Spatial Regionalization: A Hybrid Quantum Computing Approach

Amr Magdy; Federico M. Spedalieri; Ibrahim Sabek; Yunhan Chang

REVIEW 3 major objections 5 minor 22 references

Spatial Regionalization: A Hybrid Quantum Computing Approach

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper presents the first hybrid quantum-classical approach to spatial regionalization, reporting up to 62.3% runtime speedup in local optimization and up to 46.5% better seed selection quality on initial D-Wave tests.

desk verdict Promising hybrid quantum pipeline for spatial regionalization, but the local-optimization quality results rest on an unvalidated ΔH proxy that needs to be either derived properly or replaced with exact computation. read the letter →

arxiv 2506.18799 v1 pith:M6AD5KPY submitted 2025-06-23 cs.ET cs.ITmath.ITquant-ph

Yunhan Chang , Amr Magdy , Federico M. Spedalieri , Ibrahim Sabek This is my paper · ORCID

classification cs.ETcs.ITmath.ITquant-ph

keywords spatialregionalizationquantumannealinghybridquantum-classicaloptimizationQUBOmaximumindependentsetmax-mindispersionlocalD-Wave

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 argues that spatial regionalization, the NP-hard problem of grouping polygons into contiguous, internally homogeneous regions, can be tackled by a hybrid quantum-classical pipeline even though quantum hardware cannot handle the whole problem at once. The authors decompose seeding-based regionalization into stages and offload the two most promising stages to a D-Wave quantum annealer: seed selection, encoded as a maximum independent set problem in a BQM, and local optimization, encoded as a CQM that chooses border-area moves. Against the classical PRUC baseline, the initial experiments report up to 46.5% better seed dispersion quality and up to 62.3% faster local optimization at 5k polygons, while noting a slowdown at tiny sizes. The central claim is a potential quantum advantage for realistic regionalization workloads, not a demonstrated end-to-end win.

What carries the argument

The load-bearing mechanism is the reformulation of regionalization subproblems as QUBO/CQM models that fit current annealer constraints. Seed selection uses the Maximum Independent Set formulation: for a candidate distance threshold $d_m$, a graph keeps only edges with distance $\le d_m$, and a penalized objective $X=-\sum_i x_i+\lambda_{\mathrm{MIS}}\sum_{(i,j)\in E'}x_ix_j$ is minimized; a classical binary search on $d_m$ finds the largest dispersion supporting $p$ seeds. Local optimization uses a swap-or-move CQM whose objective estimates the heterogeneity change of moving area $a_m$ from donor to receiver via region statistics $(N_R,M_R,V_R)$ under the normality assumption, rather than recomputing pairwise sums, enabling the QPU to evaluate many moves at once while preserving contiguity.

What would settle it

Recompute the exact heterogeneity change for every candidate move on the paper's datasets and compare it with the estimator $\Delta\hat{H}$; if the estimator misranks a substantial share of moves, or if replacing the quantum CQM loop with a classical loop that uses the exact $\Delta H$ achieves the same or better final heterogeneity at comparable runtime, the reported quantum advantage is an artifact of the estimator rather than of annealing.

Watch

Extended reading notes

Core claim

The paper claims that the two stages common to all seeding-based regionalization algorithms, seed selection and local optimization, can each be expressed as constrained quadratic models amenable to quantum annealing, and that on a D-Wave Advantage system these quantum-encoded stages improve on the classical PRUC heuristic. Seed selection becomes a max-min dispersion problem, solved by quantum-assisted binary search over the minimum pairwise centroid distance, with each threshold check formulated as a maximum independent set QUBO. Local optimization is an iterative hybrid loop: the classical CPU identifies movable border areas via articulation points, and the QPU selects a compatible set of moves that minimizes an estimated change in regional heterogeneity, $\Delta\hat{H}=N_R\sqrt{V_R+(M_R-A_m)^2}-(N_R-1)\sqrt{V_R+(M_R-A_m)^2}$, under constraints that each area moves at most once and each region either donates or receives at most one area per iteration.

Load-bearing premise

The load-bearing premise is that the shortcut for estimating how much a border move changes a region's heterogeneity, which assumes the attribute values are normally distributed, ranks candidate moves in the same order as computing the true change in all pairwise absolute differences would.

