Pith. sign in

REVIEW 3 major objections 4 minor 2 cited by

Tensor Renormalization Group Meets Computer Assistance

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper proves, with a computer-assisted interval-arithmetic argument, that every tensor within 0.02 of the high-temperature fixed point in all 63 sectors flows to that fixed point under a new 2x1 tensor renormalization group map.

desk verdict A genuinely new computational framework for rigorous tensor RG; the high-T result is not new physics, but the method is the contribution and it deserves peer review. read the letter →

arxiv 2506.03247 v2 pith:KHOCWMP2 submitted 2025-06-03 cond-mat.stat-mech cond-mat.str-elhep-thmath-phmath.MP

Nikolay Ebel , Tom Kennedy , Slava Rychkov This is my paper · ORCID
classification cond-mat.stat-mechcond-mat.str-elhep-thmath-phmath.MP MSC 82B2082B2865G20
keywords tensorrenormalizationgroup2x1RGmaphat-tensormasterfunctioncomputer-assistedproofintervalarithmetichigh-temperaturefixedpointlatticespinmodels
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper establishes a rigorous, computer-assisted basin of attraction for the high-temperature fixed point of a new tensor renormalization group map. It introduces the 2x1 map, which coarse-grains a lattice by a factor of two in one direction and then rotates by 90 degrees, together with a graphical calculus that turns each step of the map into inequalities on tensor components. The central result is that any tensor whose 63 non-trivial sector norms are bounded by 0.02 flows to the fixed point under the normalized 2x1 map, and the free energy is analytic in that neighborhood. Applied to the 2D Ising and XY models, the method proves high-temperature phases for inverse temperatures $eta\le 0.12$ and $\beta\le 0.18995$, respectively.

What carries the argument

The central mechanism is the hat-tensor: a finite-dimensional tensor of nonnegative real numbers that bounds the Hilbert-Schmidt norm of each sector of the infinite-dimensional full tensor. The master function $M$ maps hat-tensors to hat-tensors by exactly mimicking the four steps of the 2x1 map, namely gauge transformation, disentangling and splitting, reconnection, and rotation. Because $M$ is monotonic and subhomogeneous, one rigorous check that an iterate decreases by a factor $\lambda<1$ implies all future iterates decrease geometrically, via the Key Lemma 2.3, reducing infinite-dimensional RG control to a finite numerical computation.

What would settle it

An independent re-implementation of the master function that fails to reproduce the componentwise bound $\hat b^{(15)}\le 0.96\,\hat b^{(14)}$ for $\delta=0.02$, $w_x=2.2$, $w_o=2$ would collapse Theorem 3.1. So would a single tensor in $O_{0.02}$ whose normalized 2x1 iterates fail to converge to zero.

Watch

Extended reading notes

Core claim

The central claim is Theorem 3.1: for reweighting parameters $w_x=2.2$ and $w_o=2$, the set $O_\delta$ with $\delta=0.02$ is a basin of stability, meaning the normalized 2x1 RG map can be iterated forever from any starting tensor in the set and every iterate converges to zero in the Hilbert-Schmidt norm. The proof is computer-assisted via a master function on finite-dimensional hat-tensors; iterating that function in interval arithmetic verifies the componentwise contraction $\hat b^{(15)}\le 0.96\,\hat b^{(14)}$, after which monotonicity and subhomogeneity force exponential decay forever. By Proposition 3.3 the free energy is analytic in the interior of the basin. The same framework yields rigorous high-temperature bounds for the Ising model ($\beta\le 0.12$, versus $\beta_c\approx 0.44$) and the XY model ($\beta\le 0.18995$, versus $\beta_c\approx 1.12$).

Load-bearing premise

The entire proof rests on the computer check being sound: the accompanying code must implement the master function correctly, and the interval arithmetic must truly enclose all rounding errors, so the observed contraction of the bounding box after fifteen iterations is guaranteed for every tensor in the box.

