pith. sign in

arxiv: 2605.04203 · v1 · submitted 2026-05-05 · 🪐 quant-ph

GHZ is All You Need: Quantum Sensing with VISTA

Oskar Novak , Christos N. Gagatsos , Narayanan Rengaswamy This is my paper

Pith reviewed 2026-05-08 17:07 UTC · model grok-4.3

classification 🪐 quant-ph
keywords quantum metrologyvariational quantum sensingLindblad dynamicsHeisenberg scalingswap testpassive sensingmulti-parameter estimationquantum twin ansatz
0
0 comments X

The pith

VISTA recovers near-Heisenberg scaling in noisy quantum sensing by comparing a passive probe to a physics-informed twin circuit.

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

Quantum metrology can in principle beat the standard quantum limit, but realistic noise typically reduces performance to that limit. VISTA lets a probe state evolve passively under Lindblad noise and compares it, via swap test, to a shallow parameterized quantum twin that is trained to reproduce the same dynamics. A classical optimizer then tunes only the physical parameters of interest, such as signal strength and decoherence rate, to minimize the loss. With high measurement shots the protocol temporarily restores near-Heisenberg scaling for moderate noise and a finite range of qubit numbers. A quasi-normalization step further sharpens the gradients so both signal and noise can be extracted simultaneously, and the method extends to multi-parameter vector sensing.

Core claim

By restricting variational optimization to the physical parameters of a passive probe and using a shallow twin ansatz that approximates Lindbladian evolution, VISTA extracts the coherent signal and environmental noise rate with low absolute error and achieves near-Heisenberg scaling over a finite range of system sizes when paired with a classical optimizer and sufficient shots.

What carries the argument

The quantum twin ansatz, a shallow parameterized circuit that mimics the probe's pure-state or open-system Lindbladian dynamics, whose overlap with the passive probe is measured by the swap test to supply gradients for parameter optimization.

If this is right

  • Practical sensors can operate without active control pulses or complex probe preparation.
  • Simultaneous estimation of signal and decoherence rates becomes feasible with low absolute error.
  • The protocol extends directly to multi-parameter vector metrology under transverse fields.
  • Barren plateaus are avoided by limiting the search space to physical parameters rather than arbitrary unitaries.
  • Resource-efficient sensing becomes viable on near- to intermediate-term devices with only shallow circuits.

Where Pith is reading between the lines

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

  • If the twin approximation continues to hold at larger qubit numbers, the finite range of near-Heisenberg scaling could be extended further.
  • Pairing VISTA with existing error-mitigation techniques might lengthen the useful sensing time before noise dominates.
  • The success of physics-informed twins suggests similar interpretable ansatze could improve variational methods in other quantum tasks such as state tomography or channel learning.
  • Hardware tests on current devices could quickly check whether the claimed scaling holds when shot noise and gate errors are included.

Load-bearing premise

The shallow parameterized quantum twin must approximate the open-system Lindblad dynamics of the passive probe well enough that the swap-test loss yields usable gradients for both signal and noise parameters.

What would settle it

A simulation or hardware experiment in which the achieved precision scaling remains at or near the standard quantum limit, or in which absolute errors in extracted signal and noise parameters stay large, for the moderate noise levels and system sizes where near-Heisenberg scaling was reported.

Figures

Figures reproduced from arXiv: 2605.04203 by Christos N. Gagatsos, Narayanan Rengaswamy, Oskar Novak.

Figure 1
Figure 1. Figure 1: FIG. 1. A diagram summarizing the VISTA metrology pro view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. For a moderate noise rate of view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Averaged Absolute error for a typical range of view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. FFT for extracting view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Circuit for the view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Comparison between the Unnormalized and Quasi view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Comparing the pure ans¨atz to the Quasi-Normalized view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13. Circuit for the view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14. Averaged absolute error for view at source ↗
read the original abstract