Editorial extensions

If this is right

At dataset sizes used in practice (5k polygons, comparable to all U.S. counties at 3.2k), the quantum local optimization loop runs up to 62.3% faster than PRUC's classical counterpart while maintaining or improving quality.
Quantum seed selection finds more widely scattered seeds as the number of regions $p$ grows, improving minimum seed distance by up to 46.5%, a regime where classical heuristics struggle.
Because seed selection and local optimization are shared by all seeding-based regionalization techniques, any such technique can adopt these quantum stages without altering its region-growing, enclave assignment, or constraint-adjustment logic.
The quantum advantage is scale-dependent: at 50 polygons local optimization is slower than classical, and quality gains shrink as data size grows, so the framework's benefit is positioned for larger realistic workloads.
The paper's decomposition gives a template for other spatial optimization problems too large for direct QUBO encoding, by splitting them into quantum-friendly subproblems plus classical orchestration.

Reading between the lines

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

The paper compares against PRUC only; a cleaner attribution test would compare the same five-stage pipeline with and without the QPU, separating annealing's contribution from the hybrid solver's classical heuristics and the new CQM encoding.
The normality-based $\Delta\hat{H}$ estimator is the step most likely to dominate behavior; if it is replaced by exact heterogeneity updates, the quantum loop's advantage may shrink, so validating it on skewed county-level attributes is a direct next experiment.
The max-min dispersion / maximum independent set encoding ties seed selection to other dispersion problems, such as codebook design and facility placement, so the reported quality gains could transfer to those settings with little modification.
A prediction of the framework is that the seed-selection advantage will widen as $p$ grows beyond 20 and as datasets reach county scale, which could be tested immediately on public U.S. county data.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague, here is my take on arXiv:2506.18799. The genuinely new thing is the two-stage hybrid quantum-classical decomposition for seeding-based spatial regionalization: seed selection via a MIS/BQM formulation, and local optimization via CQM. I don't know of earlier work applying hybrid quantum methods to this problem, and the decomposition itself is sensible—offload the two hardest stages to the QPU, keep region growing and adjustment classical. The authors use PRUC as a baseline, which is external and reasonable, and they honestly report slowdowns at small dataset sizes while showing a runtime speedup at 5k polygons and a 46.5% seed-quality improvement. Those results, if they hold, matter to people in spatial computing and quantum optimization. The soft spot is real and load-bearing for the local-optimization quality claim. Section 4.2 defines ΔH as N_R·sqrt(V_R + (M_R − A_m)^2) minus (N_R−1) times the same term, justified only by a normality assumption. The stress-test note is correct: the expected sum of absolute deviations under normality is not proportional to that RMS expression, so the proxy can misrank candidate moves. The paper gives no derivation, no comparison against exact ΔH, and no error bars or released code. Since the reported quality improvements are computed from the true heterogeneity H, the 6.3–22.4% numbers may reflect the proxy's bias rather than quantum optimization. That is not a minor caveat; it undercuts one of the two central claims. The runtime speedup does not depend on the proxy, and the seed selection results are separate, so those parts survive. The paper reads like a workshop paper with an honest "initial results" framing, and the authors do point to an appendix for details. But the missing validation of the estimator is exactly the sort of thing a serious referee should catch and require before the quality claim is accepted. If I were the editor, I would send this to review, not desk-reject, because the idea is novel and the pipeline is plausible. The review should ask for a correct derivation or a classical exact-ΔH baseline, plus run-to-run statistics and artifacts. I'd bring it to a reading group to discuss the proxy pitfall, and I'd cite the seed selection part in my own work on hybrid quantum optimization. Recommendation: accept for peer review, but with major revisions expected. The core contribution is worth refereeing, but the current local-optimization quality evidence is not solid.

