Pith. sign in

REVIEW 3 major objections 5 minor 59 references

Superconductivity in the $t$-$t'$ Hubbard Model from Symmetry-Preserving Neural-Network Quantum States

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A neural-network wave function that preserves translational symmetry finds d-wave superconducting order in the 1/8-doped t-t' Hubbard model, with a thermodynamic-limit order parameter of 0.022(1).

desk verdict A genuinely new symmetry-preserving neural-network ansatz with a plausible d-wave superconductivity claim, but the thermodynamic-limit evidence is thinner than the abstract suggests. read the letter →

arxiv 2608.12465 v1 pith:WYN3QWSB submitted 2026-08-12 cond-mat.str-el cond-mat.dis-nncond-mat.supr-conquant-ph

Riccardo Rende , Luciano Loris Viteritti , Antoine Georges This is my paper · ORCID
classification cond-mat.str-elcond-mat.dis-nncond-mat.supr-conquant-ph
keywords Hubbardmodeld-wavesuperconductivityneuralquantumstatesbackflowtranslationalsymmetrystripeordervariationalMonteCarlostronglycorrelatedelectrons
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 tries to settle a long-standing question: at 1/8 hole doping, with next-nearest-neighbor hopping $t'/t=-0.2$ and on-site repulsion $U/t=8.0$, does the two-dimensional Hubbard model harbor superconductivity? Its answer is yes. Using a neural-network wave function that keeps translational symmetry by construction, the authors reach lower variational energies than competing pure-stripe states on square lattices up to $24\times24$ (504 electrons), and their $d$-wave pairing correlation function saturates at long distances. Extrapolating the resulting order parameter to the thermodynamic limit gives $\Delta_\mathrm{SC}=0.022(1)$, consistent with an earlier constrained-path auxiliary-field quantum Monte Carlo estimate and with the picture of superconductivity coexisting with stripe correlations. The methodological point is that broken-symmetry variational states fall into stripe minima that suppress pairing, so a symmetry-preserving ansatz is needed to expose the superconducting ground state.

What carries the argument

The load-bearing object is the Symmetry-Preserving Backflow Pairing (SBP) ansatz: a neural-network pairing matrix $f_{ij}(n)$ whose weights depend only on the relative displacement $i-j$ and on configuration-dependent backflow vectors $y^\sigma_i(n)$ produced by a translationally equivariant transformer. The matrix element is multiplied by a normalized exponential distance factor $e^{-\lambda d(i,j)}/\sum_{i'j'}e^{-\lambda d(i',j')}$, suppressing pairing between distant sites. Because only up-down pairs are kept, the wave function is a determinant, $\Psi_\theta(n)=\det[n_\uparrow \star f(n)\star n_\downarrow]$, rather than a Pfaffian. Enforcing translational symmetry in the parametrization removes pure-stripe broken-symmetry states from the variational manifold, making the optimization landscape effectively convex so different random initializations converge to the same physical state. A configuration-independent $d_{x^2-y^2}$ nearest-neighbor seed accelerates the optimization but is checked not to alter the final result.

What would settle it

Compute $C_p(|r|)$ for the same parameters on larger periodic clusters or with a pairing kernel that is not forced to decay with Euclidean distance; if the saturated plateau moves toward zero with system size instead of converging to about $\Delta_\mathrm{SC}\approx 0.022$, the claimed thermodynamic-limit order is a variational artifact.

Watch

Extended reading notes

Core claim

The central claim is that the ground state of the $t$-$t'$ Hubbard model at $t'/t=-0.2$, $U/t=8.0$, and doping $\delta=1/8$ has genuine $d_{x^2-y^2}$ superconducting order in the thermodynamic limit. The evidence is the long-distance behavior of the $d$-wave pairing correlation function $C_p(|r|)$ on periodic clusters of linear size $L=12,16,20,24$: after a fast short-distance decay, $C_p(|r|)$ saturates to a finite plateau. Averaging the plateau over $|r|\ge d_\mathrm{max}/2$ defines the order parameter $\Delta_\mathrm{SC}$, and its extrapolation $1/L\to0$ gives $\Delta_\mathrm{SC}=0.022(1)$, matching a constrained-path auxiliary-field quantum Monte Carlo value obtained on cylinders with a pinning field. No pinning field is needed in the present calculation, and the order parameter is read directly from correlation functions on periodic square clusters. In the same state the spin correlations show an antiferromagnetic pattern modulated by a longer-wavelength stripe envelope, so superconductivity and stripe correlations coexist at the level of fluctuations while the one-body density remains uniform.

