Pith. sign in

REVIEW 3 major objections 4 minor 3 cited by

MPS-JuliQAOA: User-friendly, Scalable MPS-based Simulation for Quantum Optimization

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper claims MPS-JuliQAOA can simulate QAOA at 512 qubits and 20 rounds on the standard 3-regular MaxCut benchmark, and that users can do so without writing tensor-network code.

desk verdict A plausible open-source MPS-QAOA simulator whose headline 512-qubit/p=20 scaling claim is unverifiable from the abstract alone; worth refereeing, but the authors need to release the accuracy data. read the letter →

arxiv 2508.05883 v1 pith:65EQFCZE submitted 2025-08-07 quant-ph cs.ETcs.SE

Sean Feeney , Reuben Tate , John Golden , Stephan Eidenbenz This is my paper
classification quant-phcs.ETcs.SE
keywords QuantumApproximateOptimizationAlgorithmMatrixProductStateMaxCutTensorNetworkSimulationJuliaparameteropen-sourcesimulator
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 presents MPS-JuliQAOA, an open-source simulator for the Quantum Approximate Optimization Algorithm (QAOA). Instead of storing the full quantum state as a vector of amplitudes, it represents the state as a matrix product state, a chain of small tensors whose size is controlled by a bond dimension. The central claim is that this representation makes QAOA simulation scale to 512 qubits and 20 rounds on the standard de-facto benchmark, 3-regular MaxCut, while keeping runtime and memory manageable. The paper also claims that built-in parameter finding works inside the same tensor-network simulator, so users can optimize QAOA angles without learning MPS or automatic-differentiation details. If true, this makes large-scale QAOA experiments practical for a much wider group of researchers.

What carries the argument

The central object is the Matrix Product State (MPS) representation of the QAOA wavefunction, a chain of local tensors whose connecting bond dimension is truncated to keep memory and runtime bounded. The QAOA circuit's alternating phase-separator and mixer layers are applied locally to this tensor chain, with singular-value truncation after each step. Because the cost Hamiltonian is diagonal, phase-separator gates act on local qubit terms, which is what keeps the entanglement manageable enough for the MPS approach to scale. Built-in parameter finding runs automatic differentiation through the same tensor network, making angle optimization part of the simulation workflow.

What would settle it

Run the released simulation at 512 qubits and 20 rounds on the same 3-regular MaxCut benchmark with bond dimension doubled, and check whether the estimated MaxCut expectation value and optimized angles change substantially. If the result shifts beyond the stated accuracy tolerance when bond dimension is increased, the central scalability claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that an MPS-based simulation backend, integrated into the JuliQAOA package, lets QAOA simulation scale effortlessly to 512 qubits and 20 simulation rounds on the standard 3-regular MaxCut benchmark. The simulator accepts any optimization problem expressible as a diagonal Hamiltonian, so MaxCut is one demonstration among many. The authors further claim that the tool hides the complexity of matrix product states and automatic differentiation, so a user can define a problem, run the simulation, and perform parameter optimization without specialized tensor-network knowledge. The discovery is both a practical scale frontier for MPS-based QAOA simulation and an enginee

Load-bearing premise

The load-bearing premise is that the matrix-product-state representation of the QAOA state at 512 qubits and 20 rounds stays low-entanglement enough that truncating the bond dimension produces accurate expectation values.

Editorial extensions

If this is right

  • If the claims hold, QAOA simulations at 512 qubits and 20 rounds become routine on a single workstation, a regime far beyond dense state-vector simulators.
  • Researchers can run full parameter-optimization loops inside the MPS simulation rather than treating simulation and angle search as separate tasks.
  • Because the simulator accepts any problem with a diagonal Hamiltonian, the approach extends beyond MaxCut to other combinatorial optimization problems.
  • The reported runtime, memory, and accuracy tradeoffs give users a practical way to choose bond dimension based on available resources and desired fidelity.

