pith. sign in

arxiv: 2512.22803 · v2 · submitted 2025-12-28 · 🧮 math.PR · cs.DS· math-ph· math.MP

Fast mixing in Ising models with a negative spectral outlier via Gaussian approximation

Dan Mikulincer , Youngtak Sohn This is my paper

Pith reviewed 2026-05-16 19:52 UTC · model grok-4.3

classification 🧮 math.PR cs.DSmath-phmath.MP
keywords Ising modelsGlauber dynamicsmixing timeGaussian approximationStein's methodspectral outlierlogarithmic Sobolev inequalityanti-ferromagnetic models
0
0 comments X

The pith

Gaussian approximation via iterative Stein's method delivers operator-norm correlation bounds and near-optimal mixing times for Ising models with exactly one negative spectral outlier.

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

The paper shows that when the interaction matrix of an Ising model has precisely one negative eigenvalue outlier, a new covariance approximation technique produces an operator-norm bound on all pairwise correlations under arbitrary fields. This bound is obtained by iteratively applying Stein's method to quadratic tilts of sums of bounded variables and is then combined with localization schemes to derive a modified logarithmic Sobolev inequality. The resulting inequality yields near-optimal upper bounds on the mixing time of Glauber dynamics in regimes where standard spectral-width or log-concavity arguments break down. The analysis covers the anti-ferromagnetic Curie-Weiss model, anti-ferromagnetic models on expander graphs, and the Sherrington-Kirkpatrick model with negative-mean disorder. Complementary lower bounds establish that mixing becomes exponentially slow for low-temperature anti-ferromagnetic Ising models on sparse random regular graphs and Erdős-Rényi graphs.

Core claim

The central claim is that an operator-norm control on the full correlation structure of Ising models with a single negative spectral outlier can be obtained from a Gaussian approximation constructed by iterative application of Stein's method to quadratic tilts; when this control is inserted into localization arguments, it produces a modified logarithmic Sobolev inequality and therefore rapid mixing of Glauber dynamics in settings where existing spectral-gap techniques fail.

What carries the argument

The Gaussian covariance approximation obtained by iterative Stein's method applied to quadratic tilts of bounded random variables, which supplies the operator-norm bound on the correlation matrix.

If this is right

  • Near-optimal polynomial mixing-time bounds hold for Glauber dynamics on the anti-ferromagnetic Curie-Weiss model and on expander graphs with one negative outlier.
  • The same bounds apply to the Sherrington-Kirkpatrick model with negative-mean disorder under arbitrary external fields.
  • A modified logarithmic Sobolev inequality is obtained directly from the operator-norm correlation control.
  • Exponential lower bounds on mixing time are proved for low-temperature anti-ferromagnetic Ising models on sparse random regular and Erdős-Rényi graphs.

Where Pith is reading between the lines

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

  • The same Stein-based tilt iteration may extend to models whose interaction matrices have a bounded number of negative outliers provided the Gaussian approximation step can be controlled uniformly.
  • Operator-norm correlation estimates of this type could be used to analyze mixing of other local Markov chains or to bound approximation errors in variational inference for frustrated spin systems.
  • Testing whether the method remains effective when the negative outlier is allowed to approach zero would clarify the sharpness of the single-outlier hypothesis.

Load-bearing premise

The interaction matrix has exactly one negative spectral outlier while the remaining spectrum satisfies standard width or expansion conditions.

What would settle it

A direct computation or numerical check showing that the operator-norm bound on correlations fails when a second negative eigenvalue of comparable size is introduced.