Load-bearing premise

That the finite-size extrapolation from four periodic clusters, together with the variational bias from the built-in distance decay and the seeded $d$-wave channel, does not itself manufacture the long-range pairing order it reports.

Editorial extensions

If this is right

  • The contradiction between DMRG studies that find no superconductivity for $t'<0$ and neural-network studies that do is resolved, for this parameter point, in favor of $d$-wave superconductivity in the thermodynamic limit.
  • Stripe order and superconductivity are not mutually exclusive: the SBP ground state has stripe-modulated antiferromagnetic spin correlations, yet its one-body density is translationally uniform and it carries a finite $d$-wave order parameter.
  • Variational approaches that first break translational symmetry and later restore it may systematically underestimate pairing because they get trapped in pure-stripe local minima; enforcing the symmetry in the ansatz itself bypasses this bias.
  • Because the optimized SBP wave function is translationally invariant, it can be transferred from one lattice size to the next, which is what makes the 24x24, 504-electron calculation and the four-point $1/L$ extrapolation practical.

Reading between the lines

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

  • Our inference, not the paper's claim: a decisive test would be the same SBP extrapolation at $t'=0$; if the order parameter vanished there, it would confirm that negative $t'$ is what stabilizes pairing in this regime.
  • Our inference: the built-in Euclidean-distance decay of the pairing kernel is the main hidden modeling choice; repeating the calculation with algebraic or learnable long-range kernels would show whether the saturated $C_p(|r|)$ is robust or an artifact of the kernel.
  • Our inference: the 'effectively convex' landscape suggests a general design principle for competing orders: imposing the Hamiltonian's exact symmetries on the variational form may be more reliable than optimizing a broken-symmetry state and projecting afterward.
  • Our inference: the same symmetry-preserving construction should transfer to multi-orbital or frustrated fermion models, where stripe-type local minima are also expected; the paper mentions this as future work but does not test it.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript introduces the Symmetry-Preserving Backflow Pairing (SBP) ansatz, a neural-network variational wave function for the square-lattice t-t' Hubbard model that enforces translational symmetry by construction. The authors show that an unconstrained backflow-pairing ansatz falls into multiple stripe-ordered local minima, while SBP converges reproducibly and reaches lower variational energies on L=12,...,24 at t'/t=-0.2, U/t=8, δ=1/8. Using the long-distance plateau of the d-wave pairing correlation function, they extract a finite order parameter Δ_SC and extrapolate it to Δ_SC=0.022(1), consistent with a constrained-path AFQMC estimate. A half-filling QMC benchmark shows that the same ansatz produces no spurious d-wave order in the undoped case.

Significance. The result, if correct, would settle a longstanding controversy for this parameter point and support d-wave superconductivity coexisting with stripe correlations in the 2D Hubbard model. The methodological contribution is substantial: a symmetry-preserving, scalable fermionic NQS that reaches 24x24 with 504 electrons, with transfer learning across sizes, and a clean half-filling control demonstrating that the ansatz can yield vanishing pairing correlations. The agreement with AFQMC for Δ_SC provides genuine external support. The main weakness is the underdocumented thermodynamic-limit extrapolation and the untested seed-independence claim; these are local enough to be fixable but directly bear on the headline number.

