Pith. sign in

REVIEW 2 major objections 2 minor 46 references

Gaussian process surrogates with active learning calibrate trapped-ion entangling gates from noisy fidelity data.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Bayesian optimization with Gaussian process surrogate accelerates numerical calibration of Mølmer-Sørensen gate parameters, with performance tied to quantum projection noise.

T0 review reviewed 2026-07-02 challenge →

load-bearing objection GP surrogate with active learning speeds up simulated MS gate calibration but stays inside a known model with no mismatch test. the 2 major comments →

arxiv 2607.00284 v1 pith:JNAHKL6N submitted 2026-07-01 quant-ph physics.app-ph

Active Learning for Calibrating Entangling Gates via Surrogate-Based Optimization

classification quant-ph physics.app-ph
keywords active learningBayesian optimizationGaussian processquantum gate calibrationtrapped ionsMølmer-Sørensen gatesurrogate modelsentangling gates
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper presents an active learning framework that uses Bayesian optimization and a Gaussian process surrogate to tune laser amplitudes and frequencies for the trapped-ion Mølmer-Sørensen gate. It shows that the surrogate can capture the gate's Hamiltonian dynamics from limited measurements even in the presence of quantum projection noise. Active learning selects the next parameter sets to test, reducing the total number of experiments needed to reach high fidelity compared with non-adaptive searches. The speed of convergence and the final fidelity achieved depend directly on the noise level in the data. This establishes surrogate models as a practical tool for on-device quantum gate calibration when exact physical models are unavailable.

Core claim

We show that a Gaussian process can model the Hamiltonian dynamics. The addition of active learning accelerates the discovery of the optimal parameter set with speed and final fidelity dependent on the quantum projection noise of the data.

What carries the argument

Gaussian process surrogate inside a Bayesian optimization loop that predicts gate fidelity from control parameters and chooses the next measurement point to reduce model uncertainty while pursuing higher fidelity.

Load-bearing premise

A Gaussian process surrogate trained on limited noisy measurements will accurately enough approximate the underlying gate fidelity landscape to guide optimization toward the true optimum without excessive additional experiments.

What would settle it

Execute the active-learning procedure on a simulated Mølmer-Sørensen gate whose true optimum parameters and fidelity landscape are known in advance; if the method consistently reaches fidelities above 99 percent using far fewer evaluations than exhaustive search or random sampling, the claim holds; otherwise it does not.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Fewer on-device experiments are required to reach high-fidelity entangling gates when exact Hamiltonian models are missing.
  • The final achieved fidelity is limited by the quantum projection noise present in the fidelity estimates.
  • The surrogate approach remains effective even when small deviations exist between the implemented and ideal Hamiltonians.
  • Bayesian optimization guided by Gaussian process models becomes a viable calibration strategy for other parameter-sensitive quantum operations.

Where Pith is reading between the lines

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

  • The same surrogate-plus-active-learning loop could be tested on other trapped-ion gates or on superconducting qubit platforms to check transferability.
  • Accounting explicitly for projection noise in the acquisition function may further improve sample efficiency on real hardware.
  • If the method scales, it could reduce calibration overhead in larger ion-trap processors where exhaustive parameter searches become prohibitive.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper proposes an active learning framework based on Bayesian optimization with a Gaussian process surrogate to calibrate control parameters (laser amplitude and frequencies) for the trapped-ion Mølmer-Sørensen entangling gate. It validates the approach via numerical simulations of the gate dynamics with added projection noise, claiming that the GP models the Hamiltonian dynamics and that active learning accelerates discovery of high-fidelity parameters, with speed and final fidelity depending on noise level.

Significance. If validated beyond simulation, the method could reduce calibration overhead for quantum gates by using data-driven surrogates instead of full Hamiltonian modeling. The explicit dependence on projection noise is a useful quantitative insight. However, the exclusive use of matched-model simulations limits the immediate significance for real-device calibration.

