pith. sign in

arxiv: 2604.13177 · v1 · submitted 2026-04-14 · 🪐 quant-ph

Quantum computational displacement sensing

Sridhar Prabhu , Saeed A. Khan , Xingrui Song , Mathieu Ouellet , Ryotatsu Yanagimoto , Saswata Roy , Alen Senanian , Logan G. Wright
show 2 more authors
Valla Fatemi Peter L. McMahon
This is my paper

Pith reviewed 2026-05-10 15:03 UTC · model grok-4.3

classification 🪐 quant-ph
keywords quantum computational sensingdisplacement sensingparameterized quantum circuitssuperconducting circuitsquantum sensingbinary classificationoscillator sensing
0
0 comments X

The pith

Parameterized quantum circuits around a displacement sensor classify its class label more accurately than estimating the displacement first and processing classically.

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

The paper shows that embedding quantum computation directly into the sensing process can outperform the usual pipeline of first recovering the full physical signal and then applying classical algorithms. In the demonstrated case, an oscillator senses a single complex displacement once; parameterized circuits before and after the sensing step evolve the joint qubit-oscillator state so that a final qubit measurement directly yields the binary class prediction. Experiments on a superconducting device with circuits containing up to 24 entangling gates and 38 parameters, trained entirely in simulation, produce higher accuracy than several conventional estimation-plus-postprocessing baselines, with the gap reaching 15 percentage points on some tasks. The work establishes that quantum computational sensing is feasible on present-day noisy hardware when the goal is to extract a task-specific property rather than the raw signal.

Core claim

Quantum computational displacement sensing uses parameterized quantum circuits placed before and after a physical displacement acting on an oscillator; the circuits are trained to map the unknown displacement directly onto the ground or excited state of a coupled qubit, so that a single qubit measurement yields the binary class label without ever estimating the displacement value itself.

What carries the argument

Parameterized quantum circuits inserted before and after the displacement sensing operation, which encode the classification objective into the unitary evolution so the qubit readout directly outputs the predicted label.

If this is right

  • Deeper circuits with more parameters systematically increase expressivity and classification accuracy for the same sensing task.
  • When only a function or property of the sensed signal is required, the quantum-computational route avoids the overhead of full signal reconstruction.
  • The same pre- and post-sensing circuit structure can be applied to other physical parameters beyond displacement.
  • Task-specific training of the circuits can be performed in simulation and then transferred to hardware for immediate experimental use.

Where Pith is reading between the lines

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

  • If the observed advantage persists at larger circuit sizes, quantum computational sensing may become preferable for any sensing task whose output is a low-dimensional label or decision rather than a high-dimensional waveform.
  • The approach reframes quantum sensors as programmable processors whose native operation is part of an end-to-end computation rather than a standalone metrology device.
  • Similar pre- and post-processing circuits could be tested on other hardware platforms to determine how platform-specific noise affects the transfer of in-silico-trained models.

Load-bearing premise

Circuits optimized entirely in noiseless simulation will continue to deliver their reported accuracy advantage when run on real superconducting hardware despite decoherence, calibration drift, and simulation-reality mismatch.

What would settle it

An experiment in which the quantum computational protocol's classification accuracy falls to or below the best conventional estimation-plus-classical-postprocessing accuracy on the same displacement-classification tasks.

Figures

Figures reproduced from arXiv: 2604.13177 by Alen Senanian, Logan G. Wright, Mathieu Ouellet, Peter L. McMahon, Ryotatsu Yanagimoto, Saeed A. Khan, Saswata Roy, Sridhar Prabhu, Valla Fatemi, Xingrui Song.