major comments (3)
  1. [Inset of Fig. 4 and Eq. (8)] The central claim Δ_SC=0.022(1) is an extrapolation from four system sizes, but the paper does not state the fit function, the uncertainties on the per-size Δ_SC values, the goodness of fit, or the error bar on the intercept. Please report the fitting procedure, show the stability of the intercept under excluding L=12 or including a quadratic correction in 1/L, and state how the fit uncertainty is propagated. Without this, the 'robust evidence' claim is stronger than the displayed data support.
  2. [Methods D and Eq. (4)] The text asserts that the d-wave seed f_d only accelerates learning and that final results are independent of it, but no comparison is shown. Because the measured observable is the d-wave pairing correlation in the same symmetry channel as the seed, this is load-bearing. Please provide a direct comparison of Cp(r), Δ_SC, and variational energy with and without seed (or with different Δ_d values), and report the optimized value of λ and the sensitivity of the plateau to λ. The half-filling benchmark is reassuring but does not control the doped case.
  3. [Eq. (11) and Fig. 4] The order parameter uses only one of the four spin-resolved singlet correlators and one disconnected contraction, with the justification that the omitted terms coincide asymptotically. The finite-size Δ_SC, however, averages over distances that are not necessarily asymptotic (|r|≥d_max/2). Please quantify the finite-distance error, for example by computing the full expression (10) on 12×12, and by showing that Cp(r) is flat within the averaging shell rather than still decaying. This is a finite-size systematic that should be controlled before the extrapolation is trusted.
minor comments (5)
  1. [Section I] There is a typo '1/8-doped model Hubbard model'; the duplicated word 'model' should be removed.
  2. [Fig. 2 and Section III] The 'effectively convex' landscape claim rests on only two random seeds for SBP and BP; reporting a few more independent initializations would considerably strengthen the claim.
  3. [Table I] For the 20×20 and 24×24 lattices, no previous variational energies are listed; the 'state-of-the-art' claim for those sizes should be qualified or contextualized with the closest available results in the literature.
  4. [Inset of Fig. 4] The inset would be much easier to evaluate if the extrapolation curve and error bars on each Δ_SC point were drawn; currently only markers are shown.
  5. [Eq. (8)] The quantity d_max is not defined precisely; on a periodic square lattice the minimum-image convention for Euclidean distances should be stated explicitly.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor self-citation for the stripe-suppresses-superconductivity mechanism; the central d-wave extrapolation is otherwise self-contained.

  1. other [Section III, paragraph after Fig. 3 (discussion of BP broken-symmetry solutions)]
    "Crucially, as shown in Ref. [21], these broken-symmetry solutions typically display weak or no superconductivity."

    This sentence imports from the authors' own companion paper [21] the premise that pure stripe solutions suppress superconductivity, which is then used to motivate why a symmetry-preserving ansatz is necessary to expose superconducting order. The premise is not demonstrated from data in the present paper, so it is a genuine but minor self-citation. It is not load-bearing for the main numerical result: the extrapolated Delta_SC = 0.022(1) is computed directly from the SBP wave function and is benchmarked against half-filling QMC data and the independent AFQMC result of Ref. [6].

full rationale

The central derivation is self-contained: the SBP wave function in Eq. (2)-(4) is a variational ansatz with trainable parameters, and the d-wave order parameter in Eq. (8) is a measured expectation value, not a fitted parameter. The exponential prefactor in Eq. (4) is physically motivated as a locality prior, and the paper explicitly notes that a local pairing amplitude does not imply short-range pairing correlations. The half-filling QMC comparison (Methods A) provides a strong control showing the same ansatz yields vanishing d-wave pairing correlations when superconductivity is absent, so the d-wave plateau at doping is not forced by the ansatz alone. The d-wave seed in Methods D is stated to be numerically checked to affect only learning speed and not final results; it is also a trainable variational parameter rather than a fixed input. The finite-size extrapolation is compared with, not fitted to, the independent AFQMC result of Ref. [6]. The only circularity concern is the self-citation to Ref. [21] for the broken-symmetry, stripe-suppresses-superconductivity mechanism, which is used to motivate the methodological design but is not the source of the main predicted order parameter. Thus the circularity score is low.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the validity of variational Monte Carlo, on the expressivity and locality bias of the SBP ansatz, and on a linear finite-size extrapolation of the order parameter. No physical constants are fitted, but analysis choices (lambda, Delta_d, r_max) shape the result. No new physical entities are introduced.