major comments (2)
  1. [Results (numerical validation)] Results section (numerical validation): The experiments generate data from the exact known Hamiltonian of the Mølmer-Sørensen gate and add only projection noise. This setup does not test robustness to model mismatch (e.g., unmodeled drifts or crosstalk), which is the central motivation stated in the introduction for on-device calibration. The claim that the GP 'can model the Hamiltonian dynamics' therefore holds only under the assumption that the simulation landscape matches experiment.
  2. [Abstract] Abstract and main results: No quantitative metrics are reported (e.g., number of evaluations to reach target fidelity, achieved fidelities with error bars, or direct comparisons to grid search or other baselines). Without these, the claimed acceleration cannot be assessed for practical utility.
minor comments (2)
  1. [Methods] The choice of GP kernel and hyperparameter optimization procedure should be stated explicitly, ideally with an equation, to allow reproduction.
  2. [Figures] Figure captions could clarify whether plotted fidelities are noisy measurements or GP predictions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments. We address each major point below.

read point-by-point responses
  1. Referee: [Results (numerical validation)] Results section (numerical validation): The experiments generate data from the exact known Hamiltonian of the Mølmer-Sørensen gate and add only projection noise. This setup does not test robustness to model mismatch (e.g., unmodeled drifts or crosstalk), which is the central motivation stated in the introduction for on-device calibration. The claim that the GP 'can model the Hamiltonian dynamics' therefore holds only under the assumption that the simulation landscape matches experiment.

    Authors: We agree that the numerical experiments use data generated from the exact known Hamiltonian with only projection noise added and do not introduce model mismatch such as drifts or crosstalk. This is a genuine limitation of the current validation, which isolates the effect of projection noise in a matched-model setting rather than fully replicating real-device conditions. The introduction motivates the work by the difficulty of exact Hamiltonian modeling, but the study is a controlled numerical demonstration. We will revise the manuscript to explicitly acknowledge this scope, clarify that the GP modeling result holds under the matched assumption, and add discussion of future extensions (e.g., perturbed Hamiltonians or hardware data) to address model mismatch. revision: yes

  2. Referee: [Abstract] Abstract and main results: No quantitative metrics are reported (e.g., number of evaluations to reach target fidelity, achieved fidelities with error bars, or direct comparisons to grid search or other baselines). Without these, the claimed acceleration cannot be assessed for practical utility.

    Authors: The results section contains figures that quantify convergence speed versus noise level and compare active learning to non-active baselines, but we acknowledge that the abstract itself reports no specific numbers, error bars, or explicit baseline comparisons. We will revise the abstract to include key quantitative metrics (e.g., typical evaluations needed to reach target fidelity, achieved fidelities with noise dependence, and comparisons to grid/random search) drawn from the existing results, making the acceleration claim easier to assess. revision: yes

Circularity Check

0 steps flagged

No circularity: standard surrogate optimization applied to simulated calibration task

full rationale

The paper presents a Bayesian optimization procedure using a Gaussian process surrogate to tune control parameters for a trapped-ion gate, with validation performed by generating noisy measurements from an exactly known numerical model of the Mølmer-Sørensen Hamiltonian and comparing recovered parameters against the known optimum. No derivation step reduces to a self-definition, a fitted quantity renamed as a prediction, or a load-bearing self-citation; the GP model is trained on external simulated data whose underlying dynamics are independent of the surrogate itself. The numerical experiments therefore constitute an external benchmark rather than an internal tautology.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

Only the abstract is available, so the ledger is limited to assumptions explicitly stated there.

free parameters (1)
  • Gaussian process kernel hyperparameters
    Standard in GP surrogate modeling; must be fitted or chosen during the active learning loop.
axioms (2)
  • domain assumption A Gaussian process can model the Hamiltonian dynamics of the Mølmer-Sørensen gate
    Explicitly claimed in the abstract as a result shown by the work.
  • domain assumption Active learning speed and final fidelity are governed by quantum projection noise
    Stated as a dependency in the abstract.

reviewed 2026-07-02 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Active Learning for Calibrating Entangling Gates via Surrogate-Based Optimization." pith.science (2026). https://pith.science/paper/JNAHKL6N

@misc{pith2026260700284,
  author       = {Pith},
  title        = {Pith review of: Active Learning for Calibrating Entangling Gates via Surrogate-Based Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JNAHKL6N}},
  note         = {Machine review of arXiv:2607.00284}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

The fidelity of a quantum gate is sensitive to small deviations in the physical control parameters. Unfortunately, it is generally difficult to exactly model the implemented Hamiltonian for a set of user-defined parameters, necessitating on-device calibration. Here, we present an active learning framework based on Bayesian optimization with a Gaussian Process surrogate to find the optimal parameter set. We validate the technique through numerical calibration of the laser amplitude and frequencies that implement the trapped-ion M{\o}lmer S{\o}rensen gate. We show that a Gaussian process can model the Hamiltonian dynamics. The addition of active learning accelerates the discovery of the optimal parameter set with speed and final fidelity dependent on the quantum projection noise of the data. These results establish the utility of active learning and surrogate models for quantum calibration and control.