read the original abstract

We study the mixing time of Glauber dynamics for Ising models in which the interaction matrix contains a single negative spectral outlier. This class includes the anti-ferromagnetic Curie-Weiss model, the anti-ferromagnetic Ising model on expander graphs, and the Sherrington-Kirkpatrick model with disorder of negative mean. Existing approaches to rapid mixing rely crucially on log-concavity or spectral width bounds and therefore can break down in the presence of a negative outlier. To address this difficulty, we develop a new covariance approximation method based on Gaussian approximation. This method is implemented via an iterative application of Stein's method to quadratic tilts of sums of bounded random variables, which may be of independent interest. The resulting analysis provides an operator-norm control of the full correlation structure under arbitrary external fields. Combined with the localization schemes of Eldan and Chen, these estimates lead to a modified logarithmic Sobolev inequality and near-optimal mixing time bounds in regimes where spectral width bounds fail. We complement these results by proving exponential lower bounds on the mixing time for low temperature anti-ferromagnetic Ising models on sparse random regular graphs and Erd\"os-R\'enyi graphs, based on the existence of gapped states as in the recent work of Sellke.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims that for Ising models whose interaction matrix has exactly one negative spectral outlier, an iterative Stein-method Gaussian approximation applied to quadratic tilts of bounded variables yields uniform operator-norm bounds on the full correlation matrix under arbitrary external fields. These bounds, when combined with Eldan-Chen localization, produce a modified logarithmic Sobolev inequality and near-optimal Glauber mixing times in regimes where spectral-width conditions fail. The results cover the anti-ferromagnetic Curie-Weiss model, anti-ferromagnetic Ising on expanders, and the Sherrington-Kirkpatrick model with negative-mean disorder. Complementary exponential lower bounds on mixing time are established for low-temperature anti-ferromagnetic Ising models on sparse random regular and Erdős-Rényi graphs via the existence of gapped states.

Significance. If the claimed operator-norm control holds uniformly, the work meaningfully extends rapid-mixing theory to a natural class of models with negative outliers that defeat standard spectral or log-concavity arguments. The iterative Stein-method construction for covariance approximation is technically novel and may be reusable beyond Ising models. The pairing of upper bounds with explicit lower bounds on sparse graphs gives a sharper picture of the mixing threshold than upper bounds alone.

major comments (2)
  1. [Covariance approximation via iterative Stein's method] The central operator-norm bound on correlations (obtained via repeated Stein-method applications to quadratic tilts) must be shown to remain O(1) uniformly in the external field, including when the field has a large projection onto the negative outlier eigenvector. The abstract asserts the bound but does not indicate how the iteration count or Stein-equation remainder controls error accumulation after the quadratic tilt shifts the effective mean and variance; this uniformity is load-bearing for feeding the bound into localization.
  2. [Modified LSI and mixing-time upper bound] The passage from the operator-norm correlation bound to the modified LSI via Eldan-Chen localization requires explicit verification that the localization scheme remains compatible with the single negative outlier; no equation or lemma in the provided description confirms that the localization error terms do not degrade when the outlier interacts with the field.
minor comments (2)
  1. The abstract would be clearer if it stated the precise mixing-time upper bound obtained (e.g., O(n log n) or polylog(n) factors).
  2. Notation for the interaction matrix, its spectral decomposition, and the definition of the negative outlier should be introduced once and used consistently throughout.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. The two major points raised concern the uniformity of the covariance approximation under arbitrary fields and the compatibility of the localization argument with the negative outlier; we address each below and commit to revisions that strengthen the presentation without altering the core claims.

read point-by-point responses

  1. Referee: [Covariance approximation via iterative Stein's method] The central operator-norm bound on correlations (obtained via repeated Stein-method applications to quadratic tilts) must be shown to remain O(1) uniformly in the external field, including when the field has a large projection onto the negative outlier eigenvector. The abstract asserts the bound but does not indicate how the iteration count or Stein-equation remainder controls error accumulation after the quadratic tilt shifts the effective mean and variance; this uniformity is load-bearing for feeding the bound into localization.

    Authors: We agree that the uniformity argument merits an explicit lemma. In the full manuscript (Section 3), the iteration count is taken to be O(log n) and the Stein remainder is controlled via the boundedness of the spins together with the spectral gap away from the outlier. To make the field-uniformity transparent, we will insert a new Lemma 3.5 that tracks the mean and variance shifts induced by the quadratic tilt. The proof shows that the operator-norm error remains bounded by a constant independent of the field strength (including large projections onto the outlier eigenvector) because the Gaussian approximation error is O(1/sqrt(n)) uniformly and the iteration contracts the error at a rate determined solely by the fixed outlier gap. This lemma will be placed immediately before the application to localization. revision: yes

  2. Referee: [Modified LSI and mixing-time upper bound] The passage from the operator-norm correlation bound to the modified LSI via Eldan-Chen localization requires explicit verification that the localization scheme remains compatible with the single negative outlier; no equation or lemma in the provided description confirms that the localization error terms do not degrade when the outlier interacts with the field.