free parameters (3)
  • Pairing decay length lambda = trainable, not reported
    Eq. (4) introduces a trainable spatial decay scale in the pairing matrix; it shapes the variational manifold and could affect whether long-range pairing is captured.
  • d-wave seed amplitude Delta_d = trainable, not reported
    Methods: a configuration-independent d-wave pairing matrix f_d with trainable scalar Delta_d is added to seed the superconducting channel; authors claim final results are independent, but no comparison is shown.
  • Order parameter cutoff r_max = d_max/2
    Delta_SC is defined as an average of C_p over |r| >= r_max with r_max = d_max/2; this choice affects the finite-size estimate and the extrapolation.
assumptions (6)
  • domain assumption Variational Monte Carlo sampling gives unbiased estimates for the optimized wave function.
    The optimization and all correlation functions are evaluated with VMC using 12288 samples; the paper does not discuss autocorrelation or sampling bias.
  • ad hoc to paper The pairing amplitude f_ij is local in Euclidean distance.
    Eq. (4) imposes an exponential decay in the pairing matrix, motivated by locality but not proven to be a sufficient variational restriction.
  • domain assumption Finite-size scaling of Delta_SC is linear in 1/L.
    The inset of Fig. 4 extrapolates four system sizes to 1/L -> 0 without describing the fitting function or checking alternative scaling forms.
  • domain assumption A translationally symmetric ansatz can capture the essential ground-state physics, including stripe correlations.
    The SBP state cannot break translational symmetry, so stripe order is accessible only in correlation functions, not in one-body observables.
  • domain assumption C4v projection after optimization is adequate, even when not followed by further optimization.
    For 20x20 and 24x24 clusters, only sampling is performed after symmetry restoration, without re-optimization.
  • domain assumption The transformer architecture with 8 layers, 12 heads, and dimension 72 is expressive enough for the relevant ground-state correlations.
    Expressivity is asserted by construction and benchmarked only at half filling; no convergence test versus architecture size is shown.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Superconductivity in the $t$-$t'$ Hubbard Model from Symmetry-Preserving Neural-Network Quantum States." pith.science (2026). https://pith.science/paper/WYN3QWSB

@misc{pith2026260812465,
  author       = {Pith},
  title        = {Pith review of: Superconductivity in the $t$-$t'$ Hubbard Model from Symmetry-Preserving Neural-Network Quantum States},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WYN3QWSB}},
  note         = {Machine review of arXiv:2608.12465}
}
abstract

Despite its fundamental importance in the theory of strongly correlated electrons, the nature of the ground state of the two-dimensional doped Hubbard model remains intensely debated. Variational approaches provide a powerful route to this problem, but their conclusions can depend sensitively on the chosen wave-function parameterization, the mean-field initialization, or the pinning fields used to guide the optimization, as well as on boundary conditions. This can favor one type of symmetry breaking over another, making it difficult to distinguish the genuine interplay of intertwined or competing orders from biases induced by the variational parameterization. Here, we introduce the Symmetry-Preserving Backflow Pairing (SBP) ansatz, a neural-network wave function that respects translational symmetry by construction and thereby avoids these broken-symmetry minima. The SBP ansatz reaches state-of-the-art variational energies for the $t$-$t'$ Hubbard model on lattices up to $24\times24$ with $504$ electrons, below those of competing pure stripe solutions. By extrapolating to the thermodynamic limit, we find robust evidence for $d$-wave superconducting order, resolving a long-standing question about the $1/8$-doped model at $t'/t=-0.2$ and $U/t=8.0$. Built on general principles of symmetry and locality, the SBP wave function provides a broadly applicable variational representation for challenging interacting fermionic systems.

Figures

Figures reproduced from arXiv: 2608.12465 by the authors.

