Pith. sign in

REVIEW 3 major objections 5 minor 58 references

Quantum Interference as a Proposal Mechanism for Combinatorial Optimization

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Pith's one-line read Quantum Interference Proposal Search claims that finite-shot sampling from seed-conditioned two-layer circuits keeps a classical elite frontier competitive with a strong classical kick-and-repair search at equal proposal budgets.

desk verdict Careful, honest, novel proposal-search study whose core 'competitive at matched proposal budget' claim holds up — but the abstract's 'resource-efficient' overreaches, since only objective evaluations are counted, not circuit cost. read the letter →

arxiv 2607.27509 v1 pith:U6ZYSWVY submitted 2026-07-29 quant-ph

classification quant-ph MSC 68Q1290C2781P68 PACS 03.67.Ac
keywords quantumoptimizationQUBO/Isingmodelsfinite-shotsamplinglocalizedinterferenceproposalsearchseed-conditionedcircuitsnon-variationalalgorithmresource-awarebenchmarking
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 localized quantum interference can work as a resource-efficient proposal mechanism for combinatorial optimization, rather than as a variational state-preparation shortcut. The algorithm, QIPS, samples candidate bitstrings from shallow seed-conditioned circuits, scores them classically under the QUBO/Ising cost (a quadratic binary energy function), and feeds the winners back into an elite frontier; across six such benchmark families with 18 to 29 variables, that loop stays competitive with a matched classical kick-and-repair search that consumes the same number of proposals. The point of the claim is practical: it suggests a route to useful gate-based optimization that avoids variational parameter training, using only a fixed two-layer circuit, 100 shots per circuit, and a classical outer search. The paper is explicit that this does not establish quantum advantage; what it identifies is a distinct finite-shot proposal profile—repeated access to near-optimal states—that differs qualitatively from the classical baseline.

What carries the argument

The load-bearing object is the seed-conditioned two-layer circuit |ψ⟩ = U_B2 U_K2 U_B1 U_K1 |+⟩^{⊗ n_b}, where each layer applies a diagonal cost-phase operator exp(−iγ(K + D)) followed by parallel single-qubit rotations about X, Y, and Z axes. The seed bitstring enters through the canonical second-layer angles θ_2 = (π/2) b_j^(seed), creating a seed-centered localization pattern; stochastic deviations around the canonical angles are regulated by a Metropolis controller whose pseudo-energy depends only on seed multiplicity and the number of distinct measured bitstrings in the 100-shot record. The cost operator K supplies problem-dependent multi-qubit phases, and a static degeneracy-breaking

What would settle it

Run the same six-benchmark comparison under an end-to-end cost model that charges QIPS for gate count, circuit depth, state-vector simulation, and decoherence in addition to proposal evaluations; if QIPS's physical cost exceeds the classical proposal cost by more than a constant factor while top-K recovery is merely comparable, the resource-efficiency claim is falsified. A complementary control is the paper's own state-label shuffle, which should reduce QIPS to blind search; a setting where it does not would falsify the claimed dependence on QUBO structure.

Watch

Extended reading notes

Core claim

The central discovery is that the measurement record of a localized seed-conditioned quantum circuit can itself be a search resource. For QUBO/Ising objectives, a seed-encoded two-layer circuit produces probability on the seed plus a few detectable non-seed states; a feedback controller tunes randomized angle deviations to keep that localization. In ideal simulation across six benchmark families, QIPS stays competitive with a matched classical kick-and-repair proposer at equal proposal budget and repeatedly resamples near-optimal states. Shuffling the bitstring-to-energy mapping destroys the structure, reducing QIPS to blind search.

Load-bearing premise

The load-bearing premise is the Methods resource accounting that treats the number of 100-shot proposals as the only matched resource: if the physical cost of implementing the two-layer circuit—gate count, depth, simulation overhead, or hardware noise—is orders of magnitude larger than a classical kick-and-repair proposal, the 'resource-efficient' claim fails even though the benchmark curves are correct.