Quantum metrology holds the potential to enhance magnetic field sensing beyond current limits. However, in the presence of realistic noise, this advantage degrades to the Standard Quantum Limit. While recent algorithmic and variational techniques attempt to recover this scaling, they are hindered by stringent control requirements on the probe state that are infeasible in the near term, or by barren plateaus and interpretability issues inherent to black-box variational quantum circuits. Here, we introduce Variational Inference and Sensing with Twin Ans\"atze (VISTA), a closed-loop protocol that combines passive sensing, or where the probe state is left to evolve without any active control, with physics-informed variational optimization. In the VISTA framework, a probe state evolves under a Lindbladian master-equation, and is compared, via the Swap test, to a parameterized ``quantum twin", a shallow quantum circuit designed to mimic the underlying pure-state or Lindbladian master-equation dynamics. By restricting the optimization space to the physical parameters of interest, VISTA circumvents barren plateaus. We demonstrate that by coupling the protocol with a classical optimizer and high shot counts, VISTA can temporarily achieve near-Heisenberg scaling for moderately noisy qubits over a finite range of system sizes. Furthermore, we introduce a Quasi-Normalization technique that sharpens the loss gradients, enabling simultaneous extraction of both the coherent signal $\theta$ and the environmental noise rate $\gamma$ with low absolute error. Finally, we extend VISTA to the multi-parameter vector metrology regime, enabling simultaneous parameter extraction from a transverse-magnetic-field Hamiltonian. By eliminating the need for complex, open-loop control and processing, VISTA offers a highly practical, resource-efficient framework for near- to intermediate-term quantum sensors.

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

3 major / 3 minor

Summary. The manuscript introduces VISTA (Variational Inference and Sensing with Twin Ansätze), a closed-loop protocol combining passive probe evolution under a Lindblad master equation with a shallow parameterized quantum twin circuit. The twin is compared to the probe via swap test; optimization is restricted to physical parameters (signal θ and noise γ) to extract both quantities simultaneously. The paper claims that, with a classical optimizer and high shot counts, this yields near-Heisenberg scaling over finite system sizes in moderately noisy regimes, introduces a quasi-normalization step to sharpen gradients, and extends to multi-parameter vector metrology under a transverse-field Hamiltonian.

Significance. If the twin ansatz faithfully reproduces the target Lindblad dynamics, VISTA would provide a practical, control-light route to quantum-enhanced sensing on near-term hardware by avoiding active feedback and barren-plateaus issues. The restriction to physical parameters and the quasi-normalization technique are positive design choices that could make the method interpretable and resource-efficient; the finite-range near-Heisenberg claim, if robustly demonstrated, would be a useful benchmark for NISQ metrology protocols.

major comments (3)
  1. [Numerical demonstrations and twin-ansatz definition] The central claim that the shallow twin circuit faithfully approximates open-system Lindblad dynamics (and thereby supplies reliable gradients for joint θ,γ extraction) is load-bearing but insufficiently verified. Explicit fidelity metrics or state-vector comparisons between the twin circuit and the master-equation solution for γt ≳ 0.1 should be added, especially in the regime where the reported scaling window appears.
  2. [Scaling results] The near-Heisenberg scaling is stated to hold only temporarily and for finite N; the manuscript must delineate the precise range of N and γ where this occurs and demonstrate that the scaling is not an artifact of finite-shot swap-test noise or optimizer hyper-parameter tuning. Error bars on the extracted scaling exponent are required.
  3. [Quasi-normalization subsection] Quasi-normalization is introduced to sharpen loss gradients, yet its effect on the swap-test estimator when the twin ansatz deviates from the true Lindblad channel is not analyzed. If the underlying loss already misrepresents noise dependence, normalization may amplify rather than mitigate estimation bias; a short derivation or ablation showing gradient improvement under controlled mismatch is needed.
minor comments (3)
  1. [Methods] Notation for the swap-test loss and the quasi-normalization factor should be defined once in the main text with a single equation rather than re-introduced in multiple sections.
  2. [Multi-parameter results] The multi-parameter extension would benefit from a brief statement of the condition number of the extracted Fisher information matrix to quantify parameter correlation.
  3. [Discussion] A few sentences comparing VISTA’s resource overhead (shots, circuit depth) to existing variational metrology protocols would help readers assess practicality.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments, which have helped strengthen the presentation and verification of our results. We address each major comment point by point below, with revisions incorporated where appropriate to enhance the manuscript's rigor.

read point-by-point responses

  1. Referee: [Numerical demonstrations and twin-ansatz definition] The central claim that the shallow twin circuit faithfully approximates open-system Lindblad dynamics (and thereby supplies reliable gradients for joint θ,γ extraction) is load-bearing but insufficiently verified. Explicit fidelity metrics or state-vector comparisons between the twin circuit and the master-equation solution for γt ≳ 0.1 should be added, especially in the regime where the reported scaling window appears.