Editorial extensions

If this is right

  • Any tensor within the 0.02 box in all 63 sectors has a well-defined infinite RG trajectory converging to the high-temperature fixed point.
  • The free energy exists and is analytic on the interior of that basin, so all such tensors lie in the same high-temperature phase.
  • The Ising model at inverse temperatures $\beta\le 0.12$ and the XY model at $\beta\le 0.18995$ are rigorously inside their high-temperature phases.
  • The same master-function scheme can be rerun for other models or other reweighting parameters, producing checkable computer-assisted phase proofs.
  • The graphical language converts diagrammatic RG maps into componentwise inequalities, so future maps, including those aimed at critical fixed points, can be controlled in the same way.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An implication the authors leave implicit: replacing the scalar reweighting factors $w_x,w_o$ with sector-resolved matrices, using the same $w\,w^{-1}=1$ identity, should enlarge the provable basin beyond $\delta=0.02$.
  • The structure of the proof suggests a testable extension: running the same master-function iteration on other two-dimensional models, such as Potts or clock models, should yield explicit high-temperature intervals with only a few lines of the provided code.
  • The authors' own sketch for critical fixed points points to a finite-dimensional head plus hat-controlled tail; the hat-tensor machinery here is exactly the control component such a construction would need.
  • Because the method bounds all sectors uniformly, it automatically proves stability against symmetry-breaking perturbations such as magnetic fields, a robustness the paper notes but does not exploit numerically.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper introduces a new anisotropic tensor renormalization group map, the 2x1 map, which coarse-grains a square lattice by a factor of two in one direction and then rotates by 90 degrees. The map is defined on four-leg tensors over infinite-dimensional Hilbert spaces, and the authors develop a graphical calculus that translates the RG steps into inequalities on tensor components. A finite-dimensional "hat-tensor" bounding box is introduced, together with a master function M that controls how this box evolves under RG. The main theorem (Theorem 3.1) states that, for reweighting parameters w_x=2.2 and w_o=2, every tensor whose 63 non-trivial sectors have Hilbert-Schmidt norm at most 0.02 flows to the high-temperature fixed point; consequently the free energy is analytic in the interior of this neighborhood (Proposition 3.3). The method is then applied to obtain explicit high-temperature bounds for the 2D Ising model (β≤0.12) and the XY model (β≤0.18995). The proofs are computer-assisted, using interval arithmetic implemented in a provided Julia notebook.

Significance. If the computer-assisted checks are sound, this is a valuable contribution to the rigorous tensor RG program. The 2x1 map is a genuinely new and relatively simple RG map, and the hat-tensor/master-function framework provides a concrete finite-dimensional control of an infinite-dimensional RG flow, with explicit basin-of-attraction sizes and model-specific bounds. The paper ships documented Julia code and uses interval arithmetic, which are strengths, and the bounds for the Ising and XY models are concrete, falsifiable predictions. The main limitation is that the central computer-assisted verification is not independently reproduced, and some supporting lemmas are stated without full proof.

