pith. sign in

arxiv: 2605.03718 · v2 · pith:2NBNCJJLnew · submitted 2026-05-05 · 🧮 math.PR · cs.DS· math-ph· math.MP

Potential Hessian Ascent III: Sampling the Sherrington--Kirkpatrick Model at Beta < 1/2

Ewan Davies , Holden Lee , Juspreet Singh Sandhu , Jonathan Shi This is my paper

Pith reviewed 2026-05-20 23:41 UTC · model grok-4.3

classification 🧮 math.PR cs.DSmath-phmath.MP
keywords Sherrington-Kirkpatrick modelGibbs samplingstochastic localizationHessian ascentspin glassestotal variation distancehigh-temperature regimepolynomial-time algorithm
0
0 comments X

The pith

A polynomial-time algorithm samples the Sherrington-Kirkpatrick Gibbs measure with negligible total-variation error for inverse temperatures below 1/2.

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

The paper develops a sampling algorithm for the equilibrium distribution of the Sherrington-Kirkpatrick model that runs in polynomial time and produces output whose total variation distance to the true Gibbs measure is negligible when the inverse temperature satisfies beta less than 1/2. Earlier algorithms achieved such guarantees only up to roughly beta 0.295 or settled for weaker Wasserstein guarantees across the full replica-symmetric range beta less than 1. The method repurposes an existing optimization procedure called potential Hessian ascent to carry out stochastic localization, using covariance estimates obtained from Gaussian integration by parts, overlap concentration, and cavity methods to control the localization error.

Core claim

Potential Hessian ascent implements algorithmic stochastic localization at high temperature. Covariance estimates for the tilted Gibbs distribution, derived via Gaussian integration by parts, overlap concentration, and precise cavity estimates, yield an O(1) Wasserstein error guarantee for finite-time localization. A free-probability argument controlling the diagonal sub-algebra of the Hessian upgrades this to O(1) KL divergence. Jarzynski equality with rejection sampling together with entropy contraction on the time-T localized distribution then reduces the error to o(1) in total variation distance, after which the polarized walk completes the sampling.

What carries the argument

Potential Hessian ascent process, which performs algorithmic stochastic localization by estimating the covariance of the tilted Gibbs distribution via Gaussian integration by parts and cavity estimates.

If this is right

  • Sampling from the Gibbs measure of the SK model becomes feasible in polynomial time with negligible TVD error for all beta less than 1/2.
  • The same Hessian-ascent procedure that solves optimization tasks also produces accurate samples via stochastic localization.
  • Localization error improves from o(n) to O(1) in Wasserstein distance once covariance estimates are inserted.
  • KL-divergence control follows from a free-probability argument on the Hessian diagonal sub-algebra.
  • Jarzynski equality combined with rejection sampling and entropy contraction converts the O(1) KL guarantee into o(1) TVD error before the polarized walk finishes the sample.

Where Pith is reading between the lines

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

  • The covariance estimation techniques may extend to other mean-field spin glasses whose overlap concentrates at high temperature.
  • Pushing the same localization-plus-refinement pipeline past beta = 1/2 would require new control on the Hessian spectrum once replica symmetry breaks.
  • The free-probability argument used for the diagonal sub-algebra suggests that similar random-matrix tools could bound error in other stochastic-localization algorithms.
  • Practical tests on instances of size a few hundred spins could verify whether the theoretical O(1) constants remain small enough for concrete use.

Load-bearing premise

The covariance estimates obtained for the tilted Gibbs distribution are accurate enough to make the Hessian-ascent localization reach only O(1) Wasserstein error before the KL and TVD refinements are applied.