Referee Report

3 major / 5 minor

Summary. The paper proposes a hybrid quantum-classical method for spatial regionalization, decomposing the problem into five stages and replacing the classical seed-selection and local-optimization stages with quantum-encoded BQM and CQM formulations. Seed selection is mapped to a Maximum Independent Set problem and solved on D-Wave's BQM solver; local optimization is formulated as a CQM that chooses moves of border areas between regions, using a statistical estimator of the change in regional heterogeneity. Experiments compare the quantum pipeline against the classical PRUC baseline on datasets up to 5k polygons, reporting up to 46.5% seed-quality improvement and up to 62.3% runtime speedup in local optimization, while honestly noting a slowdown at small dataset sizes. The paper concludes with a claim of potential quantum advantage in seed selection quality and local-optimization runtime efficiency.

Significance. If validated, this would be a useful first step toward applying quantum annealers to spatial regionalization, a problem not previously addressed with quantum methods. The decomposition into quantum-friendly subproblems is sensible, the MIS-based seed selection is a clean mapping, and the reported runtime scaling of the QPU-based local optimization is encouraging. The paper also honestly reports the slowdown at small dataset sizes. However, the current evidence is preliminary: the local-optimization objective relies on an unvalidated statistical proxy, the experimental evaluation lacks error bars and ablation controls, and neither code nor data is released. The seed-selection quality result is more credible because it is independent of the proxy, but it also needs repeated-run statistics to be fully convincing.

major comments (3)

[§4.2 and Appendix B] The ΔH estimator is load-bearing for the local-optimization quality claims and is not valid as stated. The text defines ΔĤ_add = N_R·√(V_R+(M_R−A_m)^2), ΔĤ_remove = −(N_R−1)·√(V_R+(M_R−A_m)^2), and then ΔH = ΔĤ_add+ΔĤ_remove, but all three lines use the same symbol R for the region, so as written the two terms cancel to a single RMS term. If the intention is to use receiver statistics for the add term and donor statistics for the remove term, that must be written explicitly with distinct subscripts such as R_r and R_d. More importantly, no derivation is given from H = Σ_k Σ_{i<j∈R_k}|A_i−A_j|, and the RMS expression is not the correct estimate even under the stated normality assumption: for X∼N(μ,σ²), E|X−a| = σ√(2/π)exp(−(a−μ)²/(2σ²)) + |a−μ|erf(|a−μ|/(σ√2)), not √(σ²+(a−μ)²). Since the CQM objective in Appendix B minimizes this proxy, the QPU may select moves that increase true H. Figure 1a reports quality improvements computed from H, so the improvement cannot be attributed to optimizing the true objective. I ask the authors to either derive a correct estimator, validate the proxy by comparing its ranking of candidate moves with exact ΔH on random instances, or compute exact ΔH in the CQM for small n and report the discrepancy.
[§5] The experimental evaluation does not yet support the strength of the claims. Figures 1 and 2 report single curves with no number of independent runs, standard deviations, or confidence intervals, and the datasets are not described beyond polygon counts. For the local-optimization quality claim, an ablation is needed: compare the quantum CQM against (a) PRUC's classical local search, (b) a classical local search using the same ΔH proxy, and (c) a random move selection subject to the same constraints. Without (b) and (c), the positive quality numbers in Fig. 1a could reflect the proxy's bias or the classical pre-processing rather than a quantum advantage. Please also report the value of λ_MIS and a sensitivity analysis for it, since the seed-selection results may depend on this penalty coefficient.
[§4.2] The contiguity constraint 'each region may either donate or receive at most one area per iteration, but not both' is asserted to preserve spatial contiguity, but no proof or empirical verification is given. The constraint also removes the possibility of a region both donating and receiving, which may be needed for some beneficial swaps. Because the CQM's feasible set is defined by this constraint, the quality results depend on it. Please provide a proof that the final regions remain contiguous under all allowed move combinations, or at least report the fraction of iterations in which contiguity was explicitly verified.

minor comments (5)

[Appendix B] The heading reads 'CQM in Location Optimization' but the stage is called 'local optimization' throughout the rest of the paper; please standardize the terminology.
[§5] Please specify whether the reported runtimes include QPU access or API overhead or only solver time, and describe the dataset generation process (attribute distributions, polygon sizes, number of instances) so the results can be reproduced.
[References] The paper contains a printed appendix (A and B) but also cites an external Appendix URL [1]; please clarify the relationship between the two and ensure the URL is stable.
[Figures 1 and 2] Both figures use two y-axes with very different scales; please add explicit axis labels and a legend for the quality and runtime curves so the reader can distinguish them without relying on the caption.
[§3] The claim that 'none of the existing techniques leverages quantum annealing' is too strong given the quantum clustering work cited in [19,18,5]; please qualify it to 'no existing technique for spatial regionalization with explicit contiguity constraints'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity found; comparisons are against an external classical baseline and evaluation uses the true objective functions.