Figure 1
Figure 1. FIG. 1. Pictorial representation of the variational energy [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Optimization of the SBP wave function [see Eq. ( [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Local charge density [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Benchmark of the SBP ansatz against numerically exact quantum Monte Carlo (QMC) for the pure Hubbard model [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Real space spin-spin correlation function [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 22 canonical work pages

  1. [21]

    L. L. Viteritti, R. Rende, C. Roth, A. Sengupta, G. Car- leo, and A. Georges, Beyond Variational Bias: Resolv- ing Intertwined Orders in the Hubbard Model (2026), arXiv:2604.21978 [cond-mat.str-el]

    arXiv 2026

  2. [1]

    Hubbard, Electron correlations in narrow energy bands, Proceedings of the Royal Society of Lon- don

    J. Hubbard, Electron correlations in narrow energy bands, Proceedings of the Royal Society of Lon- don. A. Mathematical and Physical Sciences276, 238 (1963), https://royalsocietypublishing.org/rspa/article- pdf/276/1365/238/54456/rspa.1963.0204.pdf

    arXiv 1963

  3. [2]

    D. P. Arovas, E. Berg, S. A. Kivelson, and S. Raghu, The hubbard model, Annual Review of Condensed Mat- ter Physics13, 239–274 (2022)

    work page 2022

  4. [3]

    M. Qin, T. Sch¨ afer, S. Andergassen, P. Corboz, and E. Gull, The hubbard model: A computational perspec- tive, Annual Review of Condensed Matter Physics13, 275–302 (2022)

    2022

  5. [4]

    J. P. F. LeBlanc, A. E. Antipov, F. Becca, I. W. Bulik, G. K.-L. Chan, C.-M. Chung, Y. Deng, M. Ferrero, T. M. Henderson, C. A. Jim´ enez-Hoyos, E. Kozik, X.-W. Liu, A. J. Millis, N. V. Prokof’ev, M. Qin, G. E. Scuseria, H. Shi, B. V. Svistunov, L. F. Tocchio, I. S. Tupitsyn, S. R. White, S. Zhang, B.-X. Zheng, Z. Zhu, and E. Gull (Simons Collaboration on...

    2015

  6. [5]

    Fradkin, S

    E. Fradkin, S. A. Kivelson, and J. M. Tranquada, Collo- quium: Theory of intertwined orders in high temperature superconductors, Rev. Mod. Phys.87, 457 (2015)

    2015

  7. [6]

    Xu, C.-M

    H. Xu, C.-M. Chung, M. Qin, U. Schollw¨ ock, S. R. White, and S. Zhang, Coexistence of su- perconductivity with partially filled stripes in the hubbard model, Science384, eadh7691 (2024), https://www.science.org/doi/pdf/10.1126/science.adh7691

    doi:10.1126/science.adh7691 2024

  8. [7]

    C. Roth, A. Chen, A. Sengupta, and A. Georges, Superconductivity in the two-dimensional hubbard model revealed by neural quantum states (2025), arXiv:2511.07566 [cond-mat.supr-con]

    arXiv 2025

Show all 59 references
  1. [8]

    S. R. White, D. J. Scalapino, R. L. Sugar, E. Y. Loh, J. E. Gubernatis, and R. T. Scalettar, Numerical study of the two-dimensional hubbard model, Phys. Rev. B40, 506 (1989)

    1989

  2. [9]

    L. F. Tocchio, F. Becca, and S. Sorella, Hidden mott transition and large-usuperconductivity in the two- dimensional hubbard model, Phys. Rev. B94, 195126 (2016). 9

    2016

  3. [10]

    M. Qin, H. Shi, and S. Zhang, Benchmark study of the two-dimensional hubbard model with auxiliary-field quantum monte carlo method, Phys. Rev. B94, 085103 (2016)

    2016

  4. [11]

    ˇSimkovic, R

    F. ˇSimkovic, R. Rossi, A. Georges, and M. Ferrero, Origin and fate of the pseudogap in the doped hubbard model, Science385, 10.1126/science.ade9194 (2024)

    2024 doi

  5. [12]

    Sorella, Systematically improvable mean-field varia- tional ansatz for strongly correlated systems: Applica- tion to the hubbard model, Phys

    S. Sorella, Systematically improvable mean-field varia- tional ansatz for strongly correlated systems: Applica- tion to the hubbard model, Phys. Rev. B107, 115133 (2023)

    2023

  6. [13]

    Y. Gu, W. Li, H. Lin, B. Zhan, R. Li, Y. Huang, D. He, Y. Wu, T. Xiang, M. Qin, L. Wang, and D. Lv, Solving the hubbard model with neural quantum states (2025), arXiv:2507.02644 [cond-mat.str-el]

