arxiv: 2605.09247 · v1 · submitted 2026-05-10 · ❄️ cond-mat.quant-gas · cond-mat.stat-mech

Recognition: 2 theorem links

· Lean Theorem

Benchmarking a restricted Boltzmann machine on the mathbb{Z}₂ Bose-Hubbard chain in the adiabatic hard-core regime

Gustavo Alejandro Avalos Valent\'in, Isaac P\'erez Castillo, Roman Josu\'e Armenta Rico

Authors on Pith no claims yet

Pith reviewed 2026-05-12 03:32 UTC · model grok-4.3

classification ❄️ cond-mat.quant-gas cond-mat.stat-mech

keywords restricted Boltzmann machinevariational Monte CarloBose-Hubbard modelZ2 gauge theoryadiabatic regimeneural quantum statesphase diagramsymmetry breaking

0 comments

The pith

A shallow restricted Boltzmann machine reproduces the adiabatic phase structure of the Z2 Bose-Hubbard chain at half filling

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

The paper benchmarks a shallow restricted Boltzmann machine as a variational ansatz for finding the ground state of the Z2 Bose-Hubbard chain in the adiabatic hard-core regime at half filling. It uses variational Monte Carlo sampling to compute magnetization observables and compares the resulting phase diagram against the established adiabatic description of the model. The neural ansatz distinguishes polarized and Neel-ordered regions and recovers representative staggered spin patterns and site occupations when a weak symmetry-breaking field is applied. A sympathetic reader would care because this tests whether a minimal neural network can faithfully capture the essential ground-state features of a gauge-coupled bosonic lattice model.

Core claim

Using variational Monte Carlo with a shallow restricted Boltzmann machine as the variational ansatz, the ground state of the Z2 Bose-Hubbard chain in the adiabatic hard-core limit at half filling is studied. The variational results reproduce the overall structure of the phase diagram obtained from magnetization observables, distinguish the polarized and Neel-ordered regions, and capture representative spin patterns and site occupations for the staggered insulating configurations selected by a weak symmetry-breaking field.

What carries the argument

Shallow restricted Boltzmann machine as variational ansatz in Monte Carlo sampling, which parametrizes the wavefunction to approximate ground-state properties

If this is right

The variational results match the expected overall phase diagram from magnetization observables.
Polarized and Neel-ordered regions are distinguished by the neural ansatz.
Staggered insulating configurations under weak symmetry breaking are recovered in site occupations and spin patterns.

Where Pith is reading between the lines

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

The success on this one-dimensional gauge model suggests the same shallow ansatz could be tested on non-adiabatic regimes or small two-dimensional lattices where exact methods become costly.
If the ansatz remains accurate, it offers a route to larger system sizes in similar Z2-coupled bosonic chains without requiring deeper networks.
The ability to select symmetry-broken states with a weak field points to possible use in studying related symmetry-breaking phenomena in lattice gauge theories.

Load-bearing premise

The shallow restricted Boltzmann machine wavefunction combined with variational Monte Carlo sampling is sufficiently expressive and free of sampling bias to represent the true ground state in this regime.

What would settle it

Exact diagonalization on small chains in the same adiabatic hard-core regime at half filling that produces magnetization or order-parameter values differing substantially from the restricted Boltzmann machine results would falsify the reproduction claim.

Figures

Figures reproduced from arXiv: 2605.09247 by Gustavo Alejandro Avalos Valent\'in, Isaac P\'erez Castillo, Roman Josu\'e Armenta Rico.

read the original abstract

We study the ground state of the $\mathbb{Z}_2$ Bose-Hubbard chain in the adiabatic hard-core limit at half filling using variational Monte Carlo with a shallow restricted Boltzmann machine as the variational ansatz. In this context, the neural quantum state is compared with the established adiabatic description of the model. The variational results reproduce the overall structure of the phase diagram obtained from magnetization observables, distinguish the polarized and N\'eel-ordered regions, and capture representative spin patterns and site occupations for the staggered insulating configurations selected by a weak symmetry-breaking field. Taken together, these results show that a shallow restricted Boltzmann machine reproduces the main adiabatic phase structure of the one-dimensional $\mathbb{Z}_2$ Bose-Hubbard chain and captures the selected symmetry-broken insulating configurations at half filling.