major comments (3)
  1. [§3.1 and §2.10] The proof of Theorem 3.1 rests on the computer-assisted check that pb^(15) ≤ 0.96 pb^(14) for the master function; this check is performed only by the provided Julia code, and the paper explicitly states in §2.10 that the symbolic linearization (2.99) was not checked by hand. The correctness of the theorem therefore depends on the absence of implementation bugs and on the soundness of the interval arithmetic libraries (ArbNumerics.jl, HCubature.jl), neither of which is independently verified in the manuscript. Because this is the load-bearing step, I ask for an independent verification: for example, a pencil-and-paper derivation of at least the linearized master function (2.99), or a second, independent reimplementation of the full master function and of the check (3.2) in a different language or with a different interval arithmetic library.
  2. [§2.7.6, Proposition 2.6, Proposition 2.8] The master function M is proved monotonic and subhomogeneous by composing Propositions 2.5, 2.6, and 2.8. Proposition 2.6 is only sketched (the key quadrature estimate (2.78) is asserted to preserve subhomogeneity without a detailed proof), and Proposition 2.8 is stated without proof. These properties are essential for the Key Lemma 2.3, which converts the finite check (3.2) into infinite-time convergence. The proofs should be written out in full, at least for the non-trivial quadrature estimate (2.78) and for the normalization divisions by |N_1| and |N_2|.
  3. [Appendix B, Lemma B.3] Lemma B.3, the counting identity |T_n| = 2n^2 + 3n + 1, is stated without proof and is used in Lemma B.4 to bound the tail of the XY tensor norm. This is an elementary counting argument, but since the XY result depends on it, a short proof should be included instead of being omitted.
minor comments (4)
  1. [§2.7.6, Eq. (2.79)] In Eq. (2.79), the notation pR should be pRκ; as written, the equation appears to define a hat-tensor for R rather than for Rκ.
  2. [§2.7.4] The sentence "we represented the oo channel as a contraction of L_o and R_o" uses L_o and R_o where L_oo and R_oo are meant; the subscript is missing.
  3. [References] Reference [29] lists the title "TensorSeries.jl," but the text in Appendix A refers to TaylorSeries.jl; the bibliography entry should be corrected.
  4. [§3.1, Eq. (3.6)] The Kronecker delta δ_{i,i0} in Eq. (3.6) clashes with the radius δ = 0.02 used throughout Section 3; consider renaming one of these quantities to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the main convergence theorem rests on an explicit interval-arithmetic check of the master function, not on a fitted parameter or a self-citation chain.

full rationale

The paper's central claim, Theorem 3.1, is established by iterating the explicitly defined master function M on a bounding box of size delta = 0.02 and verifying the inequality pb^(15) <= 0.96 pb^(14) by interval arithmetic (Section 3.1). This is an independent numerical verification, not a restatement of the definitions: the master function M is constructed from the RG map in Sections 2.4-2.9, and its monotonicity and subhomogeneity are proven in Proposition 2.9 and Lemma 2.3. The reweighting parameters w_x and w_o were tuned to make the verification succeed, but choosing parameters and then checking the resulting contraction condition is standard mathematical practice and does not make the theorem circular. The self-citations to [3,4,6,7] provide technical lemmas, background, and the prior framework; they do not supply the basin-of-stability conclusion, which is proven in the present paper via the master function and the Key Lemma. The paper's admitted lack of a pencil-and-paper check of Eq. (2.99) concerns only the linearized guidance formula used to choose parameters, not the proof of Theorem 3.1, and is a verification/correctness concern rather than circularity. No step in the derivation reduces to its own inputs by construction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the computer-assisted verification and on the specific construction of the 2x1 map. The map involves hand-tuned parameters (w_x, w_o, delta) and relies on standard mathematical background (HS norms, analyticity, Cauchy-Schwarz) plus the correctness of the interval arithmetic implementation. There are no new physical entities.

free parameters (3)
  • w_x = 2.2 (general), 2.3 (Ising and XY)
    Reweighting parameter for ox and xo channels; tuned for maximal basin size while keeping W=2(1/w_x^2+1/w_o^2)<1 (Section 2.10, Section 3.1).
  • w_o = 2.0 (general, Ising), 2.4 (XY)
    Reweighting parameter for the oo channel; tuned jointly with w_x.
  • delta = 0.02
    Per-sector norm bound in the initial box O_delta; chosen by numerical search as the largest value for which the master function iterates start contracting (Section 3.1).
assumptions (4)
  • standard math Hilbert-Schmidt norm inequalities and Cauchy-Schwarz bound contractions of tensors.
    Used in Lemma 2.1 and Lemma 2.2 to propagate hat-tensors through contractions; standard functional analysis.
  • standard math Analyticity of the RG map and N-factor on the neighborhood Omega, based on analytic functions on Banach spaces.
    Needed for Proposition 2.5(b), 2.6, 2.8, 2.9 and for the free energy series; proof sketched in [4] App. A.
  • domain assumption The exact tensor network representations of the Ising and XY models are those given in Eqs. (3.8) and (3.19).
    These representations come from prior literature ([14], [3], [16]) and are not re-derived here; they carry the physical content of the high-T bounds.
  • domain assumption The interval arithmetic libraries and Julia code correctly implement rigorous bounds.
    The theorem's computational part is certified by interval arithmetic; this assumes the software (ArbNumerics.jl, HCubature.jl) and the authors' implementation are correct.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tensor Renormalization Group Meets Computer Assistance." pith.science (2026). https://pith.science/paper/KHOCWMP2