What would settle it

Numerical simulation of the full pipeline on moderate-sized SK instances at beta slightly larger than 1/2, checking whether the empirical total-variation distance to the true Gibbs measure stays o(1) or begins to grow.

read the original abstract

We give a polynomial-time algorithm to sample from the Gibbs measure of the Sherrington-Kirkpatrick model with negligible total-variation distance (TVD) error up to inverse temperature $\beta < 1/2$. Prior work obtained TVD error guarantees only up to $\beta\approx 0.295$, while results covering the entire replica-symmetric regime $\beta < 1$ gave guarantees only in Wasserstein distance. Our approach demonstrates that the same potential Hessian ascent previously developed for optimization also functions as a sampling algorithm by implementing algorithmic stochastic localization at high temperature. By estimating the covariance of the tilted Gibbs distribution via Gaussian integration by parts, overlap concentration, and precise cavity estimates, we show that a Hessian-ascent process achieves an $O(1)$ Wasserstein error guarantee for finite-time localization, improving on the previous $o(n)$. A careful comparison of stochastic localization with the Hessian ascent process and a free probability argument controlling the diagonal sub-algebra of the Hessian improves this to $O(1)$ in KL divergence. We then use Jarzynski's equality with rejection sampling, along with entropy contraction on the time-$T$ localized distribution, to refine the error to $o(1)$ in TVD up to a constant time $T$ and to complete the sampling with the polarized walk.

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

1 major / 2 minor

Summary. The manuscript claims a polynomial-time algorithm to sample from the Gibbs measure of the Sherrington-Kirkpatrick model with o(1) total-variation distance error for β < 1/2. It implements algorithmic stochastic localization via potential Hessian ascent, controls the covariance of the tilted measure through Gaussian integration by parts, overlap concentration, and cavity estimates to obtain an O(1) Wasserstein guarantee after finite time, upgrades this to O(1) KL divergence via a free-probability comparison on the diagonal sub-algebra of the Hessian, and finally applies Jarzynski rejection sampling together with entropy contraction on the time-T localized distribution to reach o(1) TVD.

Significance. If the error controls hold, the result meaningfully extends the regime of provably efficient sampling with strong TVD guarantees in the SK model beyond the previous algorithmic threshold near β ≈ 0.295 while remaining inside the replica-symmetric phase. The combination of Hessian-ascent dynamics with stochastic localization and free-probability tools supplies a new technical route that may be reusable for other mean-field disordered systems.

major comments (1)
  1. [Covariance estimation for the tilted measure and finite-time localization analysis] The central step establishing the O(1) Wasserstein bound for the Hessian-ascent process (described in the abstract and the section on covariance estimates) invokes Gaussian integration by parts, overlap concentration, and precise cavity estimates for the tilted Gibbs distribution. Explicit, uniform-in-n error bounds on these estimates must be supplied and shown to remain O(1) as β → 1/2^−; without such control the base Wasserstein guarantee is not guaranteed and the subsequent KL and TVD refinements rest on an uncontrolled foundation.
minor comments (2)
  1. [Final sampling step] The precise statement of the polarized walk and its initialization from the time-T localized measure should be given explicitly, including any auxiliary parameters.
  2. [Introduction] A short table or paragraph comparing the new TVD threshold β < 1/2 with all prior algorithmic and information-theoretic thresholds would improve readability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading and for identifying the need for greater explicitness in the error controls. We address the major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Covariance estimation for the tilted measure and finite-time localization analysis] The central step establishing the O(1) Wasserstein bound for the Hessian-ascent process (described in the abstract and the section on covariance estimates) invokes Gaussian integration by parts, overlap concentration, and precise cavity estimates for the tilted Gibbs distribution. Explicit, uniform-in-n error bounds on these estimates must be supplied and shown to remain O(1) as β → 1/2^−; without such control the base Wasserstein guarantee is not guaranteed and the subsequent KL and TVD refinements rest on an uncontrolled foundation.