Figures

Figures reproduced from arXiv: 2607.00284 by Ana Larra\~naga, Caleb Walton, Filippo Zacchei, Patricia Garc\'ia-Caspue\~nas, Sara Mouradian, Steven L. Brunton.

Figure 1
Figure 1. Figure 1: FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: (a,b) show examples of one-dimensional cuts through the learned GP posterior, when N = 100, at dif￾ferent points during the active learning process showing the evolution of the model landscape through measure￾ments of the quantum system. The surrogate is able to immediately identify the frequency peaks, and narrows its uncertainty around ∆δ = ∆ωcl = 0. In contrast, the am￾plitude cut remains broad because … view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗

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

46 extracted references · 46 canonical work pages · 1 internal anchor

  1. [1]

    Preskill, Quantum computing in the NISQ era and beyond, Quantum2, 79 (2018)

    J. Preskill, Quantum computing in the NISQ era and beyond, Quantum2, 79 (2018)

  2. [2]

    A. G. Fowler, M. Mariantoni, J. M. Martinis, and A. N. Cleland, Surface codes: Towards practical large-scale quantum computation, Physical Review A86, 032324 (2012)

  3. [3]

    C. G. Yale, A. D. Burch, M. N. Chow, B. P. Ruzic, D. S. Lobser, B. K. McFarland, M. C. Revelle, and S. M. Clark, Realization and calibration of continuously parameter- ized two-qubit gates on a trapped-ion quantum processor, IEEE Transactions on Quantum Engineering (2025)

  4. [4]

    Gerster, F

    L. Gerster, F. Mart´ ınez-Garc´ ıa, P. Hrmo, M. W. van Mourik, B. Wilhelm, D. Vodola, M. M¨ uller, R. Blatt, P. Schindler, and T. Monz, Experimental bayesian calibra- tion of trapped-ion entangling operations, PRX Quantum 3, 020350 (2022)

  5. [5]

    Nielsen, J

    E. Nielsen, J. K. Gamble, K. Rudinger, T. Scholten, K. Young, and R. Blume-Kohout, Gate set tomography, Quantum5, 557 (2021)

  6. [6]

    W. Yu, J. Sun, Z. Han, and X. Yuan, Robust and efficient hamiltonian learning, Quantum7, 1045 (2023)

  7. [7]

    Mohseni, A

    M. Mohseni, A. T. Rezakhani, and D. A. Lidar, Quantum- process tomography: Resource analysis of different strate- gies, Physical Review A77, 032322 (2008)

  8. [8]

    Navon, N

    N. Navon, N. Akerman, S. Kotler, Y. Glickman, and R. Ozeri, Quantum process tomography of a mølmer- sørensen interaction, Phys. Rev. A90, 010103 (2014)

  9. [9]

    Proctor, M

    T. Proctor, M. Revelle, E. Nielsen, K. Rudinger, D. Lob- ser, P. Maunz, R. Blume-Kohout, and K. Young, Detect- ing and tracking drift in quantum information processors, Nature Communications11, 5396 (2020)

  10. [10]

    M. A. Rol, C. C. Bultink, T. E. O’Brien, S. R. de Jong, L. S. Theis, X. Fu, F. Luthi, R. F. L. Vermeulen, J. C. de Sterke, A. Bruno, D. Deurloo, R. N. Schouten, F. K. Wilhelm, and L. DiCarlo, Restless tuneup of high-fidelity qubit gates, Phys. Rev. Appl.7, 041001 (2017). 10

  11. [11]

    Glaser, F

    N. Glaser, F. Roy, I. Tsitsilin, L. Koch, N. Bruckmoser, J. Schirk, J. Romeiro, G. Huber, F. Wallner, M. Singh, G. Krylov, A. Marx, L. S¨ odergren, C. Schneider, M. Wern- inghaus, and S. Filipp, Closed-loop optimization for high- fidelity controlled-z gates in superconducting qubits, Phys. Rev. Appl.24, 024048 (2025)

  12. [12]