@misc{pith2026250603247,
  author       = {Pith},
  title        = {Pith review of: Tensor Renormalization Group Meets Computer Assistance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KHOCWMP2}},
  note         = {Machine review of arXiv:2506.03247}
}
read the original abstract

Tensor renormalization group, originally devised as a numerical technique, is emerging as a rigorous analytical framework for studying lattice models in statistical physics. Here we introduce a new renormalization map - the 2x1 map - which coarse-grains the lattice anisotropically by a factor of two in one direction followed by a 90-degree rotation. We develop a novel graphical language that translates the action of the 2x1 map into a system of inequalities on tensor components, with rigorous estimates in the Hilbert-Schmidt norm. We define a finite-dimensional "bounding box" called the hat-tensor, and a master function governing its RG flow. Iterating this function numerically, we establish convergence to the high-temperature fixed point for tensors lying within a quantifiable neighborhood. Our main theorem shows that tensors with deviations bounded by 0.02 in 63 orthogonal sectors flow to the fixed point. We also apply the method to specific models - the 2D Ising and XY models - obtaining explicit bounds on their high-temperature phase. This work brings the Tensor RG program closer towards a rigorous, computer-assisted construction of critical fixed points.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Renormalization flows for 1D mixed states and a quantum Goursat lemma

    math-ph 2026-07 accept novelty 7.5 of 10

    Convergent renormalization trajectories of Hopf-algebra boundary MPDOs under on-site noise are classified by finite *-quantum hypergroups via a new quantum Goursat lemma.

  2. Lattice and PT symmetries in tensor-network renormalization group: Case study of a hard-square lattice gas model

    cond-mat.stat-mech 2026-03 conditional novelty 5.0 of 10

    A tensor-network RG scheme that preserves lattice rotation/reflection and PT symmetries is formulated and validated on the hard-square lattice gas, yielding improved critical-point and scaling-dimension estimates.

Reference graph

Works this paper leans on

34 extracted references · 23 canonical work pages · cited by 2 Pith papers

  1. [1]

    Tensor renormalization group approach to 2D classical lattice models,

    M. Levin and C. P. Nave, “Tensor renormalization group approach to 2D classical lattice models,” Phys. Rev. Lett.99no. 12, (2007) 120601,arXiv:cond-mat/0611687

    arXiv 2007

  2. [2]

    Algorithms for tensor network renormalization

    G. Evenbly, “Algorithms for tensor network renormalization,”arXiv:1509.07484 [cond-mat.str-el]

    work page Pith review arXiv

  3. [3]

    Tensor RG Approach to High-Temperature Fixed Point,

    T. Kennedy and S. Rychkov, “Tensor RG Approach to High-Temperature Fixed Point,”J. Statist. Phys.187no. 3, (2022) 33,arXiv:2107.11464 [math-ph]

    arXiv 2022

  4. [4]

    Tensor Renormalization Group at Low Temperatures: Discontinuity Fixed Point,

    T. Kennedy and S. Rychkov, “Tensor Renormalization Group at Low Temperatures: Discontinuity Fixed Point,”Annales Henri Poincare25no. 1, (2024) 773–841,arXiv:2210.06669 [math-ph]

    arXiv 2024

  5. [5]

    3D Tensor Renormalisation Group at High Temperatures,

    N. Ebel, “3D Tensor Renormalisation Group at High Temperatures,”Annales Henri Poincare(July,

    work page

  6. [6]

    Rotations, Negative Eigenvalues, and Newton Method in Tensor Network Renormalization Group,