    Authors: We agree that the covariance estimates require more explicit uniform-in-n bounds with clear dependence on β. In the current draft, Gaussian integration by parts (Lemma 3.2), overlap concentration (Theorem 4.1), and cavity estimates (Proposition 5.3) are used to obtain an O(1) Wasserstein guarantee after finite time, but the error terms are stated asymptotically as n→∞ for fixed β<1/2 without fully expanded constants. We will add an appendix deriving explicit bounds showing that all error terms are at most C(β)·n^{-c} for some c>0, where C(β) remains finite for each fixed β<1/2. Regarding uniformity as β→1/2^−, the constants C(β) do grow (as expected near the high-temperature boundary), but remain O(1) for any fixed β<1/2; the subsequent KL and TVD arguments are unaffected because the algorithm runs in time polynomial in n for each such β. We will clarify this distinction in the revision. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation uses independent standard tools and derives new error bounds from its localization process

full rationale

The paper estimates covariance of the tilted Gibbs distribution using Gaussian integration by parts, overlap concentration, and precise cavity estimates to establish an O(1) Wasserstein guarantee for finite-time Hessian-ascent localization. It then compares stochastic localization to the ascent process, applies a free-probability argument on the Hessian diagonal, invokes Jarzynski equality with rejection sampling, and uses entropy contraction to reach o(1) TVD. These steps build on domain-standard techniques whose validity is external to the present algorithm and do not reduce by construction to fitted parameters, self-definitions, or unverified self-citations; the O(1) improvements are obtained directly from the paper's own finite-time analysis rather than being presupposed.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on domain-standard concentration and approximation results from statistical physics applied to tilted distributions; no new free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption Overlap concentration holds for the Sherrington-Kirkpatrick model and its tilted versions at beta < 1/2
    Invoked to control covariance estimates in the localization process
  • domain assumption Precise cavity estimates remain valid under the algorithmic stochastic localization dynamics
    Used to obtain the O(1) Wasserstein guarantee for finite-time localization

pith-pipeline@v0.9.0 · 5787 in / 1597 out tokens · 44413 ms · 2026-05-20T23:41:07.228433+00:00 · methodology

Review history (2 revisions) →

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

138 extracted references · 138 canonical work pages · 3 internal anchors

  1. [1]

    Sampling from

    Wang, Chenguang and Cui, Kaiyuan and Zhao, Weichen and Yu, Tianshu , booktitle=. Sampling from. 2025 , organization=

    work page 2025

  2. [2]

    International Conference on Machine Learning , pages=

    Sampling from Binary Quadratic Distributions via Stochastic Localization , author=. International Conference on Machine Learning , pages=. 2025 , organization=

    work page 2025

  3. [3]

    2008 , publisher=

    Optimization algorithms on matrix manifolds , author=. 2008 , publisher=

    work page 2008

  4. [4]

    17th Conference on the Theory of Quantum Computation, Communication and Cryptography (TQC 2022) , pages=

    The Quantum Approximate Optimization Algorithm at High Depth for MaxCut on Large-Girth Regular Graphs and the Sherrington-Kirkpatrick Model , author=. 17th Conference on the Theory of Quantum Computation, Communication and Cryptography (TQC 2022) , pages=. 2022 , organization=

    work page 2022

  5. [5]

    Quantum , volume=

    The quantum approximate optimization algorithm and the Sherrington-Kirkpatrick model at infinite size , author=. Quantum , volume=. 2022 , publisher=

    work page 2022

  6. [6]

    Foundations and Trends in Machine Learning , volume=

    Graphical models, exponential families, and variational inference , author=. Foundations and Trends in Machine Learning , volume=. 2008 , publisher=

    work page 2008

  7. [7]

    Dynamical

    Adhikari, Arka and Brennecke, Christian and. Dynamical. Journal of Statistical Physics , volume =. doi:10.1007/s10955-021-02773-7 , langid =

    work page doi:10.1007/s10955-021-02773-7

  8. [8]

    Proceedings of the 56th Annual ACM Symposium on Theory of Computing , pages=

    Semidefinite programs simulate approximate message passing robustly , author=. Proceedings of the 56th Annual ACM Symposium on Theory of Computing , pages=

    work page

  9. [9]