Reading between the lines

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

  • If QAOA states on bounded-degree graphs keep entanglement low at moderate depth, the demonstrated 512-qubit runs may understate the method's practical reach; how far it extends to denser or more entangled problems remains untested.
  • A natural next test is to compare MPS expectation values at fixed bond dimension against classical semidefinite-programming bounds on randomized 3-regular MaxCut instances; large deviations would reveal when truncation stops being safe.
  • The same backend could plausibly support warm-started or recursion-style QAOA variants, but the paper does not address whether those variants preserve the low-entanglement property that makes MPS simulation accurate.
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. MPS-JuliQAOA is presented as an open-source Julia/ITensor-based matrix-product-state simulator for QAOA on arbitrary diagonal Hamiltonians. The central claim is that it "effortlessly scales to 512 qubits and 20 simulation rounds" on 3-regular MaxCut QAOA, while also offering built-in parameter optimization and a user-friendly interface that hides MPS and automatic differentiation details. The paper further states that runtime, memory usage, and accuracy tradeoffs are studied. The supplied full text is largely corrupted and unreadable, so the detailed derivations, figures, and tables could not be inspected; the review therefore focuses on the abstract and on the explicitly stated evidence requirements.

Significance. If the 512-qubit, p=20 benchmark were supported by controlled numerical data, this would be a useful practical contribution: a scalable, open-source MPS simulator with parameter finding would lower the barrier to benchmarking QAOA at system sizes far beyond exact statevector simulation, and the chosen 3-regular MaxCut benchmark is a standard and meaningful testbed. The release of source code is a concrete strength, and no circular reasoning is apparent: the scaling claim is empirical rather than derived from an assumption that already contains the answer. However, the manuscript as supplied does not provide the quantitative evidence needed to assess the magnitude or validity of the central claim, so the significance is currently prospective rather than established.

major comments (3)
  1. [Abstract] The central claim, "effortlessly scales to 512 qubits and 20 simulation rounds on ... 3-regular MaxCut QAOA problem," is an empirical assertion with no supporting numbers. No runtime, memory, bond dimension, discarded weight, final energy/approximation ratio, or error bars are reported. The sentence "We study the scalability ... accuracy tradeoffs" promises these data but does not give them. Please add a benchmark table/figure for the claimed 512-qubit, p=20 case, including instance-to-instance statistics and hardware/software details; otherwise the headline claim is not established.
  2. [Abstract / MPS truncation] The scalability claim requires that the MPS bond dimension stay manageable at depth 20. For a QAOA state at depth p, the Schmidt rank across a cut can grow with p (worst case ~2^p), so uncontrolled truncation could make the 512-qubit result numerically meaningless. The paper must report the maximum bond dimension and truncation error at p=20, show convergence of the expected energy as the bond dimension increases, and validate the implementation against exact statevector results at smaller n or previously published QAOA data. Without these checks, the claim of "acceptable" accuracy is unsupported.
  3. [Full text / parameter finding] The supplied manuscript body is heavily corrupted and appears to contain a header for a different arXiv identifier (arXiv:2508.05876v2 [cs.LG]); equations, figures, and tables are unreadable, so the detailed results and parameter-finding description could not be verified. The abstract also claims "built-in parameter finding capabilities" without specifying the method, number of restarts, or demonstration that optimized parameters were obtained for the claimed 512-qubit, p=20 benchmark. If these data exist in the original, they are not accessible in the supplied copy. Please resubmit a readable version and provide these details.
minor comments (4)
  1. [Abstract] The tool name is inconsistent: "MPS-JuliQAOA" in the title and first sentence versus "MPS-Juli-QAOA" later. Please unify.
  2. [Abstract] Define "simulation rounds" (presumably QAOA depth p) and state how the 3-regular MaxCut instances were generated (fixed instance? random ensemble? seed?).
  3. [Introduction / Related work] Add references for the 3-regular MaxCut QAOA benchmark and for prior MPS/tensor-network QAOA simulators so the novelty is positioned clearly.
  4. [Reproducibility] The phrase "effortlessly scales" is informal; suggest a quantitative statement such as "we demonstrate ... with controlled truncation error." Also pin the GitHub branch to a commit hash or release version for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper reports measured simulator scalability, not a derived result that reduces to its inputs.

