Pith. sign in

REVIEW 3 major objections 5 minor 39 references

Effective temperature in approximate quantum many-body states

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Approximate many-body states from variational optimization share a universal spectral pattern: squared overlaps with exact eigenstates decay exponentially in energy, with a small decay factor β̃ the paper calls the inverse effective…

desk verdict Useful diagnostic survey with a real ITES two-stage finding, but the 'universal exponential spectrum' claim is oversold until the fitting protocol is pinned down. read the letter →

arxiv 2411.18921 v1 pith:ZZY3EK42 submitted 2024-11-28 quant-ph cond-mat.str-el

Yu-Qin Chen , Shi-Xin Zhang This is my paper · ORCID
classification quant-phcond-mat.str-el
keywords effectivetemperaturespectraldecompositiontensornetworksneuralquantumstatesvariationaleigensolverimaginary-timeevolutionmany-bodysystemsexponentialdecay
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

This paper tries to establish that approximate wavefunctions produced by variational optimization—matrix product states, projected entangled-pair states, neural quantum states, variational quantum circuits, and fully parameterized vectors—all show the same spectral fingerprint when decomposed into the exact eigenstates of the target Hamiltonian: the squared overlap with eigenstate i decays roughly exponentially with its energy, |c_i|^2 ∝ exp(−β̃ ε_i), with a small decay rate β̃ interpreted as an inverse effective temperature. The authors argue this pattern is universal across ansatz structure, optimization objective, and physical system, and that β̃ tracks ansatz expressiveness and training progress. They also introduce imaginary-time evolved target states with exact exponential spectra and show a two-stage behavior: for small inverse temperature β the approximate state matches the target, while beyond a critical β* the high-energy part of the spectrum flattens and the fitted β̃ drops below β. If the picture holds, β̃ becomes a practical diagnostic for when a variational simulation is effectively hot and for comparing the capacity of different ansatzes.

What carries the argument

The machinery is the spectral decomposition |ψ⟩ = ∑_i c_i|ε_i⟩ and the exponential fit |c_i|² ∝ $e^{{−β̃ ε_i}}$; β̃ is obtained by linear regression of log|ci|² against εi over the full spectrum or, for VQE and the 4×4 lattice, the half-filling sector. The companion object is the imaginary-time evolved state |φ(β)⟩ = Z(β)^{−1}∑_i $e^{{−β ε_i/2}}$|ε_i⟩, a family whose exact spectra are exponential by construction and which provides a controlled test of whether an ansatz can represent a given inverse temperature. The transition value β* is defined by the point where the fitted slope begins to deviate from the target β.

What would settle it

For the 4×4 XXZ model in the half-filling sector, take a converged MPS ground-state approximant optimized with the fidelity objective and fit log|ci|² against εi separately over the lower third and upper third of the spectrum; if the two fitted slopes differ by more than the fit uncertainty, or if one is negative while the other is positive, the claimed universal single-exponential law fails for that state.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is an empirical rule: for approximate ground states obtained by minimizing energy or infidelity, plotting log|ci|² against eigenenergy εi yields an approximately straight line over a wide range of energies, corresponding to |ci|² ∝ $e^{{−β̃ ε_i}}$. The fitted slope β̃ is small—typically below about 0.3, and sometimes negative for converged fidelity-optimized states—so the excited-state spectrum of an approximate ground state is nearly flat. For imaginary-time evolved targets |φ(β)⟩ = $Z^{{-1}}$∑ $e^{{−β ε_i/2}}$|ε_i⟩, the optimized approximate states reproduce the target exponential spectrum for β below a critical β*, while for β > β* only the low-energy part decays exponentially and the high-energy overlaps form a plateau; the paper presents evidence that this plateau is not a machine-precision artifact and that β* is characteristic of the ansatz's expressive power.

Load-bearing premise

The universal law rests on assuming that a single fitted slope of log|ci|² versus eigenenergy, extracted over a chosen set of eigenstates, is a stable and representative description of the whole spectrum, independent of the fitting window, charge sector, and optimization run.