    Authors: We thank the referee for highlighting this gap in exposition. The manuscript applies Eldan-Chen localization after the covariance bound has been established uniformly in the field, so the localization error is controlled by the same operator-norm quantity. In the revision we will add a short subsection (Section 4.2) containing Lemma 4.2, which verifies that the localization error terms remain O(1) under the single-negative-outlier assumption: the proof uses the fact that the outlier direction is handled by the preceding tilt and does not interact adversely with the localization measure because the covariance bound already absorbs any field-induced shift. Equation (4.3) will record the resulting modified LSI constant explicitly. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation uses external Stein-method construction and localization schemes

full rationale

The paper's central chain applies an iterative Stein's method to quadratic tilts of bounded variables to obtain operator-norm covariance control, then feeds this into Eldan-Chen localization for a modified log-Sobolev inequality. Both the Stein iteration and the localization are drawn from prior external literature (no author overlap with Mikulincer-Sohn). The single-negative-outlier assumption is stated explicitly as a modeling hypothesis rather than derived from the target mixing-time bound. No equation reduces a prediction to a fitted parameter defined on the same data, and no self-citation is load-bearing for the uniqueness or validity of the approximation. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The argument relies on the existence of a single negative eigenvalue outlier together with standard assumptions on the remaining spectrum; the Stein-method iteration is treated as a black-box tool whose error terms are controlled under boundedness of the variables. No new entities are postulated.

axioms (2)
  • domain assumption The interaction matrix has exactly one negative eigenvalue whose magnitude exceeds the spectral width of the remaining eigenvalues.
    Stated in the first paragraph of the abstract as the class of models under study.
  • standard math Stein's method applied iteratively to quadratic tilts of sums of bounded random variables yields an operator-norm covariance approximation.
    Invoked as the core technical engine; treated as a known or derivable tool rather than proved from scratch in the abstract.

pith-pipeline@v0.9.0 · 5521 in / 1612 out tokens · 18757 ms · 2026-05-16T19:52:01.573514+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [1]

    el Ducatez, and Antti Knowles. Extremal eigenvalues of critical E rd\

    Johannes Alt, Rapha\"el Ducatez, and Antti Knowles. Extremal eigenvalues of critical E rd\"os- R \'enyi graphs. Ann. Probab. , 49(3):1347--1401, 2021

    work page 2021

  2. [2]

    Fast relaxation of the random field ising dynamics

    Ahmed El Alaoui, Ronen Eldan, Reza Gheissari, and Arianna Piana. Fast relaxation of the random field ising dynamics. Ann. Probab. to appear , 2023

    work page 2023

  3. [3]

    Anderson, Alice Guionnet, and Ofer Zeitouni

    Greg W. Anderson, Alice Guionnet, and Ofer Zeitouni. An Introduction to Random Matrices , volume 118 of Cambridge Studies in Advanced Mathematics . Cambridge University Press, Cambridge, 2010

    work page 2010

  4. [4]

    Entropic independence: optimal mixing of down-up random walks

    Nima Anari, Vishesh Jain, Frederic Koehler, Huy Tuan Pham, and Thuy-Duong Vuong. Entropic independence: optimal mixing of down-up random walks. In Proceedings of the 54th Annual ACM SIGACT Symposium on Theory of Computing , pages 1418--1430, 2022

    work page 2022

  5. [5]

    Trickle-down in localization schemes and applications

    Nima Anari, Frederic Koehler, and Thuy-Duong Vuong. Trickle-down in localization schemes and applications. In Proceedings of the 56th Annual ACM Symposium on Theory of Computing , STOC 2024, page 1094–1105, New York, NY, USA, 2024. Association for Computing Machinery

    work page 2024

  6. [6]

    Spectral independence in high-dimensional expanders and applications to the hardcore model

    Nima Anari, Kuikui Liu, and Shayan Oveis Gharan. Spectral independence in high-dimensional expanders and applications to the hardcore model. SIAM J. Comput. , 53(6):FOCS20--1--FOCS20--37, 2024

    work page 2024

  7. [7]

    A transportation approach to the mean-field approximation

    Fanny Augeri. A transportation approach to the mean-field approximation. Probability Theory and Related Fields , 180(1):1--32, 2021

    work page 2021

  8. [8]

    Fluctuation of the free energy of sherrington--kirkpatrick model with curie--weiss interaction: The paramagnetic regime

