2016

  16. [18]

    org/abs/2404.14469

    URL https://www.arxiv. org/abs/2404.14469. Jintao Zhang, Chendong Xiang, Haofeng Huang, Jia Wei, Haocheng Xi, Jun Zhu, and Jianfei Chen. Spargeattention: Accurate and training-free sparse attention accelerating any model inference. InProceedings of the 42nd International Conference on Machine Learning,

    Pith/arXiv arXiv