    2025 arXiv

  7. [14]

    Marino, F

    V. Marino, F. Becca, and L. F. Tocchio, Stripes in the extendedt−t ′ hubbard model: A variational monte carlo analysis, SciPost Physics12, 10.21468/scipost- phys.12.6.180 (2022)

    2022 doi

  8. [15]

    E. W. Huang, C. B. Mendl, H.-C. Jiang, B. Moritz, and T. P. Devereaux, Stripe order from the perspec- tive of the hubbard model, npj Quantum Materials3, 10.1038/s41535-018-0097-0 (2018)

    2018 doi

  9. [16]

    Jiang and T

    H.-C. Jiang and T. P. Devereaux, Superconductivity in the doped hubbard model and its interplay with next- nearest hopping t’, Science365, 1424 (2019)

    2019

  10. [17]

    Jiang, T

    H.-C. Jiang, T. P. Devereaux, and S. A. Kivelson, Com- petition between charge-density-wave and superconduct- ing orders on eight-leg square hubbard cylinders, npj Quantum Materials https://doi.org/10.1038/s41535-026- 00910-7 (2026), Arxiv:2511.18644

    2026

  11. [18]

    Carleo and M

    G. Carleo and M. Troyer, Solving the quantum many- body problem with artificial neural networks, Science 355, 602–606 (2017)

    2017

  12. [19]

    Y. Gu, Z. Han, W. Li, Z. Xiao, T. Xiang, M. Qin, L. Wang, and D. Lv, Pareto frontier of neural quantum states: Scalable, affordable, and accurate convolutional backflow for strongly correlated lattice fermions (2026), arXiv:2604.25775 [cond-mat.str-el]

    2026 arXiv

  13. [20]

    W.-Y. Liu, H. Zhai, R. Peng, Z.-C. Gu, and G. K.-L. Chan, Accurate simulation of the hubbard model with fi- nite fermionic projected entangled pair states, Phys. Rev. Lett.134, 256502 (2025)

    2025

  14. [22]

    Andersen, A

    O. Andersen, A. Liechtenstein, O. Jepsen, and F. Paulsen, Lda energy bands, low-energy hamiltonians, t′,t′′,t perp(k), andj perp, Journal of Physics and Chem- istry of Solids56, 1573 (1995), proceedings of the Con- ference on Spectroscopies in Novel Superconductors

    1995

  15. [23]

    Hirayama, Y

    M. Hirayama, Y. Yamaji, T. Misawa, and M. Imada, Ab initio effective hamiltonians for cuprate superconductors, Phys. Rev. B98, 134501 (2018)

    2018

  16. [24]

    M. T. Schmid, J.-B. Mor´ ee, R. Kaneko, Y. Yamaji, and M. Imada, Superconductivity studied by solving ab ini- tio low-energy effective hamiltonians for carrier doped cacuo2, bi 2sr2cuo6, bi 2sr2cacu2o8, and hgba 2cuo4, Phys. Rev. X13, 041036 (2023)

    2023

  17. [25]

    Luo and B

    D. Luo and B. K. Clark, Backflow transformations via neural networks for quantum many-body wave functions, Phys. Rev. Lett.122, 226401 (2019)

    2019

  18. [26]

    Zhou, Z.-W

    Y.-T. Zhou, Z.-W. Zhou, and X. Liang, Solving fermi- hubbard-type models by tensor representations of back- flow corrections, Phys. Rev. B109, 245107 (2024)

    2024

  19. [27]

    J. R. Moreno, G. Carleo, A. Georges, and J. Stokes, Fermionic wave functions from neural-network con- strained hidden states, Proceedings of the National Academy of Sciences119, e2122059119 (2022), https://www.pnas.org/doi/pdf/10.1073/pnas.2122059119

    2022 doi

  20. [28]

    Sharma, A

    L. Sharma, A. Shokry, R. Nutakki, O. Simard, M. Fer- rero, and F. Vicentini, Comparing symmetrized determi- nant neural quantum states for the hubbard model, Phys. Rev. B113, 245104 (2026)