Editorial extensions

If this is right

  • A converged approximate ground state is never truly cold: its excited-state spectrum remains nearly flat, so any observable computed from the variational state carries a high-energy tail that is easy to underestimate.
  • Operator expectations of functions of the Hamiltonian, such as ⟨ψ|H²|ψ⟩, inherit fragility from this tail, which matters for sampling-based energy-variance estimators.
  • The critical inverse temperature β* provides an ansatz-dependent ranking of how low an effective temperature each method can represent, and it correlates with fidelity.
  • Training dynamics of β̃ first rise and then fall, so the maximum β̃ reached during optimization acts as an upper bound tied to the ansatz's capacity.
  • Energy and fidelity objectives give similar β̃ dynamics within the same ansatz family, with energy typically yielding larger β̃ because it penalizes high-energy states unevenly.

Reading between the lines

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

  • If the exponential spectrum is a generic outcome of gradient-based variational optimization, the same fit should appear for random or badly converged approximate states; testing this would separate a property of the ansatz family from a property of the optimization dynamics.
  • The definition of β̃ depends on the chosen energy window and charge sector, so a practical benchmark would need a prespecified fitting protocol; the paper's reported negative slopes for converged ground states suggest the single-exponential description is not always stable.
  • The β* transition may be governed by optimization hardness rather than representational power alone; the paper's comparison with a stronger optimizer suggests that the apparent phase boundary could shift if the optimizer is changed, so β* should be interpreted as a property of the method-plus-optimizer pair.
  • A scaling study of β̃ with bond dimension, circuit depth, or network width—only touched in the paper—would sharpen the claim that β̃ is an expressiveness diagnostic.
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 / 5 minor

Summary. The paper studies the spectral decomposition of approximate quantum many-body states (obtained from MPS, PEPS, NQS/VMC, VQE, and fully parameterized vector states) into the exact eigenstates of the target Hamiltonian. It claims that the squared overlaps |c_i|^2 as a function of eigenenergy ε_i roughly follow an exponential decay |c_i|^2 ∝ exp(−β̃ ε_i), defining β̃ as an inverse effective temperature. The authors report small positive β̃ for most approximate ground states, a non-monotonic training dynamics of β̃, and a two-stage behavior when approximating imaginary-time evolved states (ITES), with a critical β* separating a regime where β̃ ≈ β from a regime where β̃ < β. The findings are presented as universal across ansatzes, objectives, and lattice geometries.

Significance. If the exponential spectral-decay pattern is shown to be robust, the effective temperature could provide a useful, fine-grained diagnostic for variational quantum and classical methods, complementing global metrics such as fidelity. The paper has notable strengths: it covers a wide range of ansatzes and objectives, provides extensive supplemental material with alternative metrics (MSE, Pearson coefficient, fit uncertainty), and includes a credible argument against a pure machine-precision explanation of the spectral plateau. The use of open-source software (TensorCircuit-NG, QuSpin) supports reproducibility. However, the central claim is currently vulnerable to the underspecified fitting procedure used to define β̃, and the reported universality is not yet established.