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: b, we plot the classification accuracy during the training phase of the simulation, evaluated on the test dataset, for a protocol of depth N = 10. The performance steadily improves and converges within a 1000 epochs. We then evaluate the performance of the trained protocol in experiment. Fig. 2c illustrates the experimentally-measured qubit- [PITH_FULL_IMAGE:figures/full_fig_p004_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_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8 [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9 [PITH_FULL_IMAGE:figures/full_fig_p021_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10 [PITH_FULL_IMAGE:figures/full_fig_p024_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11 [PITH_FULL_IMAGE:figures/full_fig_p025_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12 [PITH_FULL_IMAGE:figures/full_fig_p026_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13 [PITH_FULL_IMAGE:figures/full_fig_p027_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14 [PITH_FULL_IMAGE:figures/full_fig_p028_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: FIG. 15 [PITH_FULL_IMAGE:figures/full_fig_p029_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: FIG. 16 [PITH_FULL_IMAGE:figures/full_fig_p030_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: FIG. 17 [PITH_FULL_IMAGE:figures/full_fig_p031_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: FIG. 18 [PITH_FULL_IMAGE:figures/full_fig_p039_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: FIG. 19 [PITH_FULL_IMAGE:figures/full_fig_p042_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: FIG. 20 [PITH_FULL_IMAGE:figures/full_fig_p043_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: FIG. 21 [PITH_FULL_IMAGE:figures/full_fig_p044_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: FIG. 22 [PITH_FULL_IMAGE:figures/full_fig_p045_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: FIG. 23 [PITH_FULL_IMAGE:figures/full_fig_p046_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: FIG. 24 [PITH_FULL_IMAGE:figures/full_fig_p048_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: FIG. 25 [PITH_FULL_IMAGE:figures/full_fig_p050_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: FIG. 26 [PITH_FULL_IMAGE:figures/full_fig_p051_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: FIG. 27 [PITH_FULL_IMAGE:figures/full_fig_p052_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: FIG. 28 [PITH_FULL_IMAGE:figures/full_fig_p053_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: FIG. 29 [PITH_FULL_IMAGE:figures/full_fig_p053_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: FIG. 30 [PITH_FULL_IMAGE:figures/full_fig_p055_30.png] view at source ↗
Figure 31
Figure 31. Figure 31: FIG. 31 [PITH_FULL_IMAGE:figures/full_fig_p056_31.png] view at source ↗
Figure 32
Figure 32. Figure 32: FIG. 32 [PITH_FULL_IMAGE:figures/full_fig_p059_32.png] view at source ↗
Figure 33
Figure 33. Figure 33: FIG. 33 [PITH_FULL_IMAGE:figures/full_fig_p060_33.png] view at source ↗
read the original abstract

Quantum computational sensing (QCS) combines quantum sensing with quantum computing to extract task-relevant information from the physical world. QCS can in principle achieve an accuracy advantage for specific tasks versus the alternative of raw-signal estimation using conventional quantum sensing followed by task-specific classical postprocessing. Here we report the experimental demonstration of quantum computational displacement sensing (QCDS) with a superconducting circuit comprising a qubit coupled to an oscillator. We consider binary classification sensing tasks, where the goal is to predict the class label of a single complex-valued displacement sensed once by the oscillator. Rather than estimating the displacement, our computational-sensing protocol -- using parameterized quantum circuits before and after sensing -- attempts to determine the binary class label using quantum processing and map it onto the ground or excited state of the qubit. A single measurement of the qubit directly outputs the prediction. We implemented circuits with up to 24 entangling gates and 38 free parameters, which were trained in silico. We show that increasing the circuit depth systematically improves expressivity and classification accuracy. We experimentally obtained an accuracy advantage over a suite of protocols that first use conventional quantum sensing to estimate the displacement before using classical postprocessing to perform prediction. For certain tasks, our protocol achieves a 15-percentage-points higher classification accuracy than the best conventional approach considered. Our results establish the feasibility of quantum computational sensing with noisy superconducting hardware and illustrate how integrating quantum computation with quantum sensing can enhance performance when the goal is to estimate a property or function of a signal rather than to estimate the signal.