    2026

  21. [29]

    Zhou, Z.-W

    Y.-T. Zhou, Z.-W. Zhou, and W.-Y. Liu, Locality- induced hierarchical backflow wavefunctions for corre- lated fermions (2026), arXiv:2606.00924 [cond-mat.str- el]

    2026 arXiv

  22. [30]

    Liang, Investigating the fermi-hubbard model by the tensor-backflow method, Physical Review B113, 10.1103/dqc7-zsvc (2026)

    X. Liang, Investigating the fermi-hubbard model by the tensor-backflow method, Physical Review B113, 10.1103/dqc7-zsvc (2026)

    2026 doi

  23. [31]

    K. Ido, T. Ohgoe, and M. Imada, Competition among various charge-inhomogeneous states and d-wave su- perconducting state in hubbard models on square lat- tices, Physical Review B97, 10.1103/physrevb.97.045138 (2018)

    2018 doi

  24. [32]

    A. S. Darmawan, Y. Nomura, Y. Yamaji, and M. Imada, Stripe and superconducting order competing in the hub- bard model on a square lattice studied by a combined variational monte carlo and tensor network method, Phys. Rev. B98, 205132 (2018)

    2018

  25. [33]

    Zheng, C.-M

    B.-X. Zheng, C.-M. Chung, P. Corboz, G. Ehlers, M.- P. Qin, R. M. Noack, H. Shi, S. R. White, S. Zhang, and G. K.-L. Chan, Stripe order in the underdoped re- gion of the two-dimensional hubbard model, Science358, 1155–1160 (2017)

    2017

  26. [34]

    Wigner, W

    E. Wigner, W. P´ al, and J. Griffin,Group Theory: Its Ap- plication to the Quantum Mechanics of Atomic Spectra, Pure and applied physics a series of monographs and text books (Academic Press, 1959)

    1959

  27. [35]

    M. Reh, M. Schmitt, and M. G¨ arttner, Optimizing design choices for neural quantum states, Phys. Rev. B107, 195115 (2023)

    2023

  28. [36]

    Tahara and M

    D. Tahara and M. Imada, Variational monte carlo method combined with quantum-number projection and multi-variable optimization, Journal of the Physical So- ciety of Japan77, 114701 (2008)

    2008

  29. [37]

    Nomura and M

    Y. Nomura and M. Imada, Dirac-type nodal spin liq- uid revealed by refined quantum many-body solver using neural-network wave function, correlation ratio, and level spectroscopy, Phys. Rev. X11, 031034 (2021)

    2021

  30. [38]

    C. Liu, L. Zhu, and M. Belkin, Loss landscapes and op- timization in over-parameterized non-linear systems and neural networks, Applied and Computational Harmonic Analysis59, 85 (2022), special Issue on Harmonic Anal- ysis and Machine Learning

    2022

  31. [39]

    Garipov, P

    T. Garipov, P. Izmailov, D. Podoprikhin, D. P. Vetrov, and A. G. Wilson, Loss surfaces, mode connectivity, and fast ensembling of dnns, Advances in neural information processing systems31(2018)

    2018

  32. [40]

    Draxler, K

    F. Draxler, K. Veschgini, M. Salmhofer, and F. Ham- precht, Essentially no barriers in neural network en- ergy landscape, inProceedings of the 35th International Conference on Machine Learning, Proceedings of Ma- chine Learning Research, Vol. 80, edited by J. Dy and A. Krause ...

    2018

  33. [41]

    Nomura, A

    Y. Nomura, A. S. Darmawan, Y. Yamaji, and M. Imada, 10 Restricted boltzmann machine learning for solving strongly correlated quantum systems, Phys. Rev. B96, 205152 (2017)

    2017

  34. [42]

    Bouchaud, A

    J. Bouchaud, A. Georges, and C. Lhuillier, Pair wave functions for strongly correlated fermions and their de- terminantal representation, Journal de Physique49, 553 (1988)

    1988

  35. [43]

    Loehr and B

    K. Loehr and B. K. Clark, Enhancing neural network backflow (2025), arXiv:2510.26906 [cond-mat.str-el]

    2025

  36. [44]