Editorial extensions

If this is right

  • Optimization can proceed without variational training: useful progress comes from an ensemble of localized circuits, not from optimizing parameters of a single state.
  • At a total budget of 2000 n_b proposals, QIPS keeps a classical frontier competitive with a strong classical kick-and-repair search on sparse constraint problems, weighted MaxCut, exponential-disorder Ising, and Sherrington–Kirkpatrick instances up to n_b = 29.
  • Repeated sampling of near-optimal states is an empirical finite-shot signature; a declining rate of novel frontier updates can serve as an early-stopping diagnostic.
  • Lower Hilbert-space coverage is not a failure by itself: QIPS trades breadth for concentration of probability in the low-energy tail, so coverage must be judged together with hit rate and multiplicity.
  • The paper stops short of claiming quantum advantage; its stated next step is an experimental implementation on gate-based hardware with matched end-to-end resources and noise.

Reading between the lines

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

  • Editorial inference: An end-to-end resource model that charges gate count, circuit depth, and decoherence could change the resource-efficiency verdict; the paper's accounting treats physical implementation cost as negligible.
  • Editorial inference: The approximate seed-rank symmetry CDF(u) ≈ 1 − CDF(1 − u) reported for QIPS is a testable signature that could distinguish quantum proposals from classical proposals at larger sizes and possibly serve as a coherence diagnostic.
  • Editorial inference: Because classical proposals are strongest early in descent and QIPS is strongest near the ground state, a structured hybrid that uses quantum proposals mainly at low-rank seeds might outperform either method; the paper's simple alternation did not help, but that is a different schedule.
  • Editorial inference: Dynamic cost-operator jitter is an adjustable knob the paper credits with diversifying peaks; ablating jitter amplitude while holding feedback fixed would quantify how much of QIPS depends on this perturbation rather than bare two-layer interference.
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 proposes Quantum Interference Proposal Search (QIPS), a non-variational, seed-conditioned quantum circuit proposal generator for QUBO/Ising optimization. A fixed two-layer circuit is measured 100 times per seed, and the measured bitstrings are scored classically and used to update an elite frontier. A feedback controller based only on seed multiplicity and number of distinct measured states maintains localization. The paper compares QIPS against a matched classical kick-and-repair proposal search on six benchmark families, with 32 instances per family and size, 18 <= n_b <= 29, and a fixed budget of 2000 n_b objective evaluations. It reports competitive top-K coverage, higher repeated sampling of near-optimal states, lower Hilbert-space coverage, and dyadic-rank structure that becomes relatively more favorable as seeds approach the ground state. The paper concludes that localized quantum interference is a resource-efficient proposal mechanism.

Significance. The empirical protocol is a genuine strength: the outer loop, frontier update rule, seed selection, and proposal budget are matched between quantum and classical controls; blind-search and label-shuffling controls are included; results are aggregated over six benchmark families with 32 instances per family/size; and the Supplementary Information provides system-resolved data. If interpreted as a proposal-budget comparison, the paper provides a carefully executed falsifiable study of a finite-shot, non-variational quantum proposal generator. The finding that QIPS can repeatedly sample low-energy states while retaining rank-improving proposals is a concrete, testable signature. However, the headline claim that QIPS is a 'resource-efficient proposal mechanism' is not supported by the resource accounting used: the comparison controls only objective evaluations, not circuit implementation cost, wall-clock time, or hardware overhead. The paper is also fully dependent on unreleased custom simulation code.