    Debapratim Banerjee. Fluctuation of the free energy of sherrington--kirkpatrick model with curie--weiss interaction: The paramagnetic regime. Journal of Statistical Physics , 178(1):211--246, 2020

    work page 2020

  9. [9]

    A very simple proof of the LSI for high temperature spin systems

    Roland Bauerschmidt and Thierry Bodineau. A very simple proof of the LSI for high temperature spin systems. J. Funct. Anal. , 276(8):2582--2588, 2019

    work page 2019

  10. [10]

    Kawasaki dynamics beyond the uniqueness threshold

    Roland Bauerschmidt, Thierry Bodineau, and Benoit Dagallier. Kawasaki dynamics beyond the uniqueness threshold. Probability Theory and Related Fields , 192(1):267--302, 2025

    work page 2025

  11. [11]

    Julius Borcea, Petter Br \"a nd \'e n, and Thomas M. Liggett. Negative dependence and the geometry of polynomials. Journal of the American Mathematical Society , 22(2):521--567, 2009

    work page 2009

  12. [12]

    Spectral radii of sparse random matrices

    Florent Benaych-Georges, Charles Bordenave, and Antti Knowles. Spectral radii of sparse random matrices. Ann. Inst. Henri Poincar\'e Probab. Stat. , 56(3):2141--2161, 2020

    work page 2020

  13. [13]

    Analysis and geometry of M arkov diffusion operators , volume 348 of Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]

    Dominique Bakry, Ivan Gentil, and Michel Ledoux. Analysis and geometry of M arkov diffusion operators , volume 348 of Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] . Springer, Cham, 2014

    work page 2014

  14. [14]

    On extensions of the brunn-minkowski and prékopa-leindler theorems, including inequalities for log concave functions, and with an application to the diffusion equation

    Herm Jan Brascamp and Elliott H Lieb. On extensions of the brunn-minkowski and prékopa-leindler theorems, including inequalities for log concave functions, and with an application to the diffusion equation. Journal of Functional Analysis , 22(4):366--389, 1976

    work page 1976

  15. [15]

    Fluctuations of the free energy of the spherical S herrington- K irkpatrick model with ferromagnetic interaction

    Jinho Baik and Ji Oon Lee. Fluctuations of the free energy of the spherical S herrington- K irkpatrick model with ferromagnetic interaction. Ann. Henri Poincar\'e , 18(6):1867--1917, 2017

    work page 1917

  16. [16]

    Ferromagnetic to paramagnetic transition in spherical spin glass

    Jinho Baik, Ji Oon Lee, and Hao Wu. Ferromagnetic to paramagnetic transition in spherical spin glass. J. Stat. Phys. , 173(5):1484--1522, 2018

    work page 2018

  17. [17]

    Universality of the mean-field for the potts model

    Anirban Basak and Sumit Mukherjee. Universality of the mean-field for the potts model. Probability Theory and Related Fields , 168(3):557--600, 2017

    work page 2017

  18. [18]

    A new proof of F riedman's second eigenvalue theorem and its extension to random lifts

    Charles Bordenave. A new proof of F riedman's second eigenvalue theorem and its extension to random lifts. Ann. Sci. \'Ec. Norm. Sup\'er. (4) , 53(6):1393--1439, 2020

    work page 2020

  19. [19]

    Bobkov and Prasad Tetali

    Sergey G. Bobkov and Prasad Tetali. Modified logarithmic S obolev inequalities in discrete settings. J. Theoret. Probab. , 19(2):289--336, 2006

    work page 2006

  20. [20]

    Rapid mixing on random regular graphs beyond uniqueness

    Xiaoyu Chen, Zejia Chen, Zongchen Chen, Yitong Yin, and Xinyuan Zhang. Rapid mixing on random regular graphs beyond uniqueness. arXiv preprint arXiv:2504.03406 , 2025

    work page arXiv 2025

  21. [21]

    Factorizations of relative entropy using stochastic localization

    Pietro Caputo, Zongchen Chen, and Daniel Parisi. Factorizations of relative entropy using stochastic localization. arXiv preprint arXiv:2503.19419 , 2025

    work page arXiv 2025

  22. [22]

    Rapid mixing at the uniqueness threshold

    Xiaoyu Chen, Zongchen Chen, Yitong Yin, and Xinyuan Zhang. Rapid mixing at the uniqueness threshold. In S TOC '25--- P roceedings of the 57th A nnual ACM S ymposium on T heory of C omputing , pages 879--890. ACM, New York, [2025] 2025

    work page 2025

  23. [23]