    Chen, Z.-Q

    A. Chen, Z.-Q. Wan, A. Sengupta, A. Georges, and C. Roth, Neural network-augmented pfaffian wave- functions for scalable simulations of interacting fermions (2025), arXiv:2507.10705 [cond-mat.str-el]

    2025 arXiv

  37. [45]

    L. L. Viteritti, R. Rende, and F. Becca, Transformer vari- ational wave functions for frustrated quantum spin sys- tems, Phys. Rev. Lett.130, 236401 (2023)

    2023

  38. [46]

    L. L. Viteritti, R. Rende, A. Parola, S. Goldt, and F. Becca, Transformer wave function for two dimensional frustrated magnets: Emergence of a spin-liquid phase in the Shastry-Sutherland model, Phys. Rev. B111, 134411 (2025)

    2025

  39. [47]

    Rende, L

    R. Rende, L. L. Viteritti, L. Bardone, F. Becca, and S. Goldt, A simple linear algebra identity to optimize large-scale neural network quantum states, Communica- tions Physics7, 260 (2024)

    2024

  40. [48]

    L. L. Viteritti, R. Rende, S. Sachdev, and G. Carleo, Approaching the Thermodynamic Limit with Neural- Network Quantum States (2026), arXiv:2602.02665 [cond-mat.str-el]

    2026

  41. [49]

    Gaggioli, S

    F. Gaggioli, S. Azadi, and L. Fu, Accurate self-attention wavefunctions at large scale (2026), arXiv:2607.08616 [cond-mat.str-el]

    2026 arXiv

  42. [50]

    Trivedi and D

    N. Trivedi and D. M. Ceperley, Ground-state correlations of quantum antiferromagnets: A green-function monte carlo study, Phys. Rev. B41, 4552 (1990)

    1990

  43. [51]

    D. F. B. ten Haaf, H. J. M. van Bemmel, J. M. J. van Leeuwen, W. van Saarloos, and D. M. Ceperley, Proof for an upper bound in fixed-node monte carlo for lattice fermions, Phys. Rev. B51, 13039 (1995)

    1995

  44. [52]

    Becca, W.-J

    F. Becca, W.-J. Hu, Y. Iqbal, A. Parola, D. Poilblanc, and S. Sorella, Lanczos steps to improve variational wave functions, Journal of Physics: Conference Series640, 012039 (2015)

    2015

  45. [53]

    Qin, C.-M

    M. Qin, C.-M. Chung, H. Shi, E. Vitali, C. Hubig, U. Schollw¨ ock, S. R. White, and S. Zhang (Simons Col- laboration on the Many-Electron Problem), Absence of superconductivity in the pure two-dimensional hubbard model, Phys. Rev. X10, 031016 (2020)

    2020

  46. [54]

    H. Xu, H. Shi, E. Vitali, M. Qin, and S. Zhang, Stripes and spin-density waves in the doped two-dimensional hubbard model: Ground state phase diagram, Phys. Rev. Res.4, 013239 (2022)

    2022

  47. [55]

    Rende and L

    R. Rende and L. L. Viteritti, Are queries and keys always relevant? A case study on transformer wave functions, Machine Learning: Science and Technology6, 010501 (2025)

    2025

  48. [56]

    Rende, F

    R. Rende, F. Gerace, A. Laio, and S. Goldt, Mapping of attention mechanisms to a generalized Potts model, Phys. Rev. Res.6, 023057 (2024)

    2024

  49. [57]

    Becca and S

    F. Becca and S. Sorella,Quantum Monte Carlo Ap- proaches for Correlated Systems(Cambridge University Press, 2017)

    2017

  50. [58]

    Sorella, Green Function Monte Carlo with Stochastic Reconfiguration, Phys

    S. Sorella, Green Function Monte Carlo with Stochastic Reconfiguration, Phys. Rev. Lett.80, 4558 (1998)

    1998

  51. [59]

    Chen and M

    A. Chen and M. Heyl, Empowering deep neural quantum states through efficient optimization, Nature Physics20, 1476 (2024)

    2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.