    G. Pang, L. Yang, and G. E. Karniadakis, Neural-net- induced gaussian process regression for function approx- imation and pde solution, Journal of Computational Physics384, 270 (2019)

  13. [13]

    S. L. Brunton, N. Zolman, J. N. Kutz, and U. Fasel, Ma- chine learning for sparse nonlinear modeling and control, Annual Review of Control, Robotics, and Autonomous Systems8, 127 (2025)

  14. [14]

    Booker, Design and analysis of computer experiments, in7th AIAA/USAF/NASA/ISSMO Symposium on Mul- tidisciplinary Analysis and Optimization(1998) p

    A. Booker, Design and analysis of computer experiments, in7th AIAA/USAF/NASA/ISSMO Symposium on Mul- tidisciplinary Analysis and Optimization(1998) p. 4757

  15. [15]

    D. R. Jones, M. Schonlau, and W. J. Welch, Efficient global optimization of expensive black-box functions, Jour- nal of Global Optimization13, 455 (1998)

  16. [16]

    P. I. Frazier, A Tutorial on Bayesian Optimization (2018), arXiv:1807.02811 [stat]

  17. [17]

    Snoek, H

    J. Snoek, H. Larochelle, and R. P. Adams, Practical Bayesian Optimization of Machine Learning Algorithms, inAdvances in Neural Information Processing Systems, Vol. 25 (Curran Associates, Inc., 2012)

  18. [18]

    L. Fung, U. Fasel, and M. Juniper, Rapid bayesian iden- tification of sparse nonlinear dynamics from scarce and noisy data, Proceedings of the Royal Society A: Mathe- matical, Physical and Engineering Sciences481, 20240200 (2025), https://royalsocietypublishing.org/rspa/article- pdf/doi/10.1098/rspa.2024.0200/1230512/rspa.2024.0200.pdf

  19. [19]

    S. M. Hirsh, D. A. Barajas-Solano, and J. N. Kutz, Spar- sifying priors for bayesian uncertainty quantification in model discovery, Royal Society Open Science9, 211823 (2022)

  20. [20]

    Fasel, J

    U. Fasel, J. N. Kutz, B. W. Brunton, and S. L. Brunton, Ensemble-sindy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and con- trol, Proceedings of the Royal Society A478, 20210904 (2022)

  21. [21]

    S. L. Brunton, J. L. Proctor, and J. N. Kutz, Discovering governing equations from data by sparse identifica- tion of nonlinear dynamical systems, Proceedings of the National Academy of Sciences113, 3932 (2016), https://www.pnas.org/doi/pdf/10.1073/pnas.1517384113

  22. [22]

    Berkemeier, M

    T. Berkemeier, M. Kr¨ uger, A. Feinberg, M. M¨ uller, U. P¨ oschl, and U. K. Krieger, Accelerating models for multiphase chemical kinetics through machine learning with polynomial chaos expansion and neural networks, Geoscientific Model Development16, 2037 (2023)

  23. [23]

    C. K. Williams and C. E. Rasmussen,Gaussian processes for machine learning, Vol. 2 (MIT press Cambridge, MA, 2006)

  24. [24]

    Morita, S

    Y. Morita, S. Rezaeiravesh, N. Tabatabaei, R. Vinuesa, K. Fukagata, and P. Schlatter, Applying bayesian opti- mization with gaussian process regression to computa- tional fluid dynamics problems, Journal of Computational Physics449, 110788 (2022)

  25. [25]

    G. Chen, J. J. Jiao, Q. Liu, Z. Wang, and Y. Jin, Machine- learning-accelerated multi-objective design of fractured geothermal systems, Nexus1(2024)

  26. [26]

    F. Hase, L. M. Roch, C. Kreisbeck, and A. Aspuru-Guzik, Phoenics: a bayesian optimizer for chemistry, ACS central science4, 1134 (2018)

  27. [27]

    Burger, P

    B. Burger, P. M. Maffettone, V. V. Gusev, C. M. Aitchi- son, Y. Bai, X. Wang, X. Li, B. M. Alston, B. Li, R. Clowes,et al., A mobile robotic chemist, Nature583, 237 (2020)

  28. [28]

    D. Fan, G. Jodin, T. Consi, L. Bonfiglio, Y. Ma, L. Keyes, G. E. Karniadakis, and M. S. Triantafyllou, A robotic intelligent towing tank for learning complex fluid-structure dynamics, Science Robotics4, eaay5063 (2019)