major comments (3)
  1. [Methods 'Outer search loop and resource accounting'; Abstract] The headline claim 'resource-efficient proposal mechanism' is not supported by the resource convention used. The Methods define N_prop = 100 x 20 n_b and state 'This convention treats sampling effort as the shared resource.' The comparison therefore controls only objective evaluations, not the cost of generating proposals. A QIPS proposal requires running a two-layer circuit 100 times; for the complete-graph families (SK, exponential weak QD), the diagonal cost operator has O(n_b^2) terms, so each phase layer decomposes into O(n_b^2) two-qubit gates, and 20 n_b circuits give O(n_b^3) gates total. The matched classical kick-and-repair generator has no comparable sampling overhead. Since the Discussion itself calls for 'matched end-to-end resources' as the decisive next step, either the abstract should say 'proposal-budget-competitive' or the paper should include a gate-count/depth and wal
  2. [Supp. Note 10, Eqs. (58)-(76); Results 'Emergent structure'] The 'localized interference' is not purely emergent. The feedback controller's pseudo-energy E_QIP = E_DS(n_seed) + E_NU(n_unique) and the qualification rule (for 100 shots, n_seed >= 10 and 11 <= n_unique <= 61) explicitly select circuits that return the seed often and produce few distinct states. Thus the high seed multiplicity and limited support in Figs. 3a-d, and the repeated near-optimal sampling in Fig. 4e-f, are partly constructed by the algorithm's internal objective rather than being independent consequences of quantum interference. The search comparison remains valid because the pseudo-energy does not use the QUBO objective, but the word 'emergent' should be calibrated and the paper should clearly state that localization is a controller-specified target.
  3. [Code availability] The empirical claims rest on a custom simulation code that is 'not publicly released with this preprint.' The Supplementary Information gives an algorithm summary, but not a complete reference implementation for the accelerated sparse-probability sampling, jitter schedule, or feedback controller. Without code or a deterministic reference implementation, the aggregate curves and representative runs cannot be independently checked. Releasing the code, or providing a complete pseudo-code reference with all hyperparameters, should be a condition for the claims as stated.
minor comments (5)
  1. [Results, Fig. 3d] The statement that the extended CDF 'fits markedly well' to a two-parameter probit is not supported by any goodness-of-fit statistic or reported parameter ranges. Please add e.g. Kolmogorov-Smirnov distances or quantile-quantile summaries.
  2. [Fig. 5 caption] Error bars are omitted from all panels of Fig. 5. Since the text claims QIPS shows 'the strongest relative behavior' for the best-ranked seeds, at least one panel should include uncertainty so this claim can be assessed.
  3. [Eq. (10) and Fig. 3d] The condition q_E < 0.1 is used in the main text before q_E is formally defined in Eq. (10). Move the definition earlier or state it in the figure caption.
  4. [Fig. 4 axis] The x-axis 'Number of Rounds' is ambiguous for the classical control, because a classical 'proposal step' can consume a variable number of repair-neighbor evaluations. Please clarify how rounds are defined for the classical search.
  5. [Table 1] Minor typographical issues: 'T able 1' in the table heading and inconsistent use of 'n b' versus n_b. The title and abstract also use 'n_b' without defining the subscript consistently.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor constructed-localization circularity; central matched benchmark is independent.

  1. self definitional [Methods: 'Feedback control of localized interference'; Results: Fig. 4e,f; Supplementary Note 10, Eq. (58)-(76)]
    "These two quantities define a pseudo-energy objective that favors intermediate seed multiplicity and a finite number of distinct outcomes. ... A useful quantum interference pattern must return the seed repeatedly ... whereas QIPS exhibits greater repeated sampling of low-energy states as the frontier matures."

    The repeated-sampling signature is not independent evidence: the pseudo-energy E_QIP = E_DS(n_seed) + E_NU(n_unique) explicitly rewards n_seed in [10,70] (Supp. Note 10, Eq. 59), so circuits that return the seed are selected by the controller, and seeds are drawn from the low-energy elite frontier. Hence 'QIPS resamples near-optimal states much more frequently' follows from the feedback objective plus low-energy seed selection by construction; it is partly a restatement of what the controller defines as a 'useful' pattern. The matched top-K coverage/hit-rate benchmarks, by contrast, are not constructed by this pseudo-energy and remain independent of this step.

full rationale

The central empirical claim is self-contained and not a disguised fit: the same outer-loop protocol, frontier rule, and proposal budget are applied to QIPS and to a classical kick-and-repair control, and the QUBO energy is not used by the feedback pseudo-energy. There are no self-citations and no fitted parameters relabeled as predictions. The abstract's 'resource-efficient' wording is explicitly tied to the paper's own accounting convention ('This convention treats sampling effort as the shared resource'), and the paper itself discloses that end-to-end hardware cost is not assessed ('The physical gate depth required to implement the diagonal cost operator depends on the problem graph, native gate set and hardware connectivity' and the call for benchmarking 'under matched end-to-end resources'). That is a stated limitation and a correctness risk, not a hidden circularity. The only notable circular element is the descriptive multiplicity result, which is partly engineered by the feedback controller's seed-return objective; this does not undermine the independent matched benchmark comparison, so the overall circularity is minor.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a stack of hand-tuned algorithmic constants (feedback pseudo-energy, canonical angles, deviation scales, frontier sizes, classical baseline hyperparameters) and on the modeling assumptions that ideal simulation, proposal-count accounting, and the probit/uniform approximations faithfully represent a real device. No new physical entities are introduced.