    Authors: We agree that additional explicit verification strengthens the central claim. In the revised manuscript we have added a dedicated subsection with state-vector fidelity plots and direct comparisons between the twin circuit output and the Lindblad master-equation solution. These metrics are shown for γt values from 0.05 to 0.5, confirming fidelities above 0.95 precisely in the parameter window used for the scaling demonstrations. This supports the reliability of the gradients for joint extraction. revision: yes

  2. Referee: [Scaling results] The near-Heisenberg scaling is stated to hold only temporarily and for finite N; the manuscript must delineate the precise range of N and γ where this occurs and demonstrate that the scaling is not an artifact of finite-shot swap-test noise or optimizer hyper-parameter tuning. Error bars on the extracted scaling exponent are required.

    Authors: We have revised the scaling section to explicitly state the operating window (N = 4–16 and γt ∈ [0.01, 0.2]) where the exponent reaches 1.8–1.9. To rule out artifacts we performed additional runs at 10^5 shots and across varied optimizer hyperparameters (learning rates and initializations); the scaling persists. Error bars on the extracted exponents, computed from ten independent runs with different random seeds, are now included and remain consistent with near-Heisenberg behavior. revision: yes

  3. Referee: [Quasi-normalization subsection] Quasi-normalization is introduced to sharpen loss gradients, yet its effect on the swap-test estimator when the twin ansatz deviates from the true Lindblad channel is not analyzed. If the underlying loss already misrepresents noise dependence, normalization may amplify rather than mitigate estimation bias; a short derivation or ablation showing gradient improvement under controlled mismatch is needed.

    Authors: We have added a short analytic derivation showing that, for small deviations (fidelity ≳ 0.9), quasi-normalization sharpens gradients while preserving unbiasedness because the normalization factor is computed from the twin circuit’s own statistics. We also include an ablation study with controlled 5–10 % mismatch between twin and Lindblad channel; the study demonstrates that quasi-normalization reduces estimation variance relative to the un-normalized loss without increasing absolute bias in θ and γ extraction. revision: yes

Circularity Check

0 steps flagged

No significant circularity in VISTA derivation chain

full rationale

The paper defines VISTA via an external swap-test loss between the physical Lindblad-evolved probe and a twin circuit whose parameters are restricted to the physical quantities θ and γ. The classical optimizer then extracts those parameters; the reported near-Heisenberg scaling for finite N is a numerical outcome of that optimization under high shot counts, not a quantity defined in terms of itself. Quasi-normalization is a post-processing sharpening of the existing loss gradients and does not create a self-referential loop. No self-definitional equations, no fitted inputs relabeled as predictions, and no load-bearing self-citations appear in the derivation. The central claims therefore remain independent of the inputs and receive the default low-circularity finding.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 2 invented entities

The protocol rests on standard open-quantum-system assumptions plus the new twin-circuit construction; no machine-checked proofs or external benchmarks are mentioned.

free parameters (1)
  • twin-circuit variational parameters
    Parameters of the shallow quantum twin are optimized classically to match probe dynamics; their number and initialization are not specified in the abstract.
axioms (2)
  • domain assumption Probe state evolves according to a Lindblad master equation
    Standard model for Markovian noise in quantum sensing; invoked to define the passive evolution.
  • standard math Swap test provides a faithful estimate of state overlap
    Well-known quantum circuit primitive used to define the loss function.
invented entities (2)
  • quantum twin no independent evidence
    purpose: Parameterized shallow circuit that mimics the noisy probe state for comparison
    New construct introduced to enable physics-informed optimization without active control.
  • quasi-normalization no independent evidence
    purpose: Technique to sharpen loss gradients for joint extraction of signal and noise
    New post-processing step claimed to improve optimization; no external validation cited.

pith-pipeline@v0.9.0 · 5619 in / 1424 out tokens · 32896 ms · 2026-05-08T17:07:14.259869+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

32 extracted references · 13 canonical work pages · 1 internal anchor

  1. [1]

    Using the conditions in (V.5), whereϕis matched to correspond to someγ ′, we see that Tr (ρ(θ1, γ)ρcirc(θ2, ϕγ′)) = 1 2 1 +e −2N(γ+γ ′) cos(2N∆θ) ,(VI.1) where ∆θ=θ 1 −θ 2

    Unnormalized Case We first look at the unnormalized trace overlap to jus- tify our normalization scheme. Using the conditions in (V.5), whereϕis matched to correspond to someγ ′, we see that Tr (ρ(θ1, γ)ρcirc(θ2, ϕγ′)) = 1 2 1 +e −2N(γ+γ ′) cos(2N∆θ) ,(VI.1) where ∆θ=θ 1 −θ 2. We now see that: QHS (ρθ) := lim ∆θ→0 − ∂2Tr (ρ(θ1, γ)ρcirc(θ2, ϕγ′)) ∂∆θ2 = 2e...