  29. [29]

    Makarova, I

    A. Makarova, I. Usmanova, I. Bogunovic, and A. Krause, Risk-averse heteroscedastic bayesian optimization, inAd- vances in Neural Information Processing Systems, Vol. 34, edited by M. Ranzato, A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan (Curran Associates, Inc.,

  30. [30]

    K. P. Murphy,Machine learning: a probabilistic perspec- tive(MIT press, 2012)

  31. [31]

    W. M. Itano, J. C. Bergquist, J. J. Bollinger, J. M. Gilli- gan, D. J. Heinzen, F. L. Moore, M. G. Raizen, and D. J. Wineland, Quantum projection noise: Population fluctua- tions in two-level systems, Physical Review A47, 3554 (1993)

  32. [32]

    Q. V. Le, A. J. Smola, and S. Canu, Heteroscedastic gaussian process regression, inProceedings of the 22nd international conference on Machine learning(2005) pp. 489–496

  33. [33]

    Ankenman, B

    B. Ankenman, B. L. Nelson, and J. Staum, Stochastic kriging for simulation metamodeling, Operations research 58, 371 (2010)

  34. [34]

    D. R. Jones, M. Schonlau, and W. J. Welch, Efficient global optimization of expensive black-box functions, Jour- nal of Global optimization13, 455 (1998)

  35. [35]

    Shahriari, K

    B. Shahriari, K. Swersky, Z. Wang, R. P. Adams, and N. De Freitas, Taking the human out of the loop: A review of bayesian optimization, Proceedings of the IEEE104, 148 (2015)

  36. [36]

    Krause and J

    A. Krause and J. H¨ ubotter, Probabilistic artificial intelli- gence, arXiv preprint arXiv:2502.05244 (2025)

  37. [37]

    Ghorbani, M

    M. Ghorbani, M. Boley, P. Nakashima, and N. Birbilis, An active machine learning approach for optimal design of magnesium alloys using bayesian optimisation, Scientific Reports14, 8299 (2024)

  38. [38]

    Y. Wu, T. Ji, X. Lv, C. Zheng, Z. Ye, and F. Xie, Learning active flow control strategies of a swept wing by intelligent wind tunnel, Theoretical and Applied Mechanics Letters 14, 100543 (2024)

  39. [39]

    T. Ji, F. Jin, F. Xie, H. Zheng, X. Zhang, and Y. Zheng, Active learning of tandem flapping wings at optimizing propulsion performance, Physics of Fluids34(2022)

  40. [40]

    Leibfried, R

    D. Leibfried, R. Blatt, C. Monroe, and D. Wineland, Quantum dynamics of single trapped ions, Reviews of Modern Physics75, 281 (2003)

  41. [41]

    Sørensen and K

    A. Sørensen and K. Mølmer, Quantum computation with ions in thermal motion, Physical Review Letters82, 1971 (1999)

  42. [42]

    Sørensen and K

    A. Sørensen and K. Mølmer, Entanglement and quantum computation with ions in thermal motion, Physical Review A62, 022311 (2000)

  43. [43]

    Mart´ ınez-Garc´ ıa, L

    F. Mart´ ınez-Garc´ ıa, L. Gerster, D. Vodola, P. Hrmo, T. Monz, P. Schindler, and M. M¨ uller, Analytical and ex- perimental study of center-line miscalibrations in Mølmer– Sørensen gates, Phys. Rev. A105, 032437 (2022)

  44. [44]

    M. Kang, Y. Wang, C. Fang, B. Zhang, O. Khosravani, J. Kim, and K. R. Brown, Designing filter functions of 11 frequency-modulated pulses for high-fidelity two-qubit gates in ion chains, Phys. Rev. Appl.19, 014014 (2023)

  45. [45]

    IonSim.jl (2022), [Online; accessed 17. Jun. 2026]

  46. [46]

    Booker, E

    A. Booker, E. Cramer, P. Frank, J. Gablon- sky, and J. Dennis, Movars: Multidisciplinary op- timization via adaptive response surfaces, in48th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference(2007) p. 1927. VII. APPENDIX A. Bayesian Optimization hyperparameter choices This section justifies the selection of several empirica...

This paper was first reviewed by grok-4.3 on July 2, 2026.