free parameters (7)
  • Pseudo-energy well coefficients (30, 7000, 20, 12) and Monte Carlo temperature kT=20 = 30, 7000, 20, 12; kT=20
    Hand-selected to make the localization objective prefer intermediate seed multiplicity and unique-outcome counts; central to producing the 'localized interference' that QIPS claims to exploit (Supp. Note 10, Eq. 58–76).
  • Feedback qualification target f_target_QIP = 0.8
    Prescribed fraction of circuits that must qualify as localized; if changed, the proposal distribution can become more or less localized (Supp. Note 10).
  • Canonical angle parameters (theta1=π/4, alpha=π/2, theta2=π/2 * b_seed, zero delta/phi/gamma) = π/4, π/2, 0
    Seed encoding and localization anchor chosen by design; not derived from optimization principle (Methods, 'Seed-conditioned quantum proposal circuit').
  • Deviation scale constants a=1°, b=4°, c=5°, d=20° and initial latent variables t=u=v=0.3, w=0.7 = 1°, 4°, 5°, 20°; 0.3; 0.7
    Broadness of angle deviations around canonical values; hand-tuned to stay in 'intermediate regime' between seed-dominated and delocalized (Supp. Note 10).
  • Outer/target frontier sizes and per-round seeds/shots: N_F=100, N_target=10, N_seed=20, N_shots=100, budget factor 2000 = 100 / 10 / 20 / 100 / 2000 n_b
    Search-loop and budget constants chosen for benchmarking; not fitted to data, but they shape all reported curves.
  • Static degeneracy-breaking perturbation epsilon_stat=1e-9 with disorder scale sigma_ij^(0)=0.3 max(|J_ij|, ΔE/4) = 1e-9; 0.3
    Chosen to lift degeneracies without changing objective; needed for frontier ranking (Supp. Note 10).
  • Classical baseline hyperparameters: P(short)=0.85, short-kick decay 0.45, repair steps ≤4, long-kick range [0.25 n_b, 0. = 0.85; 0.45; 4; 0.25–0.5
    Designed to make a 'strong but generic' classical comparator; the fairness of the comparison depends on these choices.
assumptions (6)
  • standard math Born rule: measurement probabilities are p_s = |⟨s|ψ⟩|² for the final state; ideal unitary evolution of the two-layer circuit.
    All QIPS proposal distributions are defined through Eq. 27; no noise or decoherence is included.
  • domain assumption The function E0(s) + ε_stat ΔE_stat(s) defines an operational ordering that preserves the original optimization up to 1e-9, while lifting degeneracies.
    Used to define unique frontier ranks; if the perturbation changed the ordering of near-degenerate states beyond tolerance, top-K evaluation could shift (Supp. Note 10).
  • domain assumption Proposal-count (2000 n_b evaluations) is the correct shared resource for comparing quantum and classical searches; circuit execution cost and classical simulation cost are not counted.
    The matched-resource claim rests on this; physical gate depth for the diagonal operator is acknowledged to depend on hardware but not included.
  • ad hoc to paper A two-parameter probit CDF (Eq. 11) adequately represents the extended component of proposal distributions for post hoc analysis.
    Used for compact reconstruction; the paper reports good fits but the probit form is not derived from the circuit dynamics.
  • ad hoc to paper The feedback pseudo-energy depending only on n_seed and n_unique can regulate localized patterns without access to the QUBO objective.
    This is the paper's controller design; it is not justified by a general principle and its coefficients are hand-set.
  • domain assumption In accelerated simulation, the sub-threshold extended probability mass can be sampled uniformly from the full computational basis as a conservative approximation.
    Used in production sampling; the paper states it underestimates low-energy contribution, which is conservative for QIPS, but it is an approximation (Supp. Note 10).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Quantum Interference as a Proposal Mechanism for Combinatorial Optimization." pith.science (2026). https://pith.science/paper/U6ZYSWVY

@misc{pith2026260727509,
  author       = {Pith},
  title        = {Pith review of: Quantum Interference as a Proposal Mechanism for Combinatorial Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U6ZYSWVY}},
  note         = {Machine review of arXiv:2607.27509}
}
abstract