major comments (3)
  1. [Results for approximate ground states, Eq. (4)] The extraction of β̃ via a linear fit to log |c_i|^2 versus ε_i is never specified: the paper does not state the fitting window, the weighting of points, the treatment of degenerate eigenstates, or how near-zero overlaps are handled. This is a load-bearing omission because the same spectral data can yield different slopes under different protocols. Concretely, Fig. 2(b) and Fig. S10(b) report converged ground-state approximants with β̃ ≈ −0.114 and −0.165, respectively, meaning the fitted coefficients increase with energy, which contradicts the 'exponential decay' wording of Eq. (4) and the claim of a universal small positive β̃. The manuscript should specify the fitting protocol exactly and demonstrate that the reported β̃ values are stable with respect to the fit window and weighting.
  2. [Results for approximate ground states and Fig. S13] The reported β̃ depends on the charge-sector restriction. For VQE on the 4×3 lattice, fitting the full spectrum gives β̃ ≈ 0.477–0.799, whereas restricting to the half-filling sector gives β̃ ≈ 0.284–0.317 (Fig. S13), and the main text reports only the sector-restricted values. Similarly, the 4×4 calculations restrict spectral decomposition to the half-filling sector for VQE and other methods. Since the sector choice changes the numerical value of the central metric, the claimed universality of small positive β̃ is not established unless the sector selection is justified on physical grounds and consistently applied. The authors should report full-spectrum fits alongside sector-restricted fits and discuss the sensitivity.
  3. [Results for approximate imaginary-time evolved states, Fig. 4] The two-stage behavior for ITES and the critical value β* are derived from the same underspecified single-exponential fit, and for large β the spectrum is explicitly not exponential over the full energy range (Fig. 2(d) and Fig. S10(d)). A single fitted slope β̃ is therefore not a valid characterization of such spectra, and β* inherits the fitting ambiguity. In addition, the high-temperature regime where β̃ ≈ β is expected almost by construction when the fidelity objective directly targets |ϕ(β)⟩, so this regime does not independently test the exponential law. The manuscript should define β* via an explicit, reproducible criterion (for example, the point where the MSE or the fit uncertainty crosses a stated threshold) and show that the qualitative conclusions are robust to the criterion chosen.
minor comments (5)
  1. [Results for approximate ground states] The phrase 'The universality and validness of the results' should read 'validity'.
  2. [Throughout] The abbreviation 'aka.' should be written as 'a.k.a.' or 'also known as' in the main text and figure captions.
  3. [Introduction] In the sentence 'This objective reduces to the ground state target in the β → ∞limit', insert a space before 'limit'.
  4. [Abstract and Discussion] The abstract states that the effective temperature 'shows phase transition behaviors', while the Discussion more cautiously calls it a 'putative phase transition'; the wording should be aligned to avoid overclaiming.
  5. [Methods, Neural quantum states] The text uses 'VMC' for the neural-network ansatz while clarifying that sampling is not used; to avoid confusion with standard variational Monte Carlo, consider using 'NQS' consistently in the main text and figures.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the effective temperature is a fitted descriptor of the spectral data, and the paper's empirical claims rest on goodness-of-fit diagnostics rather than on a parameter reintroduced as a prediction.

full rationale

The central claim is an empirical characterization, not a derivation: the paper exactly diagonalizes small Hamiltonians, decomposes optimized variational states as |ψ⟩ = Σ_i c_i |ε_i⟩, and fits log |c_i|^2 versus ε_i to extract the slope β̃ (Eq. 4). β̃ is defined as that fitted slope, so calling it the 'inverse effective temperature' is a naming choice; the actual empirical assertion is that the plots are approximately linear. That assertion is supported by independent diagnostics reported in the SM: squared Pearson coefficients close to 1 (Fig. S16), small fit uncertainties Δβ̃ (Figs. S5, S17), and direct scatter plots across ansatzes, objectives, and systems. The existence of a best-fit slope does not by construction guarantee a good fit, and indeed the paper shows negative fitted slopes for some converged ground-state approximants (Fig. 2b, S10b), which demonstrates that β̃ is not constrained by the fitting procedure to be small or positive. The small-positive-β̃ pattern is therefore an empirical finding rather than a tautology. For imaginary-time evolved states, the small-β statement β̃ ≈ β is a consistency check of fidelity optimization against a target whose spectrum is exponential by definition (Eq. 5); the paper does not present this as an independent prediction, and the nontrivial content is the high-β plateau and the ansatz-dependent β*, which are not built into the fitting. The self-citations present ([10], [23], [30] by S.-X. Zhang, with [23] also including Y.-Q. Chen) are software and background method citations and are not load-bearing for the spectral-universality claim. The noted protocol dependence of β̃ (unspecified fit windows/weights, sector restriction for VQE, sensitivity to fitting range) is a robustness and interpretability concern about the universality claim, not a circular reduction in the paper's derivation chain.