    Localization schemes: a framework for proving mixing bounds for M arkov chains

    Yuansi Chen and Ronen Eldan. Localization schemes: a framework for proving mixing bounds for M arkov chains. Duke Math. J. , 174(8):1431--1510, 2025

    work page 2025

  24. [24]

    A new approach to strong convergence

    Chi-Fang Chen, Jorge Garza-Vargas, Joel A Tropp, and Ramon Van Handel. A new approach to strong convergence. Ann. of Math., to appear , 2024

    work page 2024

  25. [25]

    On the mixed even-spin S herrington- K irkpatrick model with ferromagnetic interaction

    Wei-Kuo Chen. On the mixed even-spin S herrington- K irkpatrick model with ferromagnetic interaction. Ann. Inst. Henri Poincar\'e Probab. Stat. , 50(1):63--83, 2014

    work page 2014

  26. [26]

    Maximally-stable local optima in random graphs and spin glasses: Phase transitions and universality

    Yatin Dandi, David Gamarnik, and Lenka Zdeborov\' a . Maximally-stable local optima in random graphs and spin glasses: Phase transitions and universality. arXiv preprint arXiv:2305.03591 , 2023

    work page arXiv 2023

  27. [27]

    Edgeworth expansions for independent bounded integer valued random variables

    Dmitry Dolgopyat and Yeor Hafouta. Edgeworth expansions for independent bounded integer valued random variables. Stochastic Processes and their Applications , 152:486--532, 2022

    work page 2022

  28. [28]

    R. L. Dobrushin. The problem of uniqueness of a G ibbsian random field and the problem of phase transitions. Funkcional. Anal. i Prilozen. , 2(4):44--57, 1968

    work page 1968

  29. [29]

    Large deviations techniques and applications , volume 38 of Stochastic Modelling and Applied Probability

    Amir Dembo and Ofer Zeitouni. Large deviations techniques and applications , volume 38 of Stochastic Modelling and Applied Probability . Springer-Verlag, Berlin, 2010

    work page 2010

  30. [30]

    Sampling from mean-field gibbs measures via diffusion processes

    Ahmed El Alaoui, Andrea Montanari, and Mark Sellke. Sampling from mean-field gibbs measures via diffusion processes. Probability and Mathematical Physics , 6(3):961--1022, 2025

    work page 2025

  31. [31]

    A spectral condition for spectral gap: fast mixing in high-temperature I sing models

    Ronen Eldan, Frederic Koehler, and Ofer Zeitouni. A spectral condition for spectral gap: fast mixing in high-temperature I sing models. Probab. Theory Related Fields , 182(3-4):1035--1051, 2022

    work page 2022

  32. [32]

    Thin shell implies spectral gap up to polylog via a stochastic localization scheme

    Ronen Eldan. Thin shell implies spectral gap up to polylog via a stochastic localization scheme. Geom. Funct. Anal. , 23(2):532--569, 2013

    work page 2013

  33. [33]

    The CLT in high dimensions: Quantitative bounds via martingale embedding

    Ronen Eldan, Dan Mikulincer, and Alex Zhai. The CLT in high dimensions: Quantitative bounds via martingale embedding . The Annals of Probability , 48(5):2494 -- 2524, 2020

    work page 2020

  34. [34]

    A proof of A lon's second eigenvalue conjecture and related problems

    Joel Friedman. A proof of A lon's second eigenvalue conjecture and related problems. Mem. Amer. Math. Soc. , 195(910):viii+100, 2008

    work page 2008

  35. [35]

    Roy J. Glauber. Time-dependent statistics of the I sing model. J. Mathematical Phys. , 4:294--307, 1963

    work page 1963

  36. [36]

    Spectral gap and edge universality of dense random regular graphs

    Yukun He. Spectral gap and edge universality of dense random regular graphs. Comm. Math. Phys. , 405(8):Paper No. 181, 40, 2024

    work page 2024

  37. [37]

    Local law and complete eigenvector delocalization for supercritical E rdos- R \'enyi graphs

    Yukun He, Antti Knowles, and Matteo Marcozzi. Local law and complete eigenvector delocalization for supercritical E rdos- R \'enyi graphs. Ann. Probab. , 47(5):3278--3302, 2019

    work page 2019

  38. [38]