Quantum Interference Proposal Search (QIPS) uses seed-conditioned quantum circuits to generate localized interference patterns as finite-shot proposal distributions for QUBO/Ising optimization. Candidate $n_b$-bit strings are sampled from these distributions, scored classically and used to update an elite frontier of low-energy solutions. QIPS uses a fixed two-layer gate-based circuit architecture with 100 shots per circuit while the Hilbert-space dimension grows as $2^{n_b}$. Across six benchmark families with $18 \le n_b \le 29$, QIPS maintains competitive progress relative to a matched classical control that preserves the same search loop, frontier update rule and proposal budget, with total proposals proportional to $n_b$. Performance is assessed using top-$K$ coverage, hit rate, multiplicity, Hilbert-space coverage and dyadic-rank metrics. The results identify localized quantum interference as a resource-efficient proposal mechanism for computational quantum optimization.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 12 canonical work pages

  1. [1]

    Engineering Optimization57(1), 208–233 (2025) https: //doi.org/10.1080/0305215X.2024.2435538

    Chicano, F., Luque, G., Dahi, Z.A., Gil-Merino, R.: Combinatorial optimization with quantum computers. Engineering Optimization57(1), 208–233 (2025) https: //doi.org/10.1080/0305215X.2024.2435538

  2. [2]

    Nature Reviews Physics6(12), 718–735 (2024) https://doi.org/10.1038/s42254-024-00770-9

    Abbas, A., Ambainis, A., Augustino, B., B¨ artschi, A., Buhrman, H., Cof- frin, C., Cortiana, G., Dunjko, V., Egger, D.J., Elmegreen, B.G., Franco, N., Fratini, F., Fuller, B., Gacon, J., Gonciulea, C., Gribling, S., Gupta, S., Had- field, S., Heese, R., Kircher, G., Kleinert, T., Koch, T., Korpas, G., Lenk, S., Marecek, J., Markov, V., Mazzola, G., Mensa...

  3. [3]

    ACM Computing Surveys56(3), 1–36 (2024) https://doi.org/10.1145/3620668

    Gemeinhardt, F., Garmendia, A., Wimmer, M., Weder, B., Leymann, F.: Quan- tum Combinatorial Optimization in the NISQ Era: A Systematic Mapping Study. ACM Computing Surveys56(3), 1–36 (2024) https://doi.org/10.1145/3620668

  4. [4]

    Journal of Physics A: Mathematical and Theoretical56(45), 453001 (2023) https://doi.org/10.1088/ 1751-8121/ad00f0

    Symons, B.C.B., Galvin, D., Sahin, E., Alexandrov, V., Mensa, S.: A practitioner’s guide to quantum algorithms for optimisation problems. Journal of Physics A: Mathematical and Theoretical56(45), 453001 (2023) https://doi.org/10.1088/ 1751-8121/ad00f0

  5. [5]

    Nature Computational Science6, 653–671 (2026) https: //doi.org/10.1038/s43588-026-00991-1

    Koch, T., Bernal Neira, D.E., Chen, Y., Cortiana, G., Egger, D.J., Heese, R., Hegade, N.N., Cadavid, A.G., Huang, R., Itoko, T., Kleinert, T., Xavier, P.M., Mohseni, N., Montanez-Barrera, J.A., Nakano, K., Nannicini, G., O’Meara, C., Pauckert, J., Proissl, M., Ramesh, A., Schicker, M., Shimada, N., Takeori, M., Valls, V., Van Bulck, D., Woerner, S., Zoufa...

  6. [6]

    Quantum Science and Technology10(3), 035052 (2025) https://doi.org/10.1088/ 2058-9565/ade6a5

    Wang, R., Roberto Arezzo, V., Thengil, K., Pecci, G., Santoro, G.E.: From exponential to quadratic: Optimal control for a frustrated Ising ring model. Quantum Science and Technology10(3), 035052 (2025) https://doi.org/10.1088/ 2058-9565/ade6a5

  7. [7]

    Reports on Progress in Physics83(5), 054401 (2020) https://doi.org/10.1088/1361-6633/ab85b8

    Hauke, P., Katzgraber, H.G., Lechner, W., Nishimori, H., Oliver, W.D.: Perspec- tives of quantum annealing: Methods and implementations. Reports on Progress in Physics83(5), 054401 (2020) https://doi.org/10.1088/1361-6633/ab85b8

  8. [8]