Assumptions & free parameters 2 free parameters · 3 assumptions · 2 invented entities

The central claim rests on fitting a single exponential to spectral overlaps and on the assumption that the optimized states are representative of the ansatz family. The only fitted constants are the slope beta_tilde and prefactor Lambda; the ITES target beta is an input parameter, not fitted. The effective temperature and beta_star are constructed quantities with no independent evidence outside the paper.

free parameters (2)
  • beta_tilde (inverse effective temperature) = varies, e.g., 0.121, -0.114, 0.353
    Fitted as the slope of a linear regression of log |c_i|^2 versus epsilon_i for each approximate state; it is the central quantity of the paper.
  • Lambda (prefactor of exponential fit) = varies by state and training stage
    Mentioned in SM Section II as the second parameter in the fit Lambda * exp(-beta_tilde * epsilon); it is fitted to the same spectral data and correlates with infidelity.
assumptions (3)
  • standard math The Hamiltonian spectral decomposition into exact eigenstates is complete and computable via exact diagonalization for L <= 16.
    Used throughout to obtain (epsilon_i, |c_i|^2) pairs; relies on Hermiticity and small Hilbert space.
  • domain assumption The state reached by gradient descent is representative of what the ansatz family can express, so beta_tilde and beta_star reflect expressiveness rather than optimization failure.
    Invoked in interpreting beta_star as a figure of merit for ansatz capacity; SM partially tests this with L-BFGS, but it is not proven.
  • ad hoc to paper A single exponential function Lambda * exp(-beta_tilde * epsilon) adequately describes the spectral overlaps over the fitted energy range.
    This is the central model assumption; the paper itself shows deviations (plateaus, negative slopes) and restricts the sector, so the adequacy is not always established.
invented entities (2)
  • Effective temperature 1/beta_tilde for approximate pure states
    purpose: Diagnostic metric characterizing the exponential spectral tail of variational approximate states.
    Introduced in this paper (Eq. 4, Fig. 1); it is defined by a fit to the same data and has no external prediction outside the paper.
  • Critical inverse effective temperature beta_star
    purpose: Marks the boundary between high-temperature and low-temperature regimes in approximating ITES, proposed as a measure of ansatz expressiveness.
    Defined operationally as the point where beta_tilde deviates from the target beta; no independent observable is tied to it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Effective temperature in approximate quantum many-body states." pith.science (2026). https://pith.science/paper/ZZY3EK42

@misc{pith2026241118921,
  author       = {Pith},
  title        = {Pith review of: Effective temperature in approximate quantum many-body states},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZZY3EK42}},
  note         = {Machine review of arXiv:2411.18921}
}
read the original abstract

In the pursuit of numerically identifying the ground state of quantum many-body systems, approximate quantum wavefunction ansatzes are commonly employed. This study focuses on the spectral decomposition of these approximate quantum many-body states into exact eigenstates of the target Hamiltonian. The energy spectral decomposition could reflect the intricate physics at the interplay between quantum systems and numerical algorithms. Here we examine various parameterized wavefunction ansatzes constructed from neural networks, tensor networks, and quantum circuits, employing differentiable programming to numerically approximate ground states and imaginary-time evolved states. Our findings reveal a consistent exponential decay pattern in the spectral contributions of approximate quantum states across different ansatzes, optimization objectives, and quantum systems, characterized by small decay rates denoted as inverse effective temperatures. The effective temperature is related to ansatz expressiveness and accuracy and shows phase transition behaviors in learning imaginary-time evolved states. The universal picture and unique features suggest the significance and potential of the effective temperature in characterizing approximate quantum states.

Figures

Figures reproduced from arXiv: 2411.18921 by the authors.