full rationale

This is a software/benchmark paper. The central claim—that MPS-JuliQAOA scales to 512 qubits and 20 QAOA rounds on 3-regular MaxCut—is presented as an empirical scalability result (runtime, memory, accuracy tradeoffs), not as a mathematical derivation from assumptions that already contain the answer. No fitted parameter is renamed as a prediction, no target quantity is defined in terms of the simulator's output, and the accuracy caveat is explicitly acknowledged in the abstract rather than hidden. Although the full text is garbled in the provided version, the visible abstract and fragments show no equation that reduces a predicted quantity to its own input, and no load-bearing uniqueness theorem or ansatz is smuggled in via self-citation. Concerns about whether truncation error at p=20 remains acceptable are legitimate correctness/validation concerns, but they are not circularity. The paper is self-contained as an empirical tool demonstration, so the circularity score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

No new particles or forces are introduced. The only free parameter is the MPS bond dimension/truncation tolerance, which is a standard numerical parameter in tensor-network simulations. The crucial unverified premise is that truncation errors are controllable at the claimed scale.

free parameters (1)
  • MPS bond dimension / truncation cutoff
    MPS simulation requires choosing a maximum bond dimension or truncation error tolerance. This is a user-adjustable parameter that controls the tradeoff between runtime, memory, and accuracy. The abstract says accuracy tradeoffs are studied but gives no default or fitted values.
assumptions (3)
  • domain assumption The QAOA circuit can be represented as a diagonal Hamiltonian for any optimization problem.
    Abstract: 'any optimization problem that can be expressed as diagonal Hamiltonian.' This is the standard QAOA formulation.
  • domain assumption Julia and ITensor provide numerically correct MPS operations.
    The tool builds on these libraries; correctness of the simulator depends on their correctness.
  • ad hoc to paper Bond dimension truncation errors remain acceptable at 512 qubits and 20 rounds.
    This is the key enabling assumption for the scalability claim; no evidence in the abstract.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MPS-JuliQAOA: User-friendly, Scalable MPS-based Simulation for Quantum Optimization." pith.science (2026). https://pith.science/paper/65EQFCZE

@misc{pith2026250805883,
  author       = {Pith},
  title        = {Pith review of: MPS-JuliQAOA: User-friendly, Scalable MPS-based Simulation for Quantum Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/65EQFCZE}},
  note         = {Machine review of arXiv:2508.05883}
}
read the original abstract

We present the MPS-JuliQAOA simulator, a user-friendly, open-source tool to simulate the Quantum Approximate Optimization Algorithm (QAOA) of any optimization problem that can be expressed as diagonal Hamiltonian. By leveraging Julia-language constructs and the ITensor package to implement a Matrix Product State (MPS) approach to simulating QAOA, MPS-Juli-QAOA effortlessly scales to 512 qubits and 20 simulation rounds on the standard de-facto benchmark 3-regular MaxCut QAOA problem. MPS-JuliQAOA also has built-in parameter finding capabilities, which is a crucial performance aspect of QAOA. We illustrate through examples that the user does not need to know MPS principles or complex automatic differentiation techniques to use MPS-JuliQAOA. We study the scalability of our tool with respect to runtime, memory usage and accuracy tradeoffs. Code available at https://github.com/lanl/JuliQAOA.jl/tree/mps.

Discussion (0). Sign in to comment.