full rationale

The derivation chain is self-contained in the sense required here. The quantum local optimization minimizes an estimated ΔH (Section 4.2), but the reported quality improvement is measured on the true regional heterogeneity H, and the estimator contains no fitted parameters and is not asserted to equal H by construction. Even if the RMS proxy misranks moves, that is an approximation/validation risk, not a circular reduction. The seed-selection BQM optimizes max-min dispersion via a standard MIS reduction (citing [20]) and is evaluated by the same min pairwise distance; measuring the objective directly is not circular. The classical counterpart is PRUC (Liu et al., VLDB 2021), an external published baseline sharing a co-author but independently testable; results are compared on absolute metrics. Self-citations ([4], [6]) provide the stage decomposition and future-work contiguity modeling; they are building blocks, not uniqueness results that force the conclusions. The only flagged omission is that complete BQM/CQM formulations are deferred to an appendix [1], a completeness limitation rather than a circular step. Accordingly, no equation in the paper reduces to its own inputs, and no fitted parameter is renamed as a prediction.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. The main loaded assumptions are the unvalidated ΔH estimator, the contiguity-preserving restriction, the reliability of D-Wave hybrid solvers, and the suitability of PRUC as a baseline. There is one unreported penalty parameter in the BQM model.

free parameters (1)

lambda_MIS
Penalty coefficient in the BQM seed selection objective (Section 4.1 and Appendix A), described only as 'sufficiently large'. Not reported; it affects whether the MIS constraint is actually satisfied by the annealer's samples.

assumptions (4)

ad hoc to paper Attributes of large datasets are normally distributed, and the change in heterogeneity from moving an area can be estimated from N_R, M_R, V_R, and A_m using the given ΔH formulas.
Stated in Section 4.2 with no validation on the evaluation datasets. The formula uses RMS deviation, not expected absolute deviation, so the estimator is not a faithful consequence of the normality assumption.
ad hoc to paper Each region may donate or receive at most one area per iteration, but not both, and this constraint preserves spatial contiguity while still allowing beneficial moves.
Imposed in Section 4.2 to prevent cascading changes; no proof that it is necessary or sufficient for contiguity, and it may exclude beneficial combinations of moves.
domain assumption D-Wave hybrid solvers (BQM and CQM) return sufficiently good solutions for the submitted models.
The empirical claims depend on the solvers finding useful samples on the specific instances; this is measured, not guaranteed, and the paper provides no statistics over repeated calls.
domain assumption PRUC is an appropriate classical baseline for both seed selection and local optimization stages.
Section 5 compares against PRUC with default settings, assuming it targets the same objectives (max-min dispersion and heterogeneity) with comparable effort. No ablation of the baseline's parameters is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Spatial Regionalization: A Hybrid Quantum Computing Approach." pith.science (2026). https://pith.science/paper/M6AD5KPY

@misc{pith2026250618799,
  author       = {Pith},
  title        = {Pith review of: Spatial Regionalization: A Hybrid Quantum Computing Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M6AD5KPY}},
  note         = {Machine review of arXiv:2506.18799}
}

read the original abstract

Quantum computing has shown significant potential to address complex optimization problems; however, its application remains confined to specific problems at limited scales. Spatial regionalization remains largely unexplored in quantum computing due to its complexity and large number of variables. In this paper, we introduce the first hybrid quantum-classical method to spatial regionalization by decomposing the problem into manageable subproblems, leveraging the strengths of both classical and quantum computation. This study establishes a foundational framework for effectively integrating quantum computing methods into realistic and complex spatial optimization tasks. Our initial results show a promising quantum performance advantage for a broad range of spatial regionalization problems and their variants.

Figures

Figures reproduced from arXiv: 2506.18799 by the authors.