Figure 1
Figure 1. Sketch of the effective temperature as a metric characterizing different systems and numerical meth￾ods. The effective temperature (1/β˜) can be extracted from the estimated slope of the approximate quantum state spectral decomposition. The metric varies with different systems, methods, objectives, and training wellness. However, universal fea￾tures remain the same and are valuable in diagnosing the capacity of diff… view at source ↗
Figure 2
Figure 2. Spectral decomposition of approximate states. The 2D XXZ model on 4 × 3 square lattice with Jx = Jy = 1, Jz = 0.8 and MPS ansatz with bond dimension χ = 32 are employed. The optimization objective is the infi￾delity between approximate states and the target states. The target states are chosen as the ground states for (a)(b) and imaginary-time evolved states |ϕ(β)⟩ with β = 0.3 (c) and β = 0.6 (d). Overlaps with dif… view at source ↗
Figure 4
Figure 4. Effective temperature and fidelity of ap￾proximate ITES with different βs. The results are ob￾tained from the 2D XXZ model on 4 × 3 square lattice with Jx = Jy = 1, Jz = 0.8. (a) The inverse effective temperature β˜ shows a two-stage behavior with varying β. β ∗ (insets) marks the deviation of β˜ from target β, separating the two stages. (b) The fidelity of approximate ITES with different βs. For high-accuracy metho… view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: The tensor network graphic representation for the periodic MPS ansatz used in this work. Each blue square corresponds to a tensor of shape (χ, χ, 2). The red legs correspond to physical indices of dimension 2 and all blue legs are virtual bonds of dimension χ. The dash…
Figure 6
Figure 6. Figure 6: The tensor network graphic representation for the two dimensional PEPS ansatz with periodic boundary conditions used in this work. Each blue square correspond to a tensor of shape (χ, χ, χ, χ, 2). The red legs correspond to physical indices of dimension 2 on each site …
Figure 8
Figure 8. Figure 8: VQE ansatz employed in this work. For 1D lat￾tice, a parameterized swap gate on qubits 1 and L is omitted. For 2D lattices, these parameterized swap gates are aligned with lattice bonds between the nearest neighbor sites. The initialization block prepares the initial s…
Figure 7
Figure 7. Figure 7: Neural network architecture for the NQS used in this work. The wavefunction amplitude and phase are evaluated separately by going through a stack of residue blocks (ResBlock). In each ResBlock, the input vector of dimension L is firstly mapped to a vector of dimension …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 26 canonical work pages

  1. [1]

    Kitaev, A

    A. Kitaev, A. H. Shen, and M. N. Vyalyi, Classi- cal and quantum computation (American Mathematical Soc., Providence, 2002)

    work page 2002

  2. [2]

    Kempe, A

    J. Kempe, A. Kitaev, and O. Regev, The complexity of the local Hamiltonian problem, SIAM J. Comput.35, 1070 (2006)

    2006

  3. [3]

    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 region of the two-dimensional Hubbard model, Science358, 1155 (2017)

    2017

  4. [4]

    S. Lee, J. Lee, H. Zhai, Y. Tong, A. M. Dalzell, A. Ku- mar, P. Helms, J. Gray, Z.-H. Cui, W. Liu, M. Kasto- ryano, R. Babbush, J. Preskill, D. R. Reichman, E. T. Campbell, E. F. Valeev, L. Lin, and G. K.-L. Chan, Evaluating the evidence for exponential quantum advan- tage in ground-state quantum chemistry, Nat. Commun. 14, 1952 (2023). 8

    work page 2023

  5. [5]

    Greiner, L

    W. Greiner, L. Neise, and H. Stöcker, Thermodynam- ics and Statistical Mechanics (Springer New York, New York, NY, 1995)

    work page 1995

  6. [6]

    D. Wu, R. Rossi, F. Vicentini, N. Astrakhantsev, F. Becca, X. Cao, J. Carrasquilla, F. Ferrari, A. Georges, M. Hibat-Allah, M. Imada, A. M. Läuchli, G. Mazzola, A. Mezzacapo, A. Millis, J. Robledo Moreno, T. Neupert, Y. Nomura, J. Nys, O. Parcollet, R. Pohle, I. Romero, M. Schmid, J. M. Silvester, S. Sorella, L. F. Tocchio, L. Wang, S. R. White, A. Wietek...

    work page 2024

  7. [7]

    D. E. Rumelhart, G. E. Hinton, and R. J. Williams, Learningrepresentationsbyback-propagatingerrors,Na- ture 323, 533 (1986)

    work page 1986

  8. [8]

    Bartholomew-Biggs, S

    M. Bartholomew-Biggs, S. Brown, B. Christianson, and L. Dixon, Automatic differentiation of algorithms, J. Comput. Appl. Math.124, 171 (2000)

    work page 2000

Show all 39 references
  1. [9]

    Liao, J.-G

    H.-J. Liao, J.-G. Liu, L. Wang, and T. Xiang, Differen- tiable Programming Tensor Networks, Phys. Rev. X9, 31041 (2019)