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.

Desk Editor's Note private letter to a colleague

This is a straightforward benchmarking study showing a shallow RBM reproduces the known adiabatic phase diagram of the Z2 Bose-Hubbard chain at half filling.

read the letter

The main point is that variational Monte Carlo with a shallow restricted Boltzmann machine recovers the established phase structure of this one-dimensional model in the adiabatic hard-core limit at half filling. The results line up with magnetization observables and pick out the polarized versus Neel-ordered regions plus representative spin patterns under a weak symmetry-breaking field. That match is the central finding. The work follows standard variational practice and the comparison target is independent, so there is no obvious circularity. The execution appears careful on the evidence given. The limitation is that the adiabatic description already exists, so the paper adds a data point on neural ansatz performance rather than new physics or a new method. A shallow RBM is a simple choice, which is fine for benchmarking, but the abstract leaves open how sensitive the results are to sampling details or convergence criteria. Those would need checking in the full text to judge robustness. This is the kind of paper that interests people testing neural quantum states on lattice models or working specifically with the Z2 Bose-Hubbard chain. It is not for readers seeking open problems or methodological breakthroughs. The methods are grounded and the comparison is fair, so it deserves a serious referee even if the scope stays narrow. I would send it to peer review for a specialized journal.

Referee Report

2 major / 2 minor

Summary. The manuscript benchmarks variational Monte Carlo sampling with a shallow restricted Boltzmann machine ansatz on the ground state of the one-dimensional Z2 Bose-Hubbard chain in the adiabatic hard-core regime at half filling. It reports that the variational results reproduce the overall structure of the phase diagram extracted from magnetization observables, distinguish polarized and Néel-ordered regions, and capture representative spin patterns together with site occupations for the staggered insulating states selected by a weak symmetry-breaking field.

Significance. If the reported reproduction holds under converged sampling, the work supplies a concrete benchmark showing that a shallow RBM can represent the essential adiabatic physics of this Z2-symmetric lattice model without apparent circularity in the comparison. Such benchmarks are useful for calibrating neural quantum states against independent analytic or numerical references in one-dimensional quantum spin systems.

major comments (2)

[§4 (Results)] §4 (Results): the claim that the RBM-VMC 'reproduces the overall structure of the phase diagram' rests on visual or qualitative agreement in magnetization observables; quantitative metrics (e.g., L2 deviation of order parameters or location of phase boundaries with error bars) are not provided, leaving the strength of the reproduction open to interpretation.
[§3 (Methods)] §3 (Methods): the description of the VMC sampling does not specify the number of samples per optimization step, the thermalization length, or the autocorrelation time used to compute the reported magnetization and spin patterns; without these, it is difficult to assess whether the captured configurations are statistically reliable.

minor comments (2)

[Notation] The notation for the symmetry-breaking field strength should be introduced once in the text and used consistently in all figure captions and tables.
[Figures] Figure 2 (or equivalent): the site-occupation plots for the Néel configurations would benefit from an overlay of the exact adiabatic reference pattern for direct visual comparison.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading of our manuscript and the constructive comments. We appreciate the positive assessment of the benchmark and the recommendation for minor revision. We address each major comment below and have revised the manuscript accordingly to improve clarity and rigor.

read point-by-point responses

Referee: [§4 (Results)] §4 (Results): the claim that the RBM-VMC 'reproduces the overall structure of the phase diagram' rests on visual or qualitative agreement in magnetization observables; quantitative metrics (e.g., L2 deviation of order parameters or location of phase boundaries with error bars) are not provided, leaving the strength of the reproduction open to interpretation.

