pith. machine review for the scientific record. sign in

arxiv: 2604.27324 · v1 · submitted 2026-04-30 · 🪐 quant-ph

Recognition: unknown

Q3SAT-GPT: A Generative Model for Discovering Quantum Circuits for the 3-SAT Problem

Pratim Ugale , Ilya Tyagin , Karunya Shirali , Kien X. Nguyen , Ilya Safro
Authors on Pith no claims yet

Pith reviewed 2026-05-07 08:58 UTC · model grok-4.3

classification 🪐 quant-ph
keywords quantum circuitsgenerative modelsMax-E3-SATQAOAcircuit discoveryadaptive constructionshallow ansatze
0
0 comments X

The pith

A generative model trained on adaptive QAOA circuits produces effective shallow circuits for new Max-E3-SAT instances without optimization.

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

The paper introduces Q3SAT-GPT, a generative model that learns to create quantum circuits for the Max-E3-SAT problem directly from examples of high-performing circuits. These examples are produced by MosaicADAPT-QAOA, an adaptive construction method that selects subsets of mixer operators at each step to keep circuit depth low. Once trained, the model generates candidate circuits for unseen problem instances that achieve strong solution quality at shallow depths. Experiments demonstrate that this generation process avoids the repeated variational optimization steps of conventional methods and scales more efficiently as problem size grows. The approach treats circuit discovery as a learned pattern rather than an instance-by-instance search.

Core claim

Q3SAT-GPT is a generative model that internalizes effective circuit design patterns from a training set of low-depth QAOA-style ansatze built by MosaicADAPT-QAOA. MosaicADAPT-QAOA constructs these training circuits by choosing subsets of mixer operators rather than inserting them sequentially, yielding high-quality supervision data. The trained model then produces circuits for new Max-E3-SAT instances that deliver competitive solution quality using only shallow depths and without any variational optimization at inference time. This results in better scaling than both the adaptive construction procedure and standard variational baselines.

What carries the argument

Q3SAT-GPT, the generative model that outputs complete circuit structures by learning design patterns from a dataset of high-performing, low-depth QAOA ansatze.

If this is right

  • Circuit generation occurs in a single forward pass, removing the need for iterative parameter tuning on each new problem.
  • Shallow generated circuits maintain solution quality while lowering the qubit and gate resources demanded from hardware.
  • Scaling behavior improves because the model avoids the growing optimization cost that affects both adaptive construction and variational methods.
  • The learned patterns supply a reusable foundation for adding instance-specific features in later generative models.

Where Pith is reading between the lines

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

  • The same training-and-generation pipeline could be applied to other combinatorial problems such as graph partitioning or scheduling.
  • Embedding problem-instance features directly into the generative model might produce circuits even better matched to individual inputs.
  • Scaling the training data to include many optimization families could yield a general-purpose circuit generator for broader quantum optimization tasks.

Load-bearing premise

The generative model captures general circuit construction logic from the MosaicADAPT-QAOA training examples that transfers usefully to entirely new 3-SAT instances.

What would settle it

Generate circuits with Q3SAT-GPT for a fresh collection of larger or structurally distinct Max-E3-SAT instances and observe that their solution quality falls substantially below circuits obtained from full variational optimization of QAOA.

Figures

Figures reproduced from arXiv: 2604.27324 by Ilya Safro, Ilya Tyagin, Karunya Shirali, Kien X. Nguyen, Pratim Ugale.

Figure 1
Figure 1. Figure 1: A Representative Conflict Illustration. Operator gener view at source ↗
Figure 2
Figure 2. Figure 2: Incompatibility graph representation for a 3-qubit view at source ↗
Figure 3
Figure 3. Figure 3: Detailed pipeline demonstrating formatting, tokenization, and embeddings. The view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of the energy convergence across adap view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of the maximum gradient norm across all view at source ↗
Figure 6
Figure 6. Figure 6: Evolution of the operator selection across adaptations. Note that the count inside a bar is that of the total number of view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of Q3SAT-GPT predicted parameters view at source ↗
read the original abstract