    The Annals of Probability , pages=

    The Largest Eigenvalues of Finite Rank Deformation of Large Wigner Matrices: Convergence and Nonuniversality of the Fluctuations , author=. The Annals of Probability , pages=. 2009 , publisher=

    work page 2009

  10. [10]

    Electron

    Free convolution with a semicircular distribution and eigenvalues of spiked deformations of Wigner matrices , author=. Electron. J. Probab , volume=

    work page

  11. [11]

    Indiana University Mathematics Journal , pages=

    On the free convolution with a semi-circular distribution , author=. Indiana University Mathematics Journal , pages=. 1997 , publisher=

    work page 1997

  12. [12]

    American Mathematical Society , year=

    Stochastic calculus: An introduction with applications , author=. American Mathematical Society , year=

    work page

  13. [13]

    Advances in Mathematics , volume=

    Rigidity of eigenvalues of generalized Wigner matrices , author=. Advances in Mathematics , volume=. 2012 , publisher=

    work page 2012

  14. [14]

    Bai, Z. D. and Yin, Y. Q. , year = 1993, journal =. Limit of the

    work page 1993

  15. [15]

    Probability Theory and Related Fields , volume =

    Spectral gap estimates for mixed p -spin models at high temperature , author =. Probability Theory and Related Fields , volume =. doi:10.1007/s00440-024-01261-9 , langid =

    work page doi:10.1007/s00440-024-01261-9

  16. [16]

    Operator

    Brennecke, Christian and Xu, Changji and Yau, Horng-Tzer , year = 2025, number =. Operator. doi:10.48550/arXiv.2307.12535 , archiveprefix =. 2307.12535 , primaryclass =

    work page doi:10.48550/arxiv.2307.12535 2025

  17. [17]

    Akhiezer, N. I. and Glazman, I. M. , year = 1993, publisher =. Theory of Linear Operators in

    work page 1993

  18. [18]

    The Annals of Probability , volume =

    Optimization of Mean-Field Spin Glasses , author =. The Annals of Probability , volume =

    work page

  19. [19]

    arXiv preprint arXiv:2410.02711 , year =

    Nets: A non-equilibrium transport sampler , author =. arXiv preprint arXiv:2410.02711 , year =

    work page arXiv

  20. [20]

    Near-optimal quantum algorithms for multivariate mean estimation,

    Entropic independence: optimal mixing of down-up random walks , shorttitle =. Proceedings of the 54th Annual ACM SIGACT Symposium on Theory of Computing , author =. doi:10.1145/3519935.3520048 , isbn =

    work page doi:10.1145/3519935.3520048

  21. [21]

    Trickle-

    Anari, Nima and Koehler, Frederic and Vuong, Thuy-Duong , year = 2024, series =. Trickle-. Proceedings of the 56th. doi:10.1145/3618260.3649622 , isbn =

    work page doi:10.1145/3618260.3649622 2024

  22. [22]

    2010 , publisher =

    An introduction to random matrices , author =. 2010 , publisher =

    work page 2010

  23. [23]

    Aubrun, Guillaume , editor =. A. S\'eminaire de. doi:10.1007/978-3-540-31449-3_22 , isbn =

    work page doi:10.1007/978-3-540-31449-3_22

  24. [24]

    Bakry, D. and. doi:10.1007/BFb0075847 , isbn =

    work page doi:10.1007/bfb0075847

  25. [25]

    Inventiones mathematicae , volume =

    Matrix Concentration Inequalities and Free Probability , author =. Inventiones mathematicae , volume =. doi:10.1007/s00222-023-01204-6 , langid =

    work page doi:10.1007/s00222-023-01204-6

  26. [26]

    A Very Simple Proof of the

    Bauerschmidt, Roland and Bodineau, Thierry , year = 2019, journal =. A Very Simple Proof of the

    work page 2019

  27. [27]

    2005 , publisher =

    Complex analysis methods in noncommutative probability , author =. 2005 , publisher =

    work page 2005

  28. [28]