Authors: We agree that the original presentation relied primarily on visual comparison of magnetization observables and the distinction of polarized versus Néel-ordered regions. To strengthen the claim, we have added quantitative metrics in the revised §4: the L2 deviation between RBM-VMC and reference order parameters at selected points in the phase diagram, together with estimated locations of the phase boundaries including statistical uncertainties obtained from the Monte Carlo sampling. These additions are supported by a new table in the supplementary material. revision: yes
Referee: [§3 (Methods)] §3 (Methods): the description of the VMC sampling does not specify the number of samples per optimization step, the thermalization length, or the autocorrelation time used to compute the reported magnetization and spin patterns; without these, it is difficult to assess whether the captured configurations are statistically reliable.

Authors: We thank the referee for highlighting this omission in the methods description. In the revised manuscript we have expanded §3 to specify the sampling parameters used throughout the study: 10^4 samples per optimization step following 5000 thermalization sweeps, with an integrated autocorrelation time of approximately 50 Monte Carlo steps for the magnetization observables (computed via the blocking method). These details confirm that the reported configurations and observables are statistically reliable. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper benchmarks a shallow RBM variational ansatz via VMC optimization on the Z2 Bose-Hubbard chain at half filling in the adiabatic hard-core limit. The central procedure optimizes the neural quantum state parameters to minimize the variational energy (standard VMC), then compares resulting magnetization observables, phase boundaries, and selected symmetry-broken configurations against an independently established adiabatic description of the model. No load-bearing step reduces the reported reproduction to a self-definition, a fitted input renamed as prediction, or a self-citation chain; the adiabatic benchmark is external and the optimization does not incorporate the target phase structure or observables as constraints. The derivation chain is therefore self-contained against an external reference.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only the abstract is available, so the ledger is inferred from the high-level description. The claim rests on the validity of the adiabatic hard-core limit as an accurate reference and on the assumption that variational Monte Carlo with a shallow RBM can converge to the true ground state without significant bias.

axioms (1)

domain assumption The adiabatic hard-core limit provides an accurate established description of the Z2 Bose-Hubbard chain at half filling.
The paper uses this description as the benchmark for all comparisons.

pith-pipeline@v0.9.0 · 5453 in / 1244 out tokens · 56150 ms · 2026-05-12T03:32:33.521228+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
We study the ground state of the Z2 Bose-Hubbard chain in the adiabatic hard-core limit at half filling using variational Monte Carlo with a shallow restricted Boltzmann machine as the variational ansatz... reproduce the overall structure of the phase diagram obtained from magnetization observables
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
The variational results reproduce the overall structure of the phase diagram... distinguish the polarized and Néel-ordered regions

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

[1]

and the early overview of RBMs in quantum physics [13], the field expanded toward recurrent and autoregres- sive wave functions with exact sampling [21], explicit ap- plications to topological order beyond conventional Landau phases [22], and deeper analyses of expressive power link- ing deep neural quantum states to tensor-network construc- tions [23]. M...

work page
[2]

The optimized state is then probed through the total and staggered magnetizations introduced in Eq. (2). Since the two symmetry-related Néel configurations differ only by the Rev. Mex. Fis.??(*?*) (????) ???–??? A LATEXTEMPLATE FOR THE RMF, RMF-E, SRMF5 sign of the staggered response, the phase map is most con- veniently summarized by the pair(m t,|m s|)....

work page
[3]

González-Cuadra, et al., Strongly Correlated Bosons on a Dynamical Lattice, Phys

D. González-Cuadra, et al., Strongly Correlated Bosons on a Dynamical Lattice, Phys. Rev. Lett. 121 (2018) 090402, 10.1103/PhysRevLett.121.090402

work page doi:10.1103/physrevlett.121.090402 2018
[4]

González-Cuadra, et al., Symmetry-Breaking Topological Insulators in theZ 2 Bose-Hubbard Model, Physical Review B 99 (2019) 045139

D. González-Cuadra, et al., Symmetry-Breaking Topological Insulators in theZ 2 Bose-Hubbard Model, Physical Review B 99 (2019) 045139

work page 2019
[5]

González-Cuadra, et al., Intertwined topological phases in- duced by emergent symmetry protection, Nature Communica- tions 10 (2019) 2694,10.1038/s41467-019-10796-8

D. González-Cuadra, et al., Intertwined topological phases in- duced by emergent symmetry protection, Nature Communica- tions 10 (2019) 2694,10.1038/s41467-019-10796-8

work page doi:10.1038/s41467-019-10796-8 2019
[6]