This work introduces Q3SAT-GPT, a generative model for discovering quantum circuits for the Max-E3-SAT problem. Our method learns from high-performing QAOA-style ans\"atze to directly generate candidate circuits. To create high-quality supervision, we also introduce Mosaic Adaptive QAOA (MosaicADAPT-QAOA), an adaptive strategy for constructing low-depth QAOA circuits by selecting subsets of mixer operators in each step, rather than inserting operators sequentially. The resulting circuits serve as training data for the generative model, allowing it to learn effective circuit design patterns while eliminating the need for costly variational optimization at inference time. Experiments show that our framework attains strong solution quality with shallow circuits and scales significantly better than both our adaptive construction procedure and conventional variational baselines. Our results establish generative modeling as a high-performance route toward the scalable discovery of quantum optimization circuits, demonstrating that these models can effectively internalize circuit logic while providing a foundation for future, instance-aware inductive biases. Reproducibility: The source code is available at https://github.com/pratimugale/Q3SAT-GPT.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces Q3SAT-GPT, a generative model that learns to produce quantum circuits for Max-E3-SAT directly from training data consisting of high-performing QAOA-style ansatze. To generate this supervision, the authors propose MosaicADAPT-QAOA, an adaptive procedure that constructs low-depth circuits by selecting subsets of mixer operators at each step. The generative model is then trained to internalize effective design patterns, enabling circuit generation for unseen instances at inference time without any variational optimization. The central empirical claim is that the resulting circuits achieve strong solution quality with shallow depths and exhibit significantly better scaling than both MosaicADAPT-QAOA and standard variational QAOA baselines.

Significance. If the reported scaling and quality advantages are robust, the work would demonstrate that generative models can internalize circuit-construction heuristics from adaptive QAOA data and apply them zero-shot, offering a route to reduce the per-instance optimization cost that currently limits variational quantum algorithms for combinatorial problems. The public code release is a positive factor for verifiability.

major comments (2)
  1. The training corpus is generated entirely by the authors' own MosaicADAPT-QAOA procedure. This creates a potential circularity: any reported advantage of Q3SAT-GPT over MosaicADAPT-QAOA could simply reflect faster inference-time approximation of the same adaptive selection rule rather than discovery of independent circuit logic. The experimental section should therefore include direct comparisons of approximation ratios and depths achieved by Q3SAT-GPT versus the original MosaicADAPT-QAOA outputs on the same held-out instances, together with an analysis of how often the generated circuits differ structurally from the training examples.
  2. The abstract states that the framework 'attains strong solution quality' and 'scales significantly better' than both the adaptive construction procedure and conventional variational baselines, yet no numerical values, problem sizes, number of instances, or statistical tests are supplied. The results section must contain explicit tables or figures reporting approximation ratios, circuit depths, wall-clock scaling, and baseline details (including standard QAOA with fixed p) so that the magnitude and statistical reliability of the claimed improvements can be assessed.
minor comments (2)
  1. The acronym 'MosaicADAPT-QAOA' and the precise mechanism for subset selection of mixer operators should be defined on first use in the introduction rather than deferred to the methods section.
  2. The representation of circuits fed to the generative model (e.g., gate-sequence tokenization, depth encoding) is not described in the abstract; a concise summary of the input/output format would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of our results and strengthen the empirical claims. We address each major point below and commit to the indicated revisions.

read point-by-point responses

  1. Referee: The training corpus is generated entirely by the authors' own MosaicADAPT-QAOA procedure. This creates a potential circularity: any reported advantage of Q3SAT-GPT over MosaicADAPT-QAOA could simply reflect faster inference-time approximation of the same adaptive selection rule rather than discovery of independent circuit logic. The experimental section should therefore include direct comparisons of approximation ratios and depths achieved by Q3SAT-GPT versus the original MosaicADAPT-QAOA outputs on the same held-out instances, together with an analysis of how often the generated circuits differ structurally from the training examples.

    Authors: We agree that this comparison is necessary to rule out simple replication of the training procedure. In the revised manuscript we will add a new subsection reporting approximation ratios and circuit depths for Q3SAT-GPT and MosaicADAPT-QAOA on the same held-out instances. We will also include a structural analysis (e.g., operator-set overlap and circuit-edit-distance statistics) quantifying how frequently the generated circuits differ from the training examples. revision: yes

  2. Referee: The abstract states that the framework 'attains strong solution quality' and 'scales significantly better' than both the adaptive construction procedure and conventional variational baselines, yet no numerical values, problem sizes, number of instances, or statistical tests are supplied. The results section must contain explicit tables or figures reporting approximation ratios, circuit depths, wall-clock scaling, and baseline details (including standard QAOA with fixed p) so that the magnitude and statistical reliability of the claimed improvements can be assessed.