    A fast berry-esseen theorem under minimal density assumptions

    Samuel GG Johnston. A fast berry-esseen theorem under minimal density assumptions. arXiv preprint arXiv:2305.18138 , 2023

    work page arXiv 2023

  39. [39]

    Sampling approximately low-rank ising models: Mcmc meets variational methods

    Frederic Koehler, Holden Lee, and Andrej Risteski. Sampling approximately low-rank ising models: Mcmc meets variational methods. In Po-Ling Loh and Maxim Raginsky, editors, Proceedings of Thirty Fifth Conference on Learning Theory , volume 178 of Proceedings of Machine Learning Research , pages 4945--4988. PMLR, 02--05 Jul 2022

    work page 2022

  40. [40]

    Infinite-ranged models of spin-glasses

    Scott Kirkpatrick and David Sherrington. Infinite-ranged models of spin-glasses. Phys. Rev. B , 17:4384--4403, Jun 1978

    work page 1978

  41. [41]

    The largest eigenvalue of sparse random graphs

    Michael Krivelevich and Benny Sudakov. The largest eigenvalue of sparse random graphs. Combin. Probab. Comput. , 12(1):61--72, 2003

    work page 2003

  42. [42]

    Levin, Malwina J

    David A. Levin, Malwina J. Luczak, and Yuval Peres. Glauber dynamics for the mean-field ising model: cut-off, critical power law, and metastability. Probability Theory and Related Fields , 146(1):223, 2008

    work page 2008

  43. [43]

    Le, Elizaveta Levina, and Roman Vershynin

    Can M. Le, Elizaveta Levina, and Roman Vershynin. Concentration and regularization of random graphs. Random Structures Algorithms , 51(3):538--561, 2017

    work page 2017

  44. [44]

    Kuikui Liu, Sidhanth Mohanty, Amit Rajaraman, and David X. Wu. Fast mixing in sparse random I sing models. In 2024 IEEE 65th A nnual S ymposium on F oundations of C omputer S cience--- FOCS 2024 , pages 120--128. IEEE Computer Soc., Los Alamitos, CA, [2024] 2024

    work page 2024

  45. [45]

    D. S. Mitrinovi \'c , J. E. Pecari \'c , and A. M. Fink. Classical and new inequalities in analysis , volume 61 of Mathematics and its Applications (East European Series) . Kluwer Academic Publishers Group, Dordrecht, 1993

    work page 1993

  46. [46]

    On perfectly friendly bisections of random graphs

    Dor Minzer, Ashwin Sah, and Mehtaab Sawhney. On perfectly friendly bisections of random graphs . The Annals of Probability , 52(6):2281 -- 2341, 2024

    work page 2024

  47. [47]

    Normal approximations with Malliavin calculus: from Stein's method to universality , volume 192

    Ivan Nourdin and Giovanni Peccati. Normal approximations with Malliavin calculus: from Stein's method to universality , volume 192. Cambridge University Press, 2012

    work page 2012

  48. [48]

    Polynomial mixing of the critical ising model on sparse erdos-renyi graphs

    Kyprianos-Iason Prodromidis and Allan Sly. Polynomial mixing of the critical ising model on sparse erdos-renyi graphs. arXiv preprint arXiv:2510.07254 , 2025

    work page arXiv 2025

  49. [49]

    The spectral gap of random regular graphs

    Amir Sarid. The spectral gap of random regular graphs. Random Structures Algorithms , 63(2):557--587, 2023

    work page 2023

  50. [50]

    Exponentially slow mixing of the low temperature SK model

    Mark Sellke. Exponentially slow mixing of the low temperature SK model. arXiv preprint arXiv:2511.22621 , 2025

    work page arXiv 2025

  51. [51]

    Solvable Model of a Spin-Glass

    David Sherrington and Scott Kirkpatrick. Solvable Model of a Spin-Glass . Phys. Rev. Lett. , 35(26):1792--1796, 1975

    work page 1975

  52. [52]

    On graphs with a fixed number of negative eigenvalues

    Aleksandar Torgasev. On graphs with a fixed number of negative eigenvalues. Discrete Math. , 57(3):311--317, 1985

    work page 1985

  53. [53]

    Combinatorial criteria for uniqueness of G ibbs measures

    Dror Weitz. Combinatorial criteria for uniqueness of G ibbs measures. Random Structures Algorithms , 27(4):445--475, 2005

    work page 2005