Statistical Mechanics of Interacting Run-and-Tumble Bacteria

D. González-Cuadra, et al., Dynamical Solitons and Boson Fractionalization in Cold-Atom Topological Insulators, Phys. Rev. Lett. 125 (2020) 265301,10.1103/PhysRevLett. 125.265301

work page doi:10.1103/physrevlett 2020
[7]

Chanda, et al., Devil’s staircase of topological Peierls in- sulators and Peierls supersolids, SciPost Phys

T. Chanda, et al., Devil’s staircase of topological Peierls in- sulators and Peierls supersolids, SciPost Phys. 12 (2022) 076, 10.21468/SciPostPhys.12.2.076

work page doi:10.21468/scipostphys.12.2.076 2022
[8]

Smidman, M

T. Chanda, et al., Recent progress on quantum simulations of non-standard Bose-Hubbard models, Reports on Progress in Physics 88 (2025) 044501,10.1088/1361-6633/ adc3a7

work page doi:10.1088/1361-6633/ 2025
[9]

González-Cuadra, et al.,Z n solitons in intertwined topo- logical phases, Physical Review B 102 (2020) 245137,10

D. González-Cuadra, et al.,Z n solitons in intertwined topo- logical phases, Physical Review B 102 (2020) 245137,10. 1103/PhysRevB.102.245137

work page 2020
[10]

Watanabe, S

Y . Watanabe, S. Watabe, and T. Nikuni, Transverse Field De- pendence of the Ground State in theZ 2 Bose-Hubbard Model (2025)

work page 2025
[11]

S. R. White, Density matrix formulation for quantum renormal- ization groups, Phys. Rev. Lett. 69 (1992) 2863,10.1103/ PhysRevLett.69.2863

work page 1992
[12]

Schollwöck,The density-matrix renormalization group in the age of matrix product states, Annals of Physics326(1), 96 (2011), doi:10.1016/j.aop.2010.09.012

U. Schollwöck, The density-matrix renormalization group in the age of matrix product states, Annals of Physics 326 (2011) 96,10.1016/j.aop.2010.09.012

work page doi:10.1016/j.aop.2010.09.012 2011
[13]

Catarina and B

G. Catarina and B. Murta, Density-matrix renormal- ization group: a pedagogical introduction, The Euro- pean Physical Journal B 96 (2023),10.1140/epjb/ s10051-023-00575-2

work page doi:10.1140/epjb/ 2023
[14]

Carleo and M

G. Carleo and M. Troyer, Solving the Quantum Many-Body Problem with Artificial Neural Networks, Science 355 (2017) 602,10.1126/science.aag2302

work page doi:10.1126/science.aag2302 2017
[15]

Melko, et al., Restricted Boltzmann Machines in Quan- tum Physics, Nature Physics 15 (2019),10.1038/ s41567-019-0545-1

R. Melko, et al., Restricted Boltzmann Machines in Quan- tum Physics, Nature Physics 15 (2019),10.1038/ s41567-019-0545-1

work page 2019
[16]

Lange, et al., From Architectures to Applications: A Review of Neural Quantum States, Quantum Science and Technology 9 (2024) 040501,10.1088/2058-9565/ad7168

H. Lange, et al., From Architectures to Applications: A Review of Neural Quantum States, Quantum Science and Technology 9 (2024) 040501,10.1088/2058-9565/ad7168

work page doi:10.1088/2058-9565/ad7168 2024
[17]

Saito, Solving the Bose-Hubbard Model with Machine Learning, Journal of the Physical Society of Japan 86 (2017) 093001,10.7566/JPSJ.86.093001

H. Saito, Solving the Bose-Hubbard Model with Machine Learning, Journal of the Physical Society of Japan 86 (2017) 093001,10.7566/JPSJ.86.093001

work page doi:10.7566/jpsj.86.093001 2017
[18]