  2. [2]

    ∂ρθ ∂θ 2# + 2Tr ρθ ∂2ρθ ∂θ2 = 0. (A.3) We use (A.3) to define a new quantityQ HS (ρθ) as: QHS (ρθ) :=−lim θ′→θ ∂2 ∂θ2 Tr (ρθ′ρθ) =−Tr ρθ ∂2ρθ ∂θ2 (A.4) = Tr

    Quasi-Normalized Case Unlike the Uhlmann Fidelity, the Hilbert Schmidt In- ner Product is bounded by the purity Tr(ρ 2) of the state in question. We can normalize the Hilbert Schmidt Inner Product Tr (ρ(θ1, γ)ρcirc(θ2, ϕγ′)) by: Tr (ρ(θ1, γ)ρcirc(θ2, ϕγ′))→ Tr (ρ(θ1, γ)ρcirc(θ2, ϕγ′))p Tr(ρ(θ1, γ)2)Tr(ρcirc(θ2, ϕγ′)2) . (VI.4) However, we do not knowρ(θ 1...

    work page arXiv

  3. [3]

    S. Zhou, M. Zhang, J. Preskill, and L. Jiang, Nature com- munications9, 78 (2018), URLhttps://www.nature. com/articles/s41467-017-02510-3

    2018

  4. [4]

    T. K. Le, H. Q. Nguyen, and L. B. Ho, Scientific Reports 13, 17775 (2023)

    2023

  5. [5]

    J. J. Meyer, J. Borregaard, and J. Eisert, npj Quantum Information7, 89 (2021)

    2021

  6. [6]

    Koczor, S

    B. Koczor, S. Endo, T. Jones, Y. Matsuzaki, and S. C. Benjamin, New Journal of Physics22, 083038 (2020)

    2020

  7. [7]

    R. R. Allen, F. Machado, I. L. Chuang, H.-Y. Huang, and S. Choi,Quantum computing enhanced sensing(2025), 2501.07625, URLhttps://arxiv.org/abs/2501.07625

    work page arXiv 2025

  8. [8]

    C. O. Marrero, R. J. Tang, and N. Wiebe,Encoded quantum signal processing for heisenberg-limited metrol- ogy(2026), 2603.22798, URLhttps://arxiv.org/abs/ 2603.22798

    work page arXiv 2026

  9. [9]

    H. Xu, T. Xiao, J. Huang, J. Fan, and G. Zeng,Learning to restore heisenberg limit in noisy quantum sensing via quantum digital twin(2025), 2508.11198, URLhttps: //arxiv.org/abs/2508.11198

    work page arXiv 2025

  10. [10]

    H. Kwon, U. R. Fischer, S.-W. Lee, and L. Jiang,Restor- ing heisenberg scaling in time via autonomous quan- tum error correction(2026), 2504.13168, URLhttps: //arxiv.org/abs/2504.13168. 12

    work page arXiv 2026

  11. [11]

    Larocca, S

    M. Larocca, S. Thanasilp, S. Wang, K. Sharma, J. Bia- monte, P. J. Coles, L. Cincio, J. R. McClean, Z. Holmes, and M. Cerezo, Nature Reviews Physics7, 174–189 (2025), ISSN 2522-5820, URLhttp://dx.doi.org/10. 1038/s42254-025-00813-9

    2025

  12. [12]

    Wang and P

    Y. Wang and P. S. Krstic, Physical Review A102 (2020), ISSN 2469-9934, URLhttp://dx.doi.org/10. 1103/PhysRevA.102.042609

    2020

  13. [13]

    E. T. Escudero, D. A. Alamo, O. M. G´ omez, and P. G. Bringas,Assessing the Impact of Noise on Quantum Neural Networks: An Experimental Anal- ysis(Springer Nature Switzerland, 2023), p. 314–325, ISBN 9783031407253, URLhttp://dx.doi.org/10. 1007/978-3-031-40725-3_27

    2023

  14. [14]

    Banchi and G

    L. Banchi and G. E. Crooks, Quantum5, 386 (2021), ISSN 2521-327X, URLhttp://dx.doi.org/10.22331/ q-2021-01-25-386

    2021

  15. [15]

    X. Wang, C. Liu, X. Lin, Y. Tian, Y. Li, X. Nie, Y. Feng, Y. Zheng, Y. Dong, X. Wang, et al.,Noise-resilient quantum metrology with quantum computing(2025), 2509.00771, URLhttps://arxiv.org/abs/2509.00771

    work page arXiv 2025

  16. [16]