    Science 345(6195), 420–424 (2014) https://doi.org/10.1126/science.1252319 21

    Rønnow, T.F., Wang, Z., Job, J., Boixo, S., Isakov, S.V., Wecker, D., Martinis, J.M., Lidar, D.A., Troyer, M.: Defining and detecting quantum speedup. Science 345(6195), 420–424 (2014) https://doi.org/10.1126/science.1252319 21

Show all 58 references
  1. [9]

    Journal of Physics A: Mathematical and Theoretical 56(46), 465304 (2023) https://doi.org/10.1088/1751-8121/ad0439

    Villanueva, A., Najafi, P., Kappen, H.J.: Why adiabatic quantum annealing is unlikely to yield speed-up. Journal of Physics A: Mathematical and Theoretical 56(46), 465304 (2023) https://doi.org/10.1088/1751-8121/ad0439

  2. [10]

    4OR21(3), 363–403 (2023) https://doi.org/10.1007/s10288-023-00549-1

    Grange, C., Poss, M., Bourreau, E.: An introduction to variational quantum algorithms for combinatorial optimization problems. 4OR21(3), 363–403 (2023) https://doi.org/10.1007/s10288-023-00549-1

  3. [11]

    Reviews of Modern Physics94(1), 015004 (2022) https://doi.org/10

    Bharti, K., Cervera-Lierta, A., Kyaw, T.H., Haug, T., Alperin-Lea, S., Anand, A., Degroote, M., Heimonen, H., Kottmann, J.S., Menke, T., Mok, W.-K., Sim, S., Kwek, L.-C., Aspuru-Guzik, A.: Noisy intermediate-scale quantum algo- rithms. Reviews of Modern Physics94(1), 015004 (2...

  4. [12]

    Physics Reports1068, 1–66 (2024) https://doi.org/10.1016/j.physrep.2024.03

    Blekos, K., Brand, D., Ceschini, A., Chou, C.-H., Li, R.-H., Pandya, K., Summer, A.: A review on Quantum Approximate Optimization Algorithm and its variants. Physics Reports1068, 1–66 (2024) https://doi.org/10.1016/j.physrep.2024.03. 002

  5. [13]

    Nature Com- munications12(1), 6961 (2021) https://doi.org/10.1038/s41467-021-27045-6

    Wang, S., Fontana, E., Cerezo, M., Sharma, K., Sone, A., Cincio, L., Coles, P.J.: Noise-induced barren plateaus in variational quantum algorithms. Nature Com- munications12(1), 6961 (2021) https://doi.org/10.1038/s41467-021-27045-6

  6. [14]

    Advanced Quantum Technologies6(9), 2300101 (2023) https://doi.org/10.1002/qute.202300101

    Li, H.-M., Liang, J.-M., Wang, Z.-X., Fei, S.-M.: Ising Hamiltonians for Con- strained Combinatorial Optimization Problems and the Metropolis-Hastings Warm-Starting Algorithm. Advanced Quantum Technologies6(9), 2300101 (2023) https://doi.org/10.1002/qute.202300101

  7. [15]

    IEEE Transactions on Quantum Engineering6, 1–14 (2025) https://doi.org/10.1109/TQE.2025

    Shirai, T., Togawa, N.: Compressed Space Quantum Approximate Optimiza- tion Algorithm for Constrained Combinatorial Optimization. IEEE Transactions on Quantum Engineering6, 1–14 (2025) https://doi.org/10.1109/TQE.2025. 3602404

  8. [16]

    Quantum8, 1313 (2024) https://doi

    Fitzek, D., Jonsson, R.S., Dobrautz, W., Sch¨ afer, C.: Optimizing Variational Quantum Algorithms with qBang: Efficiently Interweaving Metric and Momen- tum to Navigate Flat Energy Landscapes. Quantum8, 1313 (2024) https://doi. org/10.22331/q-2024-04-09-1313

  9. [17]

    Physical Review Research4(3), 033029 (2022) https://doi.org/10.1103/PhysRevResearch

    Zhu, L., Tang, H.L., Barron, G.S., Calderon-Vargas, F.A., Mayhall, N.J., Barnes, E., Economou, S.E.: Adaptive quantum approximate optimization algo- rithm for solving combinatorial problems on a quantum computer. Physical Review Research4(3), 033029 (2022) https://doi.org/10.1...