Saito and M

H. Saito and M. Kato, Machine Learning Technique to Find Quantum Many-Body Ground States of Bosons on a Lattice, Journal of the Physical Society of Japan 87 (2018) 014001, 10.7566/JPSJ.87.014001

work page doi:10.7566/jpsj.87.014001 2018
[19]

McBrian, G

K. McBrian, G. Carleo, and E. Khatami, Ground state phase diagram of the one-dimensional Bose-Hubbard model from re- stricted Boltzmann machines, Journal of Physics: Conference Series 1290 (2019) 012005,10.1088/1742-6596/1290/ 1/012005

work page doi:10.1088/1742-6596/1290/ 2019
[20]

Vargas-Calderón, H

V . Vargas-Calderón, H. Vinck-Posada, and F. A. González, Phase Diagram Reconstruction of the Bose-Hubbard Model with a Restricted Boltzmann Machine Wavefunction, Journal of the Physical Society of Japan 89 (2020) 094002,10.7566/ JPSJ.89.094002

work page 2020
[21]

Çeven, M

K. Çeven, M. Ö. Oktel, and A. Kele¸ s, Neural-network quan- tum states for a two-leg Bose-Hubbard ladder under magnetic flux, Physical Review A 106 (2022) 063320,10.1103/ PhysRevA.106.063320

work page 2022
[22]

M. Y . Pei and S. R. Clark, Specialising neural-network quan- tum states for the Bose Hubbard model, Journal of Physics B: Atomic, Molecular and Optical Physics 57 (2024) 215301, 10.1088/1361-6455/ad7e87

work page doi:10.1088/1361-6455/ad7e87 2024
[23]

Hibat-Allah, et al., Recurrent neural network wave func- tions, Physical Review Research 2 (2020) 023358,10.1103/ PhysRevResearch.2.023358

M. Hibat-Allah, et al., Recurrent neural network wave func- tions, Physical Review Research 2 (2020) 023358,10.1103/ PhysRevResearch.2.023358

work page 2020
[24]

Hibat-Allah, R

M. Hibat-Allah, R. G. Melko, and J. Carrasquilla, Investigat- ing topological order using recurrent neural networks, Physical Review B 108 (2023) 075152,10.1103/PhysRevB.108. 075152

work page doi:10.1103/physrevb.108 2023
[25]

Sharir, A

O. Sharir, A. Shashua, and G. Carleo, Neural tensor con- tractions and the expressive power of deep neural quantum states, Physical Review B 106 (2022) 205136,10.1103/ PhysRevB.106.205136

work page 2022
[26]

Chen and M

A. Chen and M. Heyl, Empowering deep neural quantum states through efficient optimization, Nature Physics 20 (2024) 1476, 10.1038/s41567-024-02566-1

work page doi:10.1038/s41567-024-02566-1 2024
[27]

Sprague and S

K. Sprague and S. Czischek, Variational Monte Carlo with Large Patched Transformers, Communications Physics 7 (2024) 90,10.1038/s42005-024-01584-y

work page doi:10.1038/s42005-024-01584-y 2024
[28]

Denis and G

Z. Denis and G. Carleo, Accurate neural quantum states for in- teracting lattice bosons, Quantum 9 (2025) 1772,10.22331/ q-2025-06-17-1772 Rev. Mex. Fis.??(*?*) (????) ???–??? 10RMF EDITORIAL TEAM

work page 2025
[29]

W. P. Su, J. R. Schrieffer, and A. J. Heeger, Solitons in Poly- acetylene, Phys. Rev. Lett. 42 (1979) 1698,10.1103/ PhysRevLett.42.1698

work page 1979
[30]

Vicentini, et al., NetKet 3: Machine Learning Toolbox for Many-Body Quantum Systems (2021)

F. Vicentini, et al., NetKet 3: Machine Learning Toolbox for Many-Body Quantum Systems (2021)

work page 2021
[31]

Sorella, Green Function Monte Carlo with Stochastic Re- configuration, Phys

S. Sorella, Green Function Monte Carlo with Stochastic Re- configuration, Phys. Rev. Lett. 80 (1998) 4558,10.1103/ PhysRevLett.80.4558 Rev. Mex. Fis.??(*?*) (????) ???–???

work page 1998