Forward citations

Cited by 3 Pith papers

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

  1. Quantum Variational Approaches to the Maximum Independent Set Problem at Utility Scale

    quant-ph 2026-06 unverdicted novelty 7.0 of 10

    Variational quantum methods with spectral preprocessing, CVaR optimization, and ancilla-assisted superposition solve maximum independent set to optimality on graphs up to 180 vertices, claimed as the largest such gate...

  2. Quantum Variational Approaches to the Maximum Independent Set Problem at Utility Scale

    quant-ph 2026-06 unverdicted novelty 6.0 of 10

    Variational quantum methods with spectral reordering, sparsification, CVaR optimization, and ancilla-assisted superposition solve MIS to optimality on 64-, 99-, and 180-vertex graphs, the largest such gate-based demon...

  3. Quantum Variational Approaches to the Maximum Independent Set Problem at Utility Scale

    quant-ph 2026-06 conditional novelty 6.0 of 10

    Ancilla-assisted multi-seed superposition plus excitation-preserving VQE and classical post-processing recovers exact MIS on 64-, 99-, and 180-node graphs, with partial hardware transfer.

Reference graph

Works this paper leans on

32 extracted references · 14 canonical work pages · cited by 1 Pith paper

  1. [1]

    11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 = #1 \@IEEEnotcompsoconly \@IEEEcompsoconly #1 * [1] 0pt [0pt][0pt] #1 * [1] 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEcompsocnotconfonly \@IEEEauthorblockAstyle \@IEEEcompsocnotconfonly \@IEEEcompsocconfonly \@IEEEauthordefaulttextstyle \@IEEEcompsocnotconfonly \@IEEEauthor...

    work page

  2. [2]

    Di Meglio, K

    A. Di Meglio, K. Jansen, I. Tavernelli, C. Alexandrou, S. Arunachalam, C. W. Bauer, K. Borras, S. Carrazza, A. Crippa, V. Croft et al., ``Quantum computing for high-energy physics: State of the art and challenges,'' PRX Quantum, vol. 5, no. 3, p. 037001, 2024

    work page 2024

  3. [3]

    Y. Cao, J. Romero, J. P. Olson, M. Degroote, P. D. Johnson, M. Kieferov \'a , I. D. Kivlichan, T. Menke, B. Peropadre, N. P. Sawaya et al., ``Quantum chemistry in the age of quantum computing,'' Chemical reviews, vol. 119, no. 19, pp. 10\,856--10\,915, 2019

    work page 2019

  4. [4]

    X. Xu, S. Benjamin, J. Sun, X. Yuan, and P. Zhang, ``A herculean task: Classical simulation of quantum computers,'' 2023. [Online]. Available: https://arxiv.org/abs/2302.08880

    arXiv 2023

  5. [5]

    Shor, ``Algorithms for quantum computation: discrete logarithms and factoring,'' in Proceedings 35th Annual Symposium on Foundations of Computer Science, 1994, pp

    P. Shor, ``Algorithms for quantum computation: discrete logarithms and factoring,'' in Proceedings 35th Annual Symposium on Foundations of Computer Science, 1994, pp. 124--134

    work page 1994

  6. [6]

    L. K. Grover, ``A fast quantum mechanical algorithm for database search,'' in Proceedings of the Twenty-Eighth Annual ACM Symposium on Theory of Computing, ser. STOC '96. 1em plus 0.5em minus 0.4em New York, NY, USA: Association for Computing Machinery, 1996, p. 212–219. [Online]. Available: https://doi.org/10.1145/237814.237866

    arXiv 1996

  7. [7]

    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,'' Nature Communications, vol. 5, no. 1, Jul. 2014. [Online]. Available: http://dx.doi.org/10.1038/ncomms5213

    doi:10.1038/ncomms5213 2014

  8. [8]

    Farhi, J

    E. Farhi, J. Goldstone, and S. Gutmann, ``A quantum approximate optimization algorithm,'' 2014. [Online]. Available: https://arxiv.org/abs/1411.4028

    arXiv 2014

Show all 32 references
  1. [9]