**Figure 2.** Performance in seed selection vs. p Based on the separator theorem [16], the number of potential movable border areas ranges from 2 √ 𝑛𝑝 to 2𝑝 √ 𝑛 in every single iteration. For space limitations, our CQM formulation is detailed in [1]. 5 Empirical Results The proposed BQM and CQM models for seed selection and location optimization, respectively, are implemented using Python 3.12 and constructed using the D-Wave O… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

[1]

http://cs.ucr.edu/~amr/AppendixQR.pdf, 2025

Appendix. http://cs.ucr.edu/~amr/AppendixQR.pdf, 2025

work page 2025
[2]

Alrashid, Y

H. Alrashid, Y. Liu, and A. Magdy. SMP: Scalable Max-P Regionalization. In SIGSPATIAL, 2022

work page 2022
[3]

Alrashid, Y

H. Alrashid, Y. Liu, and A. Magdy. PAGE: parallel scalable regionalization frame- work.ACM Transactions on Spatial Algorithms and Systems, 9(3):1–26, 2023

work page 2023
[4]

Alrashid and A

H. Alrashid and A. Magdy. A Scalable Unified System for Seeding Regionalization Queries. InProceedings of the 18th International Symposium on Spatial and Temporal Data, SSTD, pages 96–105, 2023

work page 2023
[5]

Arthur and P

D. Arthur and P. Date. Balancedk-means clustering on an adiabatic quantum computer.Quantum Information Processing, 20(9):294, 2021

work page 2021
[6]

Quantum Modeling of Spatial Contiguity Constraints

Y. Chang, A. Magdy, and F. M. Spedalieri. Quantum Modeling of Spatial Contigu- ity Constraints, 2025. https://arxiv.org/abs/2505.12608. To appear at the ACM SIGMOD Workshop on Quantum Computing and Quantum-Inspired Technology for Data-Intensive Systems and Applications (Q-Data 2025)

work page Pith review arXiv 2025
[7]

Real-World Quantum Applications at Business Scale

D-Wave. Real-World Quantum Applications at Business Scale. https://www. dwavequantum.com/learn/customer-success-stories/, 2025

work page 2025
[8]

D -Wave Systems Inc., 2025

D-Wave Systems Inc.Leap Service’s Hybrid Solvers. D -Wave Systems Inc., 2025. Accessed 2025-06-12

work page 2025

Show all 22 references

[9]

Ebadi, T

S. Ebadi, T. T. Wang, H. Levine, A. Keesling, G. Semeghini, K. Pichler, H. Bernien, et al. Quantum optimization of maximum independent set using rydberg atom arrays.Science, 376(6598):1209–1215, 2022

2022
[10]

Spatial Clustering — pygeoda 1.0.0 documentation

GeoDa Center for Geospatial Analysis and Computation. Spatial Clustering — pygeoda 1.0.0 documentation. https://geodacenter.github.io/pygeoda/spatial_ clustering.html, 2025. Accessed: 2025-06-12

2025
[11]

Glover, G

F. Glover, G. Kochenberger, R. Hennig, and Y. Du. Quantum bridge analytics i: a tutorial on formulating and using qubo models.Annals of Operations Research, 314(1):141–183, 2022

2022
[12]

Kadowaki and H

T. Kadowaki and H. Nishimori. Quantum annealing in the transverse ising model. Phys. Rev. E, 58:5355–5363, Nov 1998

1998
[13]

Kang and A

Y. Kang and A. Magdy. EMP: Max-P Regionalization with Enriched Constraints. In2022 IEEE 38th International Conference on Data Engineering (ICDE), 2022

2022
[14]

A. D. King et al. Quantum critical dynamics in a 5,000-qubit programmable spin glass.Nature, 617(7959):61–66, 2023. Spatial Regionalization: A Hybrid Quantum Computing Approach SIGSPATIAL ’25, 978-1-4503-XXXX-X/2025/06

2023
[15]

Koshka and M

Y. Koshka and M. A. Novotny. Comparison of D-Wave Quantum Annealing and Classical Simulated Annealing for Local Minima Determination. InProceedings of the 2020 IEEE International Conference on Rebooting Computing (ICRC), 2020

2020
[16]

R. J. Lipton and R. E. Tarjan. A Separator Theorem for Planar Graphs.SIAM Journal on Applied Mathematics, 36(2):177–189, 1979