    Authors: We acknowledge that the current manuscript does not supply the requested quantitative details. We will revise the abstract to include concrete numerical values (approximation ratios, scaling exponents, problem sizes) and expand the results section with tables and figures that report approximation ratios, circuit depths, wall-clock scaling, the number of instances tested, and direct comparisons to standard QAOA with fixed p, together with appropriate statistical tests. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces MosaicADAPT-QAOA as a new procedure to generate training circuits and then trains the Q3SAT-GPT generative model on that data. The central experimental claim is that the resulting model produces effective circuits for unseen Max-E3-SAT instances at inference time without further variational optimization and scales better than both the adaptive procedure and standard QAOA. This does not constitute a derivation that reduces by construction to its inputs: the training data generation and the model's generalization performance are distinct steps, with the latter being an empirical outcome that can be (and is claimed to be) verified independently on held-out instances. No self-definitional equations, fitted inputs renamed as predictions, load-bearing self-citations, or smuggled ansatzes appear in the abstract or described framework.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The central claim depends on the unproven assumption that patterns learned from the authors' custom adaptive QAOA circuits generalize to new Max-E3-SAT instances and that the generative model adds value beyond simply replaying the training distribution.

axioms (1)
  • domain assumption QAOA-style ansatze with selected mixer subsets can produce high-performing shallow circuits for Max-E3-SAT
    Invoked when creating the training data for the generative model.
invented entities (2)
  • Q3SAT-GPT no independent evidence
    purpose: Generative model that directly outputs quantum circuits for 3-SAT
    New model introduced to replace variational optimization at inference time.
  • MosaicADAPT-QAOA no independent evidence
    purpose: Adaptive QAOA variant that selects subsets of mixer operators to build low-depth circuits
    New construction procedure used to generate training data.

pith-pipeline@v0.9.0 · 5510 in / 1323 out tokens · 55384 ms · 2026-05-07T08:58:01.306903+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

49 extracted references · 17 canonical work pages · 2 internal anchors

  1. [1]

    Strengths and weaknesses of quantum computing,

    C. H. Bennett, E. Bernstein, G. Brassard, and U. Vazirani, “Strengths and weaknesses of quantum computing,”SIAM journal on Computing, vol. 26, no. 5, pp. 1510–1523, 1997

    1997

  2. [2]

    Complexity limitations on quantum compu- tation,

    L. Fortnow and J. Rogers, “Complexity limitations on quantum compu- tation,”Journal of Computer and System Sciences, vol. 59, no. 2, pp. 240–252, 1999

    1999

  3. [3]

    How much structure is needed for huge quantum speedups?

    S. Aaronson, “How much structure is needed for huge quantum speedups?”arXiv preprint arXiv:2209.06930, 2022

    work page arXiv 2022

  4. [4]

    Report for the ASCR Workshop on Basic Research Needs in Quantum Computing and Networking-2023,

    P. Lougovski, O. Parekh, J. Broz, M. Byrd, J. C. Chapman, Y . Chembo, W. A. de Jong, E. Figueroa, T. S. Humble, J. Larsonet al., “Report for the ASCR Workshop on Basic Research Needs in Quantum Computing and Networking-2023,” USDOE Office of Science (SC)(United States), Tech. Rep., 2023

    2023

  5. [5]

    Assessing requirements to scale to practical quantum advantage

    M. E. Beverland, P. Murali, M. Troyer, K. M. Svore, T. Hoefler, V . Kliuchnikov, G. H. Low, M. Soeken, A. Sundaram, and A. Vaschillo, “Assessing requirements to scale to practical quantum advantage,”arXiv preprint arXiv:2211.07629, 2022

    work page internal anchor Pith review arXiv 2022

  6. [6]

    Early fault- tolerant quantum computing,

    A. Katabarwa, K. Gratsea, A. Caesura, and P. D. Johnson, “Early fault- tolerant quantum computing,”PRX quantum, vol. 5, no. 2, p. 020101, 2024

    2024

  7. [7]

    Quantum computing for fi- nance.Nature Reviews Physics, 5(8):450–465, 2023

    D. Herman, C. Googin, X. Liu, Y . Sun, A. Galda, I. Safro, M. Pistoia, and Y . Alexeev, “Quantum computing for finance,”Nature Reviews Physics, vol. 5, no. 8, p. 450–465, Jul. 2023. [Online]. Available: http://dx.doi.org/10.1038/s42254-023-00603-1

    work page doi:10.1038/s42254-023-00603-1 2023

  8. [8]