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 / 1 minor

Summary. The manuscript reports an experimental realization of quantum computational displacement sensing (QCDS) on a superconducting qubit-oscillator system. Parameterized quantum circuits (up to 24 entangling gates and 38 free parameters, trained entirely in classical simulation) are inserted before and after the displacement sensing operation to perform direct binary classification of the displacement via a single qubit measurement. The central experimental claim is that this integrated quantum-computational protocol achieves up to 15 percentage points higher classification accuracy than conventional quantum sensing followed by classical post-processing pipelines, with accuracy improving systematically as circuit depth increases. The work positions this as a demonstration of the feasibility of quantum computational sensing on noisy hardware.

Significance. If the reported accuracy advantage is statistically robust, this would constitute a meaningful experimental step toward practical quantum computational sensing. It supplies concrete hardware evidence that task-specific quantum processing can outperform the conventional pipeline of full signal estimation plus classical post-processing, at least for selected binary classification tasks. The systematic depth-scaling study and the direct hardware implementation are strengths that, if properly validated, support the broader paradigm of integrating quantum computation with quantum sensing.

major comments (2)
  1. [Experimental Results] Experimental Results section: The headline claim that the protocol 'achieves a 15-percentage-points higher classification accuracy than the best conventional approach considered' is presented without error bars, confidence intervals, the number of experimental shots or repetitions, or any description of the statistical procedure used to obtain the figure. This information is load-bearing for the central experimental result and is required to assess whether the advantage is statistically significant or potentially affected by post-hoc task selection.
  2. [Circuit Training and Hardware Implementation] Circuit Training and Hardware Implementation section: Circuits with up to 38 parameters and 24 entangling gates are trained exclusively in silico, yet the manuscript contains no quantitative simulated-versus-experimental accuracy comparison for the same circuits. Without this comparison, it is impossible to confirm that the claimed 15 pp advantage survives the accumulated decoherence, gate infidelity, and model mismatch expected on the superconducting device.
minor comments (1)
  1. [Methods] Figure captions and methods: The description of the parameterized circuit ansatz would benefit from an explicit statement of the gate set and the precise parameterization (e.g., which angles are trainable) to facilitate independent reproduction.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments. We appreciate the recognition of the experimental demonstration and the systematic depth study. We address each major comment below and will revise the manuscript to incorporate the requested clarifications and additions.

read point-by-point responses

  1. Referee: [Experimental Results] Experimental Results section: The headline claim that the protocol 'achieves a 15-percentage-points higher classification accuracy than the best conventional approach considered' is presented without error bars, confidence intervals, the number of experimental shots or repetitions, or any description of the statistical procedure used to obtain the figure. This information is load-bearing for the central experimental result and is required to assess whether the advantage is statistically significant or potentially affected by post-hoc task selection.

    Authors: We agree that these statistical details are essential for assessing the robustness of the central claim. In the revised manuscript we will add error bars to the reported accuracies, computed via binomial statistics from the finite shot counts. We will explicitly state the number of shots per circuit (typically several thousand) and the number of independent hardware repetitions. We will also describe the statistical procedure used and confirm that the observed advantage remains statistically significant. The binary classification tasks were defined a priori from the sensing scenario and were not selected post-hoc. revision: yes

  2. Referee: [Circuit Training and Hardware Implementation] Circuit Training and Hardware Implementation section: Circuits with up to 38 parameters and 24 entangling gates are trained exclusively in silico, yet the manuscript contains no quantitative simulated-versus-experimental accuracy comparison for the same circuits. Without this comparison, it is impossible to confirm that the claimed 15 pp advantage survives the accumulated decoherence, gate infidelity, and model mismatch expected on the superconducting device.