    2019

  2. [10]

    Zhang, Z.-Q

    S.-X. Zhang, Z.-Q. Wan, and H. Yao, Automatic dif- ferentiable Monte Carlo: Theory and application, Phys. Rev. Res.5, 033041 (2023)

    2023

  3. [11]

    S. R. White, Density Matrix Formulation for Quan- tum Renormalization Groups, Phys. Rev. Lett.69, 2863 (1992)

    1992

  4. [12]

    Phys.326, 96 (2011)

    U.Schollwöck,Thedensity-matrixrenormalizationgroup in the age of matrix product states, Ann. Phys.326, 96 (2011)

    2011

  5. [13]

    Stoudenmire and S

    E. Stoudenmire and S. R. White, Studying Two- Dimensional Systems with the Density Matrix Renormal- ization Group, Annu. Rev. Condens. Matter Phys.3, 111 (2012)

    2012

  6. [14]

    Verstraete and J

    F. Verstraete and J. I. Cirac, Renormalization algorithms for Quantum-Many Body Systems in two and higher di- mensions, arXiv:cond-mat/0407066 (2004)

    2004 arXiv

  7. [15]

    Verstraete and J

    F. Verstraete and J. I. Cirac, Valence-bond states for quantum computation, Phys. Rev. A70, 060302 (2004)

    2004

  8. [16]

    Carleo and M

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

    2017

  9. [17]

    D.-L. Deng, X. Li, and S. Das Sarma, Quantum En- tanglement in Neural Network States, Phys. Rev. X7, 021021 (2017)

    2017

  10. [18]

    Cerezo, A

    M. Cerezo, A. Arrasmith, R. Babbush, S. C. Benjamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio, and P. J. Coles, Variational quantum algo- rithms, Nat. Rev. Phys.3, 625 (2021)

    2021

  11. [19]

    Bharti, A

    K. Bharti, A. Cervera-Lierta, T. H. Kyaw, T. Haug, S. Alperin-Lea, A. Anand, M. Degroote, H. Heimonen, J. S. Kottmann, T. Menke, W.-K. Mok, S. Sim, L.-C. Kwek, and A. Aspuru-Guzik, Noisy intermediate-scale quantum algorithms, Rev. Mod. Phys.94, 015004 (2022)

    2022

  12. [20]

    Peruzzo, J

    A. Peruzzo, J. McClean, P. Shadbolt, M.-H. Yung, X.-Q. Zhou, P. J. Love, A. Aspuru-Guzik, and J. L. O’Brien, A variational eigenvalue solver on a photonic quantum processor, Nat. Commun.5, 4213 (2014)

    2014

  13. [21]

    Tilly, H

    J. Tilly, H. Chen, S. Cao, D. Picozzi, K. Setia, Y. Li, E. Grant, L. Wossnig, I. Rungger, G. H. Booth, and J. Tennyson, The Variational Quantum Eigensolver: A review of methods and best practices, Phys. Rep.986, 1 (2022)

    2022

  14. [22]

    Farhi, J

    E. Farhi, J. Goldstone, and S. Gutmann, A Quantum Approximate Optimization Algorithm, arXiv:1411.4028 (2014)

    2014 arXiv

  15. [23]

    Zhang, J

    S.-X. Zhang, J. Allcock, Z.-Q. Wan, S. Liu, J. Sun, H. Yu, X.-H. Yang, J. Qiu, Z. Ye, Y.-Q. Chen, C.- K. Lee, Y.-C. Zheng, S.-K. Jian, H. Yao, C.-Y. Hsieh, and S. Zhang, TensorCircuit: a Quantum Software Framework for the NISQ Era, Quantum7, 912 (2023). https://github.com/tens...