    Network community detection on small quantum computers,

    R. Shaydulin, H. Ushijima-Mwesigwa, I. Safro, S. Mniszewski, and Y . Alexeev, “Network community detection on small quantum computers,”Advanced Quantum Technologies, vol. 2, no. 9, p. 1900029,

  9. [9]

    Available: https://dx.doi.org/10.1002/qute.201900029

    [Online]. Available: https://dx.doi.org/10.1002/qute.201900029

    work page doi:10.1002/qute.201900029

  10. [10]

    Towards practical applications in quantum computational biology,

    A. Fedorov and M. Gelfand, “Towards practical applications in quantum computational biology,”Nature Computational Science, vol. 1, no. 2, pp. 114–119, 2021

    2021

  11. [11]

    Artificial intelligence for quantum computing,

    Y . Alexeev, M. H. Farag, T. L. Patti, M. E. Wolf, N. Ares, A. Aspuru- Guzik, S. C. Benjamin, Z. Cai, S. Cao, C. Chamberlandet al., “Artificial intelligence for quantum computing,”Nature Communications, vol. 16, no. 1, p. 10829, 2025

    2025

  12. [12]

    Generative ai for quantum circuits and quantum code: A technical review and taxonomy,

    J. Merilehto, “Generative ai for quantum circuits and quantum code: A technical review and taxonomy,”arXiv preprint arXiv:2603.16216, 2026

    work page arXiv 2026

  13. [13]

    Some optimal inapproximability results,

    J. H ˚astad, “Some optimal inapproximability results,”Journal of the ACM (JACM), vol. 48, no. 4, pp. 798–859, 2001

    2001

  14. [14]

    SAT Competitions,

    H. van Maaren and J. Franco, “SAT Competitions,” https:// satcompetition.github.io/

  15. [15]

    The international sat solver competitions,

    M. J ¨arvisalo, D. Le Berre, O. Roussel, and L. Simon, “The international sat solver competitions,”AI Magazine, vol. 33, no. 1, pp. 89–92, 2012

    2012

  16. [17]

    Challenges and opportunities in quantum optimization,

    A. Abbas, A. Ambainis, B. Augustino, A. B ¨artschi, H. Buhrman, C. Coffrin, G. Cortiana, V . Dunjko, D. J. Egger, B. G. Elmegreenet al., “Challenges and opportunities in quantum optimization,”Nature Reviews Physics, pp. 1–18, 2024

    2024

  17. [18]

    Graph representation learning for parameter transferability in quantum approximate optimiza- tion algorithm,

    J. Falla, Q. Langfitt, Y . Alexeev, and I. Safro, “Graph representation learning for parameter transferability in quantum approximate optimiza- tion algorithm,”Quantum Machine Intelligence, vol. 6, no. 2, p. 46, 2024

    2024

  18. [20]

    Transferability of optimal qaoa parameters between random graphs,

    A. Galda, X. Liu, D. Lykov, Y . Alexeev, and I. Safro, “Transferability of optimal qaoa parameters between random graphs,” in2021 IEEE International Conference on Quantum Computing and Engineering (QCE). IEEE, 2021, pp. 171–180

    2021

  19. [21]

    Warm-starting quantum optimization,

    D. J. Egger, J. Mare ˇcek, and S. Woerner, “Warm-starting quantum optimization,”Quantum, vol. 5, p. 479, 2021

    2021

  20. [22]

    Multi-angle quantum approximate optimization algorithm,

    R. Herrman, P. C. Lotshaw, J. Ostrowski, T. S. Humble, and G. Siopsis, “Multi-angle quantum approximate optimization algorithm,”Scientific Reports, vol. 12, no. 1, p. 6781, 2022

    2022

  21. [23]

    Quantum approximate optimization algorithm with sparsified phase operator,

    X. Liu, R. Shaydulin, and I. Safro, “Quantum approximate optimization algorithm with sparsified phase operator,” in2022 IEEE International Conference on Quantum Computing and Engineering (QCE). IEEE, 2022, pp. 133–141. Parameter Value Description layer_stopper_max20Maximum Adaptations (Number of layers) permitted in the circuit floor_stopper_energy−∞Floor...