  10. [18]

    arXiv (2025)

    Gaye, M.A., Shehab, O., Titum, P., Quiroz, G.: Quantum Optimization with Classical Chaos. arXiv (2025). https://doi.org/10.48550/ARXIV.2510.01334 22

  11. [19]

    Quantum Science and Technology7(1), 015021 (2022) https://doi.org/10.1088/ 2058-9565/ac3e54

    Amaro, D., Modica, C., Rosenkranz, M., Fiorentini, M., Benedetti, M., Lubasch, M.: Filtering variational quantum algorithms for combinatorial optimization. Quantum Science and Technology7(1), 015021 (2022) https://doi.org/10.1088/ 2058-9565/ac3e54

  12. [20]

    New Journal of Physics27(5), 054505 (2025) https://doi

    Marin-Sanchez, G., Amaro, D.: Performance analysis of a filtering variational quantum algorithm. New Journal of Physics27(5), 054505 (2025) https://doi. org/10.1088/1367-2630/add365

  13. [21]

    Science Advances9(45), 0487 (2023) https: //doi.org/10.1126/sciadv.adi0487

    Dupont, M., Evert, B., Hodson, M.J., Sundar, B., Jeffrey, S., Yamaguchi, Y., Feng, D., Maciejewski, F.B., Hadfield, S., Alam, M.S., Wang, Z., Grabbe, S., Lott, P.A., Rieffel, E.G., Venturelli, D., Reagor, M.J.: Quantum-enhanced greedy combinatorial optimization solver. Science...

  14. [22]

    IEEE Transactions on Quantum Engineering5, 1–14 (2024) https://doi.org/10.1109/TQE.2024

    Wurtz, J., Sack, S.H., Wang, S.-T.: Solving Nonnative Combinatorial Optimiza- tion Problems Using Hybrid Quantum–Classical Algorithms. IEEE Transactions on Quantum Engineering5, 1–14 (2024) https://doi.org/10.1109/TQE.2024. 3443660

  15. [23]

    Physica Scripta99(5), 055104 (2024) https://doi.org/10.1088/1402-4896/ad2e55

    Tang, L., Wang, H., Li, Z., Tang, H., Zhang, C., Li, S.: Quantum dueling: An efficient solution for combinatorial optimization. Physica Scripta99(5), 055104 (2024) https://doi.org/10.1088/1402-4896/ad2e55

  16. [24]

    Quantum 9, 1906 (2025) https://doi.org/10.22331/q-2025-11-06-1906 arXiv:2404.01412 [quant-ph]

    Maciejewski, F.B., Biamonte, J., Hadfield, S., Venturelli, D.: Improving Quantum Approximate Optimization by Noise-Directed Adaptive Remapping. Quantum 9, 1906 (2025) https://doi.org/10.22331/q-2025-11-06-1906 arXiv:2404.01412 [quant-ph]

  17. [25]

    Advanced Quantum Technologies9(2), 2400484 (2026) https: //doi.org/10.1002/qute.202400484

    Wu, S.-Y., Song, Y.-Q., Li, R.-Z., Qin, S.-J., Wen, Q.-Y., Gao, F.: Resource- Efficient Adaptive Variational Quantum Algorithm for Combinatorial Optimiza- tion Problems. Advanced Quantum Technologies9(2), 2400484 (2026) https: //doi.org/10.1002/qute.202400484

  18. [26]

    Nature Computational Science5, 1168–1177 (2025) https://doi.org/10.1038/ s43588-025-00873-y

    Kotil, A., Pelofske, E., Riedm¨ uller, S., Egger, D.J., Eidenbenz, S., Koch, T., Woerner, S.: Quantum approximate multi-objective optimization. Nature Computational Science5, 1168–1177 (2025) https://doi.org/10.1038/ s43588-025-00873-y

  19. [27]

    Nzongani, U., Mermoud, D.L., Molfetta, G.D., Simonetto, A.: Sampled-based guided quantum walk: Non-variational quantum algorithm for combinatorial optimization (2025) arXiv:2509.15138 [quant-ph]

  20. [28]

    In: 2024 IEEE International Conference on Quantum Comput- ing and Engineering (QCE), pp

    Bennett, T., Noakes, L., Wang, J.: Non-Variational Quantum Combinatorial Optimisation. In: 2024 IEEE International Conference on Quantum Comput- ing and Engineering (QCE), pp. 31–41. IEEE, Montreal, QC, Canada (2024). https://doi.org/10.1109/QCE60285.2024.00014 23