    Cerezo, A

    M. Cerezo, A. Arrasmith, R. Babbush, S. C. Benjamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio et al., ``Variational quantum algorithms,'' Nature Reviews Physics, vol. 3, no. 9, pp. 625--644, 2021

    2021

  2. [10]

    Berezutskii, M

    A. Berezutskii, M. Liu, A. Acharya, R. Ellerbrock, J. Gray, R. Haghshenas, Z. He, A. Khan, V. Kuzmin, D. Lyakh et al., ``Tensor networks for quantum computing,'' arXiv preprint arXiv:2503.08626, 2025

    2025

  3. [11]

    Verstraete and J

    F. Verstraete and J. I. Cirac, ``Renormalization algorithms for quantum-many body systems in two and higher dimensions,'' arXiv preprint cond-mat/0407066, 2004

    2004 arXiv

  4. [12]

    Cincio, J

    L. Cincio, J. Dziarmaga, and M. M. Rams, ``Multiscale entanglement renormalization ansatz in two dimensions: quantum ising model,'' Physical review letters, vol. 100, no. 24, p. 240603, 2008

    2008

  5. [13]

    S. R. White, ``Density matrix formulation for quantum renormalization groups,'' Phys. Rev. Lett., vol. 69, pp. 2863--2866, Nov 1992. [Online]. Available: https://link.aps.org/doi/10.1103/PhysRevLett.69.2863

    1992 doi

  6. [14]

    Paeckel, T

    S. Paeckel, T. Köhler, A. Swoboda, S. R. Manmana, U. Schollwöck, and C. Hubig, ``Time-evolution methods for matrix-product states,'' Annals of Physics, vol. 411, p. 167998, 2019. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0003491619302532

    2019

  7. [15]

    Golden, A

    J. Golden, A. Baertschi, D. O’Malley, E. Pelofske, and S. Eidenbenz, ``Juliqaoa: Fast, flexible qaoa simulation,'' in Proceedings of the SC ’23 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, ser. SC-W 2023. 1em plus 0.5...

    2023

  8. [16]

    Javadi-Abhari, M

    A. Javadi-Abhari, M. Treinish, K. Krsulich, C. J. Wood, J. Lishman, J. Gacon, S. Martiel, P. D. Nation, L. S. Bishop, A. W. Cross, B. R. Johnson, and J. M. Gambetta, ``Quantum computing with Q iskit,'' 2024

    2024

  9. [17]

    T. Bode, D. Bagrets, A. Misra-Spieldenner, T. Stollenwerk, and F. K. Wilhelm, ``Qaoa.jl: Toolkit for the quantum and mean-field approximate optimization algorithms,'' Journal of Open Source Software, vol. 8, no. 86, p. 5364, 2023. [Online]. Available: https://doi.org/10.21105/...

    2023 doi

  10. [18]

    Misra-Spieldenner, T

    A. Misra-Spieldenner, T. Bode, P. K. Schuhmacher, T. Stollenwerk, D. Bagrets, and F. K. Wilhelm, ``Mean-field approximate optimization algorithm,'' PRX Quantum, vol. 4, p. 030335, Sep 2023. [Online]. Available: https://link.aps.org/doi/10.1103/PRXQuantum.4.030335

    2023 doi

  11. [19]

    Fishman, S

    M. Fishman, S. R. White, and E. M. Stoudenmire, `` The ITensor Software Library for Tensor Network Calculations ,'' SciPost Phys. Codebases, p. 4, 2022. [Online]. Available: https://scipost.org/10.21468/SciPostPhysCodeb.4

    2022 doi

  12. [20]

    Lykov, A

    D. Lykov, A. Chen, H. Chen, K. Keipert, Z. Zhang, T. Gibbs, and Y. Alexeev, ``Performance evaluation and acceleration of the qtensor quantum circuit simulator on gpus,'' in 2021 IEEE/ACM Second International Workshop on Quantum Computing Software (QCS). 1em plus 0.5em minus 0....