    Nearly d-linear convergence bounds for diffu- sion models via stochastic localization.arXiv preprint arXiv:2308.03686,

    Nearly d -linear convergence bounds for diffusion models via stochastic localization , author =. arXiv preprint arXiv:2308.03686 , year =

    work page arXiv

  29. [29]

    Mathematische Zeitschrift , volume =

    Processes with Free Increments , author =. Mathematische Zeitschrift , volume =. doi:10.1007/PL00004363 , langid =

    work page doi:10.1007/pl00004363

  30. [30]

    Scale-Free Networks: Complex Webs in Nature and Technology

    Boucheron, Stéphane and Lugosi, Gábor and Massart, Pascal , title =. 2013 , isbn =. doi:10.1093/acprof:oso/9780199535255.001.0001 , url =

    work page doi:10.1093/acprof:oso/9780199535255.001.0001 2013

  31. [31]

    2016 , publisher =

    Metastability: a potential-theoretic approach , author =. 2016 , publisher =

    work page 2016

  32. [32]

    The Two Point Function of the

    Brennecke, Christian and Schertzer, Adrien and Xu, Changji and Yau, Horng-Tzer , year = 2024, journal =. The Two Point Function of the. doi:10.2140/pmp.2024.5.131 , langid =

    work page doi:10.2140/pmp.2024.5.131 2024

  33. [33]

    Cadel, Agnese and Rovira, Carles , year = 2010, journal =. The. 44239929 , eprinttype =

    work page 2010

  34. [34]

    Localization Schemes:

    Chen, Yuansi and Eldan, Ronen , year = 2025, journal =. Localization Schemes:

    work page 2025

  35. [35]

    Sudakov--

    Celentano, Michael , year = 2024, journal =. Sudakov--

    work page 2024

  36. [36]

    Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions.arXiv preprint arXiv:2209.11215,

    Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions , author =. arXiv preprint arXiv:2209.11215 , year =

    work page arXiv

  37. [37]

    Improved Analysis of Score-Based Generative Modeling: User-Friendly Bounds under Minimal Smoothness Assumptions , shorttitle =

    Chen, Hongrui and Lee, Holden and Lu, Jianfeng , year = 2023, series =. Improved Analysis of Score-Based Generative Modeling: User-Friendly Bounds under Minimal Smoothness Assumptions , shorttitle =. Proceedings of the 40th

    work page 2023

  38. [38]

    Proceedings of the National Academy of Sciences , volume=

    The overlap gap property: A topological barrier to optimizing over random structures , author=. Proceedings of the National Academy of Sciences , volume=. 2021 , publisher=

    work page 2021

  39. [39]

    The Interaction between Multioverlaps in the High Temperature Phase of the

    Crawford, Nicholas , year = 2008, journal =. The Interaction between Multioverlaps in the High Temperature Phase of the

    work page 2008

  40. [40]

    2019 , publisher =

    Probability: theory and examples , author =. 2019 , publisher =

    work page 2019

  41. [41]

    Proceedings of Symposia in Applied Mathematics , author =

    Computing the volume of convex bodies: a case where randomness provably helps , shorttitle =. Proceedings of Symposia in Applied Mathematics , author =. doi:10.1090/psapm/044/1141926 , isbn =

    work page doi:10.1090/psapm/044/1141926

  42. [42]

    Sampling from the

    Alaoui, Ahmed El and Montanari, Andrea and Sellke, Mark , year = 2022, pages =. Sampling from the. 2022

    work page 2022

  43. [43]

    Bounds on the Covariance Matrix of the

    Alaoui, Ahmed El and Gaitonde, Jason , year = 2024, journal =. Bounds on the Covariance Matrix of the

    work page 2024

  44. [44]

    Eldan, Ronen , year = 2013, journal =. Thin

    work page 2013

  45. [45]