    N. Ebel, T. Kennedy, and S. Rychkov, “Rotations, Negative Eigenvalues, and Newton Method in Tensor Network Renormalization Group,”arXiv:2408.10312 [cond-mat.stat-mech]. – 44 –

    arXiv

  7. [7]

    Transfer Matrix and Lattice Dilatation Operator for High-Quality Fixed Points in Tensor Network Renormalization Group,

    N. Ebel, T. Kennedy, and S. Rychkov, “Transfer Matrix and Lattice Dilatation Operator for High-Quality Fixed Points in Tensor Network Renormalization Group,”arXiv:2409.13012 [cond-mat.stat-mech]. [8]https://github.com/ebelnikola/2x1

    arXiv

  8. [9]

    Tensor Network Renormalization,

    G. Evenbly and G. Vidal, “Tensor Network Renormalization,”Phys. Rev. Lett.115no. 18, (2015) 180405,arXiv:1412.0732 [cond-mat.str-el]

    arXiv 2015

Show all 34 references
  1. [10]

    Entanglement Renormalization,

    G. Vidal, “Entanglement Renormalization,”Phys. Rev. Lett.99no. 22, (2007) 220405, arXiv:cond-mat/0512165

    2007 arXiv

  2. [11]

    Ruelle,Statistical Mechanics: Rigorous Results

    D. Ruelle,Statistical Mechanics: Rigorous Results. World Scientific, 1999

    1999

  3. [12]

    Friedli and Y

    S. Friedli and Y. Velenik,Statistical Mechanics of Lattice Systems: A Concrete Mathematical Introduction. Cambridge University Press, 2017

    2017

  4. [13]

    Cluster expansion for abstract polymer models,

    R. Koteck´ y and D. Preiss, “Cluster expansion for abstract polymer models,”Commun. Math. Phys. no. 103, (1986) 491–498

    1986

  5. [14]

    Tensor-entanglement-filtering renormalization approach and symmetry-protected topological order,

    Z.-C. Gu and X.-G. Wen, “Tensor-entanglement-filtering renormalization approach and symmetry-protected topological order,”Phys. Rev. B80no. 15, (2009) 155131,arXiv:0903.1069 [cond-mat.str-el]

    2009 arXiv

  6. [15]

    Simon,The Statistical Mechanics of Lattice Gases, Volume I

    B. Simon,The Statistical Mechanics of Lattice Gases, Volume I. 2014

    2014

  7. [16]

    NIST Digital Library of Mathematical Functions

    “NIST Digital Library of Mathematical Functions.”https://dlmf.nist.gov/, release 1.2.4 of 2025-03-15. F. W. J. Olver, A. B. Olde Daalhuis, D. W. Lozier, B. I. Schneider, R. F. Boisvert, C. W. Clark, B. R. Miller, B. V. Saunders, H. S. Cohl, and M. A. McClain, eds

    2025

  8. [17]

    Tensor renormalization group study of classicalXYmodel on the square lattice,

    J. F. Yu, Z. Y. Xie, Y. Meurice, Y. Liu, A. Denbleyker, H. Zou, M. P. Qin, J. Chen, and T. Xiang, “Tensor renormalization group study of classicalXYmodel on the square lattice,”Phys. Rev. E89 (Jan, 2014) 013308,arXiv:1309.4963 [cond-mat.stat-mech]

    2014 arXiv

  9. [18]

    Critical analysis of two-dimensional classical XY model,

    R. G. Jha, “Critical analysis of two-dimensional classical XY model,”Journal of Statistical Mechanics: Theory and Experiment2020no. 8, (Aug, 2020) 083203,arXiv:2004.06314 [hep-lat]

    2020 arXiv

  10. [19]

    Resolving the Berezinskii-Kosterlitz-Thouless transition in the two-dimensional XY model with tensor-network-based level spectroscopy,

    A. Ueda and M. Oshikawa, “Resolving the Berezinskii-Kosterlitz-Thouless transition in the two-dimensional XY model with tensor-network-based level spectroscopy,”Phys. Rev. B104no. 16, (Oct., 2021) 165132,arXiv:2105.11460 [cond-mat.stat-mech]