    Authors: We acknowledge this gap in the original presentation. Although the circuits were trained in simulation, we possess the corresponding simulated accuracies for the trained parameter sets. In the revision we will add a direct quantitative comparison (in the main text or as a supplementary figure) of simulated versus experimental classification accuracies for the same circuits. This will quantify the degradation due to hardware noise while showing that the advantage over conventional sensing-plus-postprocessing pipelines is retained on the device. revision: yes

Circularity Check

0 steps flagged

No circularity; experimental results are directly measured

full rationale

The manuscript describes an experimental protocol for quantum computational displacement sensing using a superconducting circuit. Parameterized circuits with up to 38 parameters are trained in silico and then executed on hardware to perform binary classification of displacements. The accuracy advantage is reported from direct experimental measurements comparing the quantum protocol to conventional estimation followed by classical postprocessing. There is no derivation chain involving predictions or first-principles results that reduce to the inputs by construction. No self-citations are load-bearing for the central claim, and the work relies on empirical data rather than theoretical self-consistency.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The protocol relies on standard superconducting circuit quantum mechanics and in-silico training of circuit parameters; no new physical entities are introduced.

free parameters (1)
  • circuit parameters = up to 38
    Up to 38 adjustable parameters in the parameterized quantum circuits, optimized via in-silico training to map sensed displacements to binary labels.
axioms (2)
  • standard math Standard quantum mechanics governs the qubit-oscillator system and gate operations.
    Invoked throughout the description of the sensing and circuit implementation.
  • domain assumption In-silico training accurately models the hardware noise and dynamics for parameter optimization.
    Required for the transfer of trained circuits to physical hardware.

pith-pipeline@v0.9.0 · 5604 in / 1379 out tokens · 51523 ms · 2026-05-10T15:03:00.799190+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

37 extracted references · 37 canonical work pages

  1. [1]

    We define the qubit drive waveform as: Ω(t) = Aπ 2 eiϕhΩ(t),(D2) whereAdefines the amplitude of the pulse,ϕits phase, andh Ω(t) its temporal waveform (see Appendix C 1)

    Arbitrary qubit rotations When performing arbitrary qubit rotations, we consider the cavity drive signals to be turned off (ε= 0). We define the qubit drive waveform as: Ω(t) = Aπ 2 eiϕhΩ(t),(D2) whereAdefines the amplitude of the pulse,ϕits phase, andh Ω(t) its temporal waveform (see Appendix C 1). For a pulse of lengtht p, the pulse waveform is normaliz...

    work page

  2. [2]

    The first operation applies a coherent driveε(t) to the oscillator, whose action can be denoted as a unitary operationU ε

    Echoed conditional displacement (ECD) gate The ECD gate is composed of three elementary operations. The first operation applies a coherent driveε(t) to the oscillator, whose action can be denoted as a unitary operationU ε. This is followed by aπ-pulse on the qubit defined asR π ≡R(θ=π, ϕ= 0) (see Appendix D 1). Lastly, we apply a second coherent driveε ′(...

    work page

  3. [3]

    Fromt 0 tot − m a storage driveε(t) is applied

    work page

  4. [4]

    Fromt − m tot + m a qubitπ-pulse is applied

    work page

  5. [5]

    We typically considerε ′(t) =−ε(t)

    Finally, fromt + m tot f a second storage driveε ′(t) is applied. We typically considerε ′(t) =−ε(t). For the first storage cavity drive, the evolution is given by Eq. (D19), reproduced below: |ψ(t− m)⟩=U ε|ψ(t0)⟩=−i h ˆD(αg(t− m))e−iτ χ 2 ˆa†ˆae−iϕg ˆD†(αg(t0))|g⟩⟨g|+ ˆD(αe(t− m))e+iτ χ 2 ˆa†ˆae−iϕe ˆD†(αe(t0))|e⟩⟨e| i |ψ(t0)⟩. (D24) Upon completion of t...

    work page

  6. [6]

    (D33a), (D33b) and Eqs

    We therefore obtain analytic forms of the accrued drive-induced phases: ϕg(t) = |ε| 2 