    Probability Theory and Related Fields , volume =

    Taming Correlations through Entropy-Efficient Measure Decompositions with Applications to Mean-Field Approximation , author =. Probability Theory and Related Fields , volume =. doi:10.1007/s00440-019-00924-2 , langid =

    work page doi:10.1007/s00440-019-00924-2

  46. [46]

    A Spectral Condition for Spectral Gap: Fast Mixing in High-Temperature

    Eldan, Ronen and Koehler, Frederic and Zeitouni, Ofer , year = 2022, journal =. A Spectral Condition for Spectral Gap: Fast Mixing in High-Temperature. doi:10.1007/s00440-021-01085-x , langid =

    work page doi:10.1007/s00440-021-01085-x 2022

  47. [47]

    Sampling from Mean-Field

    Alaoui, Ahmed El and Montanari, Andrea and Sellke, Mark , year =. Sampling from Mean-Field. Probability and Mathematical Physics , volume =. doi:10.2140/pmp.2025.6.961 , langid =

    work page doi:10.2140/pmp.2025.6.961 2025

  48. [48]

    Friedli and Y

    Friedli, Sacha and Velenik, Yvan , year = 2017, edition =. Statistical Mechanics of Lattice Systems: A Concrete Mathematical Introduction , shorttitle =. doi:10.1017/9781316882603 , isbn =

    work page doi:10.1017/9781316882603 2017

  49. [49]

    Complexity analysis of normalizing constant estimation: from

    Guo, Wei and Tao, Molei and Chen, Yongxin , journal =. Complexity analysis of normalizing constant estimation: from

    work page

  50. [50]

    Sampling from spherical spin glasses in total variation via algorithmic stochastic localization

    Sampling from spherical spin glasses in total variation via algorithmic stochastic localization , author =. arXiv preprint arXiv:2404.15651 , year =

    work page arXiv

  51. [51]

    , year = 2025, series =

    Huang, Brice and Mohanty, Sidhanth and Rajaraman, Amit and Wu, David X. , year = 2025, series =. Weak. Proceedings of the 57th. doi:10.1145/3717823.3718236 , isbn =

    work page doi:10.1145/3717823.3718236 2025

  52. [52]

    , year = 1959, journal =

    Hubbard, J. , year = 1959, journal =. Calculation of

    work page 1959

  53. [53]

    Continuous Martingales and Brownian Motion

    Revuz, Daniel and Yor, Marc. Continuous Martingales and Brownian Motion

    work page

  54. [54]

    , year = 1997, journal =

    Jarzynski, C. , year = 1997, journal =. Nonequilibrium

    work page 1997

  55. [55]

    Probability theory and related fields , volume=

    The adaptive interpolation method: a simple scheme to prove replica formulas in Bayesian inference , author=. Probability theory and related fields , volume=. 2019 , publisher=

    work page 2019

  56. [56]

    arXiv preprint arXiv:2506.19940 , year=

    Strong convergence to operator-valued semicirculars , author=. arXiv preprint arXiv:2506.19940 , year=

    work page arXiv

  57. [57]

    Infinite Dimensional Analysis, Quantum Probability and Related Topics , volume=

    Strong convergence for reduced free products , author=. Infinite Dimensional Analysis, Quantum Probability and Related Topics , volume=. 2016 , publisher=

    work page 2016

  58. [58]

    Stochastic Processes and their Applications , volume=

    On the concavity of the TAP free energy in the SK model , author=. Stochastic Processes and their Applications , volume=. 2023 , publisher=

    work page 2023

  59. [59]

    Potential

    Jekel, David and Sandhu, Juspreet Singh and Shi, Jonathan , year = 2025, pages =. Potential. Proceedings of the 2025

    work page 2025

  60. [60]

    Jekel, David and Sandhu, Juspreet Singh and Shi, Jonathan , title =

    work page

  61. [61]

    Propagation of Chaos in the Random Field

    Kabluchko, Zakhar and L. Propagation of Chaos in the Random Field. Journal of Statistical Mechanics: Theory and Experiment , volume =. doi:10.1088/1742-5468/ad8f2a , langid =

    work page doi:10.1088/1742-5468/ad8f2a

  62. [62]