    2021

  13. [21]

    J. C. Bridgeman and C. T. Chubb, ``Hand-waving and interpretive dance: an introductory course on tensor networks,'' Journal of Physics A: Mathematical and Theoretical, vol. 50, no. 22, p. 223001, May 2017. [Online]. Available: http://dx.doi.org/10.1088/1751-8121/aa6dc3

    2017 doi

  14. [22]

    Biamonte and V

    J. Biamonte and V. Bergholm, ``Tensor networks in a nutshell,'' 2017. [Online]. Available: https://arxiv.org/abs/1708.00006

    2017 arXiv

  15. [23]

    Orús, ``A practical introduction to tensor networks: Matrix product states and projected entangled pair states,'' Annals of Physics, vol

    R. Orús, ``A practical introduction to tensor networks: Matrix product states and projected entangled pair states,'' Annals of Physics, vol. 349, p. 117–158, Oct. 2014. [Online]. Available: http://dx.doi.org/10.1016/j.aop.2014.06.013

    2014 doi

  16. [24]

    Verstraete and J

    F. Verstraete and J. I. Cirac, ``Matrix product states represent ground states faithfully,'' Physical Review B, vol. 73, no. 9, Mar. 2006. [Online]. Available: http://dx.doi.org/10.1103/PhysRevB.73.094423

    2006 doi

  17. [25]

    Feeney, R

    S. Feeney, R. Tate, and S. Eidenbenz, ``The better solution probability metric: Optimizing qaoa to outperform its warm-start solution,'' 2024. [Online]. Available: https://arxiv.org/abs/2409.09012

    2024 arXiv

  18. [26]

    Kus and M

    G. Kus and M. A. Bessa, ``Gradient-free neural topology optimization: Towards effective fracture-resistant designs,'' 2024. [Online]. Available: https://arxiv.org/abs/2403.04937

    2024 arXiv

  19. [27]

    A. G. Baydin, B. A. Pearlmutter, A. A. Radul, and J. M. Siskind, ``Automatic differentiation in machine learning: a survey,'' 2018. [Online]. Available: https://arxiv.org/abs/1502.05767

    2018 arXiv

  20. [28]

    Fang, H.-Z

    Y.-H. Fang, H.-Z. Lin, J.-J. Liu, and C.-J. Lin, ``A step-by-step introduction to the implementation of automatic differentiation,'' 2024. [Online]. Available: https://arxiv.org/abs/2402.16020

    2024 arXiv

  21. [29]

    C. Guo, Y. Fan, Z. Xu, and H. Shang, ``Differentiable matrix product states for simulating variational quantum computational chemistry,'' Quantum, vol. 7, p. 1192, Dec. 2023. [Online]. Available: http://dx.doi.org/10.22331/q-2023-12-04-1192

    2023 doi

  22. [30]

    Hadfield, ``On the representation of boolean and real functions as hamiltonians for quantum computing,'' ACM Transactions on Quantum Computing, vol

    S. Hadfield, ``On the representation of boolean and real functions as hamiltonians for quantum computing,'' ACM Transactions on Quantum Computing, vol. 2, no. 4, pp. 1--21, 2021

    2021

  23. [31]

    Lucas, ``Ising formulations of many np problems,'' Frontiers in physics, vol

    A. Lucas, ``Ising formulations of many np problems,'' Frontiers in physics, vol. 2, p. 5, 2014

    2014

  24. [32]

    Schollwöck, ``The density-matrix renormalization group in the age of matrix product states,'' Annals of Physics, vol

    U. Schollwöck, ``The density-matrix renormalization group in the age of matrix product states,'' Annals of Physics, vol. 326, no. 1, pp. 96--192, 2011, january 2011 Special Issue. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0003491610001752

    2011

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.