    2022

  22. [24]

    Hybrid quantum- classical algorithms for approximate graph coloring,

    S. Bravyi, A. Kliesch, R. Koenig, and E. Tang, “Hybrid quantum- classical algorithms for approximate graph coloring,”Quantum, vol. 6, p. 678, 2022

    2022

  23. [25]

    Adaptive quantum approximate optimization algorithm for solving combinatorial problems on a quantum computer,

    L. Zhu, H. L. Tang, G. S. Barron, F. A. Calderon-Vargas, N. J. Mayhall, E. Barnes, and S. E. Economou, “Adaptive quantum approximate optimization algorithm for solving combinatorial problems on a quantum computer,”Phys. Rev. Res., vol. 4, p. 033029, Jul 2022. [Online]. Available: https://link.aps.org/doi/10.1103/PhysRevResearch.4.033029

    work page doi:10.1103/physrevresearch.4.033029 2022

  24. [26]

    Layer VQE: A variational approach for combinatorial optimization on noisy quantum computers,

    X. Liu, A. Angone, R. Shaydulin, I. Safro, Y . Alexeev, and L. Cincio, “Layer VQE: A variational approach for combinatorial optimization on noisy quantum computers,”IEEE Transactions on Quantum Engineer- ing, vol. 3, pp. 1–20, 2022

    2022

  25. [27]

    Feedback-based quantum optimization,

    A. B. Magann, K. M. Rudinger, M. D. Grace, and M. Sarovar, “Feedback-based quantum optimization,”Physical Review Letters, vol. 129, no. 25, p. 250502, 2022

    2022

  26. [28]

    Tetris-adapt-vqe: An adaptive algorithm that yields shal- lower, denser circuit ans ¨atze,

    P. G. Anastasiou, Y . Chen, N. J. Mayhall, E. Barnes, and S. E. Economou, “Tetris-adapt-vqe: An adaptive algorithm that yields shal- lower, denser circuit ans ¨atze,”Physical Review Research, vol. 6, no. 1, p. 013254, 2024

    2024

  27. [29]

    QAOA-GPT: Efficient generation of adaptive and regular quantum ap- proximate optimization algorithm circuits,

    I. Tyagin, M. H. Farag, K. Sherbert, K. Shirali, Y . Alexeev, and I. Safro, “QAOA-GPT: Efficient generation of adaptive and regular quantum ap- proximate optimization algorithm circuits,” in2025 IEEE International Conference on Quantum Computing and Engineering (QCE), vol. 1. IEEE, 2025, pp. 1505–1515

    2025

  28. [30]

    Approximation algorithms for combinatorial problems,

    D. S. Johnson, “Approximation algorithms for combinatorial problems,” Journal of Computer and System Sciences, vol. 9, no. 3, pp. 256– 278, 1974. [Online]. Available: https://www.sciencedirect.com/science/ article/pii/S0022000074800449

    1974

  29. [31]

    Optimization, approximation, and complexity classes,

    C. H. Papadimitriou and M. Yannakakis, “Optimization, approximation, and complexity classes,”Journal of Computer and System Sciences, vol. 43, no. 3, pp. 425–440, 1991. [Online]. Available: https: //www.sciencedirect.com/science/article/pii/002200009190023X

    work page arXiv 1991

  30. [32]

    Satlib: An online resource for research on sat,

    H. H. Hoos and T. St ¨utzle, “Satlib: An online resource for research on sat,”Sat, vol. 2000, pp. 283–292, 2000

    2000

  31. [33]

    Balanced random sat benchmarks,

    I. Spence, “Balanced random sat benchmarks,”SAT COMPETITION 2017, vol. 53, 2017

    2017

  32. [34]

    Hard and easy distributions of sat problems,

    D. Mitchell, B. Selman, H. Levesqueet al., “Hard and easy distributions of sat problems,” inAaai, vol. 92, 1992, pp. 459–465

    1992

  33. [35]

    A Quantum Approximate Optimization Algorithm

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

    work page internal anchor Pith review arXiv 2014

  34. [36]

    For fixed control parameters the quantum approximate optimization algorithm’s objective function value concentrates for typical instances,

    F. G. S. L. Brandao, M. Broughton, E. Farhi, S. Gutmann, and H. Neven, “For fixed control parameters the quantum approximate optimization algorithm’s objective function value concentrates for typical instances,”

  35. [37]

    Available: https://arxiv.org/abs/1812.04170

    [Online]. Available: https://arxiv.org/abs/1812.04170

    work page arXiv

  36. [38]

    Similarity-based parameter transferability in the quantum approximate optimization algorithm,

    A. Galda, E. Gupta, J. Falla, X. Liu, D. Lykov, Y . Alexeev, and I. Safro, “Similarity-based parameter transferability in the quantum approximate optimization algorithm,”Frontiers in Quantum Science and Technology, vol. V olume 2 - 2023, 2023. [Online]. Available: https://www.frontiersin.org/journals/ quantum-science-and-technology/articles/10.3389/frqst....