  21. [29]

    Physical Review A113(3), 032603 (2026) https: //doi.org/10.1103/sdlc-wl67

    Bennett, T., Smith, A., Matwiejew, E., Wang, J.B.: Benchmarking quantum heuristics: Nonvariational quantum-walk-based optimization algorithm for the weighted MaxCut problem. Physical Review A113(3), 032603 (2026) https: //doi.org/10.1103/sdlc-wl67

  22. [30]

    Physical Review Letters125(26), 260505 (2020) https://doi.org/10.1103/PhysRevLett.125.260505

    Bravyi, S., Kliesch, A., Koenig, R., Tang, E.: Obstacles to Variational Quan- tum Optimization from Symmetry Protection. Physical Review Letters125(26), 260505 (2020) https://doi.org/10.1103/PhysRevLett.125.260505

  23. [31]

    npj Quantum Information9(1), 73 (2023) https://doi

    Lykov, D., Wurtz, J., Poole, C., Saffman, M., Noel, T., Alexeev, Y.: Sampling frequency thresholds for the quantum advantage of the quantum approximate optimization algorithm. npj Quantum Information9(1), 73 (2023) https://doi. org/10.1038/s41534-023-00718-4

  24. [32]

    end-game

    B¨ arligea, A., Poggel, B., Lorenz, J.M.: Scalability challenges in variational quan- tum optimization under stochastic noise. Physical Review A112(3), 032407 (2025) https://doi.org/10.1103/rgyh-8xw8 24 Supplementary Information Quantum Interference as a Proposal Mechanism for...

  25. [33]

    The actual circuit angles are obtained by adding stochastically generated deviations to these canonical values. Two-layer quantum proposal circuit For qubitjin layerℓ, define the unit vector nj ℓ = sinθ j ℓ cosϕ j ℓ,sinθ j ℓ sinϕ j ℓ,cosθ j ℓ ,(21) and the corresponding single...

  26. [34]

    quantum proposals only

  27. [35]

    classical kick-and-repair proposals only

  28. [36]

    blind uniform proposals

  29. [37]

    quantum proposals followed by classical proposals

  30. [38]

    classical proposals followed by quantum proposals

  31. [39]

    The principal quantum ablations include:

    interleaved classical and quantum proposals. The principal quantum ablations include:

  32. [40]

    disabling the cost phase separator

  33. [41]

    replacing the XYZ mixers by XZ mixers

  34. [42]

    disabling the longitudinal random fields

  35. [43]

    disabling dynamic cost-operator jitter

  36. [44]

    shuffling the correspondence between computational-basis states and energy values

  37. [45]

    holding the seed fixed

  38. [46]

    holding the angle deviations fixed

  39. [47]

    The energy-label shuffling control destroys the quadratic correspondence between nearby computational- basis states and the cost spectrum while preserving the same set of energies

    disabling accelerated sparse probability sampling. The energy-label shuffling control destroys the quadratic correspondence between nearby computational- basis states and the cost spectrum while preserving the same set of energies. Under this control, the structured QIPS propo...

  40. [48]

    Construct the QUBO or Ising energy spectrum and add a negligible static perturbation to define an ordering through exact degeneracies

  41. [49]

    Initialize the elite frontier from randomly sampled states

  42. [50]

    Select 20 seed states from the frontier using the exponential rank-biased distribution in Eq. (80)

  43. [51]

    For each seed, construct its canonical two-layer circuit and sample a trial set of latent deviation parameters around the most recently accepted values

  44. [52]

    Optionally perturb the cost phase by the adaptive jitter spectrum

  45. [53]

    Evaluate the two-layer circuit and generate 100 measured bitstrings

  46. [54]

    Merge the measured states into the elite list and truncate the frontier to its 100 lowest operational-energy states

  47. [55]

    Calculaten seed,n unique, and the pseudo-energy in Eq. (58)

  48. [56]

    Accept or reject the trial latent parameters using the Metropolis rule in Eq. (73)

  49. [57]

    Update the qualification-rate controller, jitter success rates, and frontier-contraction statistics

  50. [58]

    The algorithm therefore separates three roles

    Repeat for all seeds and rounds until the fixed proposal budget is exhausted. The algorithm therefore separates three roles. The canonical angles encode the selected seed, the randomized deviations generate a diverse ensemble of localized interference patterns, and the classic...

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.