    2021 arXiv

  11. [20]

    A comparison of plane rotor and Ising models,

    M. Aizenman and B. Simon, “A comparison of plane rotor and Ising models,”Physics Letters A76 no. 3, (1980) 281–282

    1980

  12. [21]

    Completely Analytical Gibbs Fields,

    R. L. Dobrushin and S. B. Shlosman, “Completely Analytical Gibbs Fields,” inStatistical Physics and Dynamical Systems: Rigorous Results, Y. M. Sinai, ed., pp. 371–403. Birkh¨ auser Boston, Boston, MA, 1985

    1985

  13. [22]

    Phase Diagram of the Two-Dimensional Ising Antiferromagnet (Computer-Assisted Proof),

    R. L. Dobrushin, J. Kolafa, and S. B. Shlosman, “Phase Diagram of the Two-Dimensional Ising Antiferromagnet (Computer-Assisted Proof),”Communications in Mathematical Physics102(1985) 89–103

    1985

  14. [23]

    Completely Analytical Interactions: Constructive Description,

    R. L. Dobrushin and S. B. Shlosman, “Completely Analytical Interactions: Constructive Description,”Journal of Statistical Physics46no. 5–6, (1987) 983–1014

    1987

  15. [24]

    A Fixed-Point Equation for the High-Temperature Phase of Discrete Lattice Spin Systems,

    T. Kennedy, “A Fixed-Point Equation for the High-Temperature Phase of Discrete Lattice Spin Systems,”Journal of Statistical Physics59no. 1–2, (1990) 1–29

    1990

  16. [25]

    Julia: A fresh approach to numerical computing,

    J. Bezanson, A. Edelman, S. Karpinski, and V. B. Shah, “Julia: A fresh approach to numerical computing,”SIAM Review59no. 1, (2017) 65–98

    2017

  17. [26]

    “Julia.”https://julialang.org/

  18. [27]

    TensorOperations.jl,

    L. Devos, M. Van Damme, J. Haegeman, and contributors, “TensorOperations.jl,” Oct, 2023. https://github.com/Jutho/TensorOperations.jl. – 45 –

    2023

  19. [28]

    High-performance symbolic-numerics via multiple dispatch,

    S. Gowda, Y. Ma, A. Cheli, M. Gwozdz, V. B. Shah, A. Edelman, and C. Rackauckas, “High-performance symbolic-numerics via multiple dispatch,”arXiv:2105.03949 [cs.CL]

    arXiv

  20. [29]

    TensorSeries.jl,

    L. Benet and D. P. Sanders, “TensorSeries.jl,” Apr, 2025. https://github.com/JuliaDiff/TaylorSeries.jl

    2025

  21. [30]

    ArbNumerics.jl

    J. Sarnoff and contributors, “ArbNumerics.jl.” https://github.com/JeffreySarnoff/ArbNumerics.jl

  22. [31]

    SpecialFunctions.jl

    “SpecialFunctions.jl.”https://github.com/JuliaMath/SpecialFunctions.jl

  23. [32]

    HCubature.jl

    S. G. Johnson, “HCubature.jl.”https://github.com/JuliaMath/HCubature.jl, 2017

    2017

  24. [33]

    Remarks on algorithm 006: An adaptive algorithm for numerical integration over ann-dimensional rectangular region,

    A. C. Genz and A. A. Malik, “Remarks on algorithm 006: An adaptive algorithm for numerical integration over ann-dimensional rectangular region,”Journal of Computational and Applied Mathematics6(1980) 295–302

    1980

  25. [34]

    Makie.jl: Flexible high-performance data visualization for Julia,

    S. Danisch and J. Krumbiegel, “Makie.jl: Flexible high-performance data visualization for Julia,” Journal of Open Source Software6no. 65, (2021) 3349. – 46 –

    2021

  26. [2024]

    ,arXiv:2401.04229 [cond-mat.stat-mech]

    arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.