    2014 , publisher =

    Brownian motion and stochastic calculus , author =. 2014 , publisher =

    work page 2014

  63. [63]

    2012 , publisher =

    Introduction to stochastic calculus with applications , author =. 2012 , publisher =

    work page 2012

  64. [64]

    Proceedings of Thirty Fifth Conference on Learning Theory , author =

    Sampling Approximately Low-Rank Ising Models: MCMC meets Variational Methods , shorttitle =. Proceedings of Thirty Fifth Conference on Learning Theory , author =

    work page

  65. [65]

    arXiv preprint arXiv:2407.00868 , year =

    Sampling from the Continuous Random Energy Model in Total Variation Distance , author =. arXiv preprint arXiv:2407.00868 , year =

    work page arXiv

  66. [66]

    Computing

    Lehner, Franz , year = 1999, journal =. Computing

    work page 1999

  67. [67]

    Characterizing

    Liang, Jiaming and Mitra, Siddharth and Wibisono, Andre , year = 2026, journal =. Characterizing

    work page 2026

  68. [68]

    2025 , eprint =

    Fast mixing in Ising models with a negative spectral outlier via Gaussian approximation , author =. 2025 , eprint =

    work page 2025

  69. [69]

    Advances in neural information processing systems , volume=

    A statistical model for tensor PCA , author=. Advances in neural information processing systems , volume=

    work page

  70. [70]

    2018 , publisher=

    Statistical inference and the sum of squares method , author=. 2018 , publisher=

    work page 2018

  71. [71]

    Mathematical Programming , volume=

    A relationship between the second derivatives of a convex function and of its conjugate , author=. Mathematical Programming , volume=. 1977 , publisher=

    work page 1977

  72. [72]

    2020 , publisher=

    An introduction to statistical mechanics and thermodynamics , author=. 2020 , publisher=

    work page 2020

  73. [73]

    2016 IEEE 57th Annual Symposium on Foundations of Computer Science (FOCS) , pages=

    Polynomial-time tensor decompositions with sum-of-squares , author=. 2016 IEEE 57th Annual Symposium on Foundations of Computer Science (FOCS) , pages=. 2016 , organization=

    work page 2016

  74. [74]

    Aubin, Benjamin and Loureiro, Bruno and Maillard, Antoine and Krzakala, Florent and Zdeborov. The. IEEE Transactions on Information Theory , volume =

    work page

  75. [75]

    Signal, image and video processing , volume=

    Application of semi-circle law and Wigner spiked-model in GPS jamming confronting , author=. Signal, image and video processing , volume=. 2023 , publisher=

    work page 2023

  76. [76]

    Random Structures & Algorithms , volume=

    On the K-sat model with large number of clauses , author=. Random Structures & Algorithms , volume=. 2018 , publisher=

    work page 2018

  77. [77]

    arXiv preprint arXiv:2504.20539 , year=

    Randomstrasse101: Open problems of 2024 , author=. arXiv preprint arXiv:2504.20539 , year=

    work page arXiv 2024

  78. [78]

    49th International Colloquium on Automata, Languages, and Programming (ICALP 2022) , pages=

    Limitations of Local Quantum Algorithms on Random MAX-k-XOR and Beyond , author=. 49th International Colloquium on Automata, Languages, and Programming (ICALP 2022) , pages=. 2022 , organization=

    work page 2022

  79. [79]

    14th Innovations in Theoretical Computer Science Conference (ITCS 2023) , pages=

    Random Max-CSPs Inherit Algorithmic Hardness from Spin Glasses , author=. 14th Innovations in Theoretical Computer Science Conference (ITCS 2023) , pages=. 2023 , organization=

    work page 2023

  80. [80]

    2017 , publisher =

    Free probability and random matrices , author =. 2017 , publisher =

    work page 2017

Showing first 80 references.