    work page doi:10.3389/frqst.2023.1200975 2023

  37. [39]

    Grimsley, Sophia E

    H. R. Grimsley, S. E. Economou, E. Barnes, and N. J. Mayhall, “An adaptive variational algorithm for exact molecular simulations on a quantum computer,”Nature Communications, vol. 10, no. 1, Jul. 2019. [Online]. Available: http://dx.doi.org/10.1038/s41467-019-10988-2

    work page doi:10.1038/s41467-019-10988-2 2019

  38. [40]

    Obstacles to variational quantum optimization from symmetry protection,

    S. Bravyi, A. Kliesch, R. Koenig, and E. Tang, “Obstacles to variational quantum optimization from symmetry protection,”Physical review let- ters, vol. 125, no. 26, p. 260505, 2020

    2020

  39. [41]

    Quantum Computation by Adiabatic Evolution

    E. Farhi, J. Goldstone, S. Gutmann, and M. Sipser, “Quantum computation by adiabatic evolution,” 2000. [Online]. Available: https://arxiv.org/abs/quant-ph/0001106

    work page Pith review arXiv 2000

  40. [42]

    24 Jure Leskovec, Jon M

    S. Lamm, P. Sanders, C. Schulz, D. Strash, and R. F. Werneck, “Finding near-optimal independent sets at scale,”J. Heuristics, vol. 23, no. 4, pp. 207–229, 2017. [Online]. Available: https: //doi.org/10.1007/s10732-017-9337-x

    work page doi:10.1007/s10732-017-9337-x 2017

  41. [43]

    finding near-optimal weight independent sets at scale,

    ernestine großmann, sebastian lamm, christian schulz, and darren strash, “finding near-optimal weight independent sets at scale,” in proceedings of the genetic and evolutionary computation conference, gecco 2023, lisbon, portugal, july 15-19, 2023, sara silva and lu´ıs paquete, Eds. acm, 2023, pp. 293–302. [Online]. Available: https://doi.org/10.1145/3583...

    work page doi:10.1145/3583131.3590353 2023

  42. [44]

    Hespe, C

    D. Hespe, C. Schulz, and D. Strash, “Scalable kernelization for maximum independent sets,”ACM Journal of Experimental Algorithmics, vol. 24, no. 1, pp. 1.16:1–1.16:22, 2019. [Online]. Available: https://doi.org/10.1145/3355502

    work page doi:10.1145/3355502 2019

  43. [45]

    Nocedal and S

    J. Nocedal and S. J. Wright,Numerical Optimization. New York: Springer-Verlag, 1999

    1999

  44. [46]

    Gurobi Optimizer Reference Manual,

    Gurobi Optimization, LLC, “Gurobi Optimizer Reference Manual,”

  45. [47]

    Available: https://www.gurobi.com

    [Online]. Available: https://www.gurobi.com

  46. [48]

    Satqubolib: A python framework for creating and benchmarking (max- )3sat qubos,

    S. Zielinski, M. Benkard, J. N ¨ußlein, C. Linnhoff-Popien, and S. Feld, “Satqubolib: A python framework for creating and benchmarking (max- )3sat qubos,” inInnovations for Community Services, F. Phillipson, G. Eichler, C. Erfurth, and G. Fahrnberger, Eds. Cham: Springer Nature Switzerland, 2024, pp. 48–66

    2024

  47. [49]

    Adapt-qaoa with a classically inspired initial state,

    V . K. Sridhar, Y . Chen, B. Gard, E. Barnes, and S. E. Economou, “Adapt-qaoa with a classically inspired initial state,”arXiv preprint arXiv:2310.09694, 2023

    work page arXiv 2023

  48. [50]

    Russell, S

    D. Selsam, M. Lamm, B. B ¨unz, P. Liang, L. de Moura, and D. L. Dill, “Learning a sat solver from single-bit supervision,” 2019. [Online]. Available: https://arxiv.org/abs/1802.03685

    work page arXiv 2019

  49. [51]

    Sat-gatv2: A dynamic attention-based graph neural network for solving boolean satisfiability problem,

    W. Chang and W. Liu, “Sat-gatv2: A dynamic attention-based graph neural network for solving boolean satisfiability problem,” Electronics, vol. 14, no. 3, 2025. [Online]. Available: https: //www.mdpi.com/2079-9292/14/3/423

    2025