    2023

  16. [24]

    Weinberg and M

    P. Weinberg and M. Bukov, QuSpin: a Python package for dynamics and exact diagonalisation of quantum many body systems part I: spin chains, SciPost Phys.2, 003 (2017)

    2017

  17. [25]

    Yunoki, Numerical study of the spin-flop transition in anisotropic spin-1/2 antiferromagnets, Phys

    S. Yunoki, Numerical study of the spin-flop transition in anisotropic spin-1/2 antiferromagnets, Phys. Rev. B65, 092402 (2002)

    2002

  18. [26]

    Díez-Valle, D

    P. Díez-Valle, D. Porras, and J. J. García-Ripoll, Quantum Approximate Optimization Algorithm Pseudo- Boltzmann States, Phys. Rev. Lett.130, 050601 (2023)

    2023

  19. [27]

    P. C. Lotshaw, G. Siopsis, J. Ostrowski, R. Herrman, R. Alam, S. Powers, and T. S. Humble, Approximate Boltzmann distributions in quantum approximate opti- mization, Phys. Rev. A108, 042411 (2023)

    2023

  20. [28]

    J. M. Silvester, G. Carleo, and S. R. White, Unusual en- ergy spectra of matrix product states, arXiv:2408.13616 (2024)

    2024 arXiv

  21. [29]

    Yokoyama and H

    H. Yokoyama and H. Shiba, Variational Monte-Carlo Studies of Hubbard Model. I, J. Phys. Soc. Japan56, 1490 (1987)

    1987

  22. [30]

    Zhang, Z.-Q

    S.-X. Zhang, Z.-Q. Wan, C.-K. Lee, C.-Y. Hsieh, S. Zhang, and H. Yao, Variational Quantum-Neural Hy- brid Eigensolver, Phys. Rev. Lett.128, 120502 (2022)

    2022

  23. [31]

    X. Yuan, J. Sun, J. Liu, Q. Zhao, and Y. Zhou, Quantum Simulation with Hybrid Tensor Networks, Phys. Rev. Lett. 127, 040501 (2021)

    2021

  24. [32]

    Effective Temperature in Approximate Quantum Many-body States

    G. E. Crooks, Gradients of parameterized quantum gates using the parameter-shift rule and gate decomposition, arXiv:1905.13311 (2019). 9 Supplemental Material for “Effective Temperature in Approximate Quantum Many-body States” I. TRAINING DYNAMICS FOR EFFECTIVE TEMPERATURE TOW...

    2019 arXiv

  25. [33]

    Optimized approximate ground states admit very small˜βs indicating a nearly flat spectrum across excited states

  26. [34]

    Moreover, a smaller˜β often indicates a better accuracy at the late stage of training

    During training, ˜β first increases and then decreases in general, implying an upper bound for the possible˜β of a given ansatz. Moreover, a smaller˜β often indicates a better accuracy at the late stage of training. Besides, initial ˜β ≈ 0 unless the ansatz has a strong physic...

  27. [35]

    The energy objective imposes unequal penalties for different excited state components, i.e

    For each method, the training dynamics is qualitatively similar for both energy and fidelity objectives though ˜β values are typically smaller when optimizing fidelity. The energy objective imposes unequal penalties for different excited state components, i.e. the higher energ...

  28. [36]

    VMC of different widths or tensor networks with different bond dimensions, usually display similar training dynamics

    The same family of ansatzes with different hyperparameters, eg. VMC of different widths or tensor networks with different bond dimensions, usually display similar training dynamics. This observation further supports the conclusion that effective temperature can reflect the int...

  29. [37]

    Many flat spectrum plateaus give|ci|2 in the order 10−4 ∼ 10−8, which are numerically significant and much higher than the machine precision floor10−16 for double precision (complex128) arithmetic

  30. [38]

    S10, demonstrate distinct overlap behaviors for the same energy levels across different targetβ values

    Many approximate ITES instances exhibit the two-stage behavior such as Fig. S10, demonstrate distinct overlap behaviors for the same energy levels across different targetβ values. For example, points with target overlap 11 10 4 10 2 100 (E − E0)/|E0| 0.0 0.1 0.2 ˜β (a) VQE (d ...

  31. [39]

    This fact further implies that the effective temperature is numerically meaningful

    The critical effective temperatureβ∗s differ among different ansatzes. This fact further implies that the effective temperature is numerically meaningful. If the underlying mechanism were solely the machine precision floor,β∗ would be the same across different methods. The spe...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.