1979
[17]

Y. Liu, A. R. Mahmood, A. Magdy, and S. Rey. PRUC: P-regions with User-Defined Constraint.Proc. VLDB Endow., 15(3), 2021

2021
[18]

Negre, H

C. Negre, H. Ushijima-Mwesigwa, and S. Mniszewski. Detecting multiple com- munities using quantum annealing on the d-wave system.PLOS ONE, 15, 2020

2020
[19]

Ushijima-Mwesigwa, C

H. Ushijima-Mwesigwa, C. F. A. Negre, and S. M. Mniszewski. Graph partitioning using quantum annealing on the d-wave system. InProceedings of the Second International Workshop on Post Moores Era Supercomputing, PMES’17, 2017

2017
[20]

Yukiyoshi, T

K. Yukiyoshi, T. Mikuriya, H. S. Rou, G. T. F. de Abreu, and N. Ishikawa. Quantum speedup of the dispersion and codebook design problems.IEEE Transactions on Quantum Engineering, 2024. Appendix A BQM in Seed Selection The seed selection stage is modeled as a max-min dispersion...

2024
[21]

This is enforced for every area𝑎𝑖 in the set of movable areas

One Move per Area.Each movable area 𝑎𝑖 can be transferred at most once. This is enforced for every area𝑎𝑖 in the set of movable areas. Given that for a specific area𝑎𝑖, its donor region𝑅𝑑 is fixed, we sum over all possible receiver regions𝑅𝑟 : ∀𝑎𝑖∈MovableAreas, ∑︁ 𝑟|(𝑖,𝑑,𝑟)∈𝑆 𝑥𝑖,𝑑,𝑟≤1
[22]

Region Stability.Each region 𝑅𝑘 can either act as a donor for one move or as a receiver for one move, but not both. This is enforced for every region𝑘from1to𝑝: ∀𝑘∈{1,...,𝑝}, ∑︁ (𝑖,𝑘,𝑟)∈𝑆 𝑥𝑖,𝑘,𝑟+ ∑︁ (𝑖,𝑑,𝑘)∈𝑆 𝑥𝑖,𝑑,𝑘≤1 This constraint ensures that all selected moves are mutually...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

http://cs.ucr.edu/~amr/AppendixQR.pdf, 2025

Appendix. http://cs.ucr.edu/~amr/AppendixQR.pdf, 2025

work page 2025

[2] [2]

Alrashid, Y

H. Alrashid, Y. Liu, and A. Magdy. SMP: Scalable Max-P Regionalization. In SIGSPATIAL, 2022

work page 2022

[3] [3]

Alrashid, Y

H. Alrashid, Y. Liu, and A. Magdy. PAGE: parallel scalable regionalization frame- work.ACM Transactions on Spatial Algorithms and Systems, 9(3):1–26, 2023

work page 2023

[4] [4]

Alrashid and A

H. Alrashid and A. Magdy. A Scalable Unified System for Seeding Regionalization Queries. InProceedings of the 18th International Symposium on Spatial and Temporal Data, SSTD, pages 96–105, 2023

work page 2023

[5] [5]

Arthur and P

D. Arthur and P. Date. Balancedk-means clustering on an adiabatic quantum computer.Quantum Information Processing, 20(9):294, 2021

work page 2021

[6] [6]

Quantum Modeling of Spatial Contiguity Constraints

Y. Chang, A. Magdy, and F. M. Spedalieri. Quantum Modeling of Spatial Contigu- ity Constraints, 2025. https://arxiv.org/abs/2505.12608. To appear at the ACM SIGMOD Workshop on Quantum Computing and Quantum-Inspired Technology for Data-Intensive Systems and Applications (Q-Data 2025)

work page Pith review arXiv 2025

[7] [7]

Real-World Quantum Applications at Business Scale

D-Wave. Real-World Quantum Applications at Business Scale. https://www. dwavequantum.com/learn/customer-success-stories/, 2025

work page 2025

[8] [8]

D -Wave Systems Inc., 2025

D-Wave Systems Inc.Leap Service’s Hybrid Solvers. D -Wave Systems Inc., 2025. Accessed 2025-06-12

work page 2025