    S. L. Braunstein and C. M. Caves, Phys. Rev. Lett. 72, 3439 (1994), URLhttps://link.aps.org/doi/10. 1103/PhysRevLett.72.3439

    1994

  17. [17]

    M. G. Paris, International Journal of Quantum Informa- tion7, 125 (2009)

    2009

  18. [18]

    Manzano, Aip advances10(2020)

    D. Manzano, Aip advances10(2020)

    2020

  19. [19]

    D. P. Kingma and J. Ba,Adam: A method for stochastic optimization(2017), 1412.6980, URLhttps://arxiv. org/abs/1412.6980

    work page internal anchor Pith review arXiv 2017

  20. [20]

    Novak and N

    O. Novak and N. Rengaswamy, IEEE Transactions on Quantum Engineering7, 1 (2026)

    2026

  21. [21]

    Fishman, S

    M. Fishman, S. White, and E. M. Stoudenmire, SciPost Physics Codebases p. 004 (2022)

    2022

  22. [22]

    Speed Optimized Two-Qubit Gates with Laser Coherent Control Techniques for Ion Trap Quan- tum Computing

    G. Vidal, Phys. Rev. Lett.91, 147902 (2003), URL https://link.aps.org/doi/10.1103/PhysRevLett.91. 147902

    work page doi:10.1103/physrevlett.91 2003

  23. [23]

    Haegeman, J

    J. Haegeman, J. I. Cirac, T. J. Osborne, I. Piˇ zorn, H. Verschelde, and F. Verstraete, Phys. Rev. Lett. 107, 070601 (2011), URLhttps://link.aps.org/doi/ 10.1103/PhysRevLett.107.070601

    work page doi:10.1103/physrevlett.107.070601 2011

  24. [24]

    Westhoff, M

    P. Westhoff, M. Moroder, U. Schollw¨ ock, and S. Paeckel, A tensor network framework for lindbladian spectra and steady states(2025), 2509.07709, URLhttps://arxiv. org/abs/2509.07709

    work page arXiv 2025

  25. [25]

    Quantum fingerprinting

    H. Buhrman, R. Cleve, J. Watrous, and R. de Wolf, Phys. Rev. Lett.87, 167902 (2001), URLhttps://link.aps. org/doi/10.1103/PhysRevLett.87.167902

    work page doi:10.1103/physrevlett.87.167902 2001

  26. [26]

    N. F. Ramsey, Physical Review78, 695 (1950)

    1950

  27. [27]

    Lambert, E

    N. Lambert, E. Gigu‘ere, P. Menczel, B. Li, P. Hopf, G. Su’arez, M. Gali, J. Lishman, R. Gadhvi, R. Agar- wal, et al., Physics Reports1153, 1 (2026), ISSN 0370- 1573, URLhttps://www.sciencedirect.com/science/ article/pii/S0370157325002704

    2026

  28. [28]

    Ouyang and N

    Y. Ouyang and N. Rengaswamy, Phys. Rev. A107, 022620 (2023), URLhttps://link.aps.org/doi/10. 1103/PhysRevA.107.022620

    2023

  29. [29]

    Quantum Algorithms for Nearest-Neighbor Methods for Supervised and Unsupervised Learning

    N. Wiebe, A. Kapoor, and K. Svore, arXiv preprint arXiv:1401.2142 (2014)

    work page Pith review arXiv 2014

  30. [30]

    Fanizza, M

    M. Fanizza, M. Rosati, M. Skotiniotis, J. Cal- samiglia, and V. Giovannetti, Phys. Rev. Lett.124, 060503 (2020), URLhttps://link.aps.org/doi/10. 1103/PhysRevLett.124.060503

    2020

  31. [31]

    Fang and Q

    W. Fang and Q. Wang, in33rd Annual European Sym- posium on Algorithms (ESA 2025), edited by A. Benoit, H. Kaplan, S. Wild, and G. Herman (Schloss Dagstuhl – Leibniz-Zentrum f¨ ur Informatik, Dagstuhl, Germany, 2025), vol. 351 ofLeibniz International Proceedings in Informatics (LIPIcs), pp. 4:1–4:12, ISBN 978-3-95977- 395-9, ISSN 1868-8969, URLhttps://drop...

    work page doi:10.4230/lipics.esa.2025.4 2025

  32. [32]

    Zhang and Q

    Z. Zhang and Q. Zhuang, Quantum Science & Technology 6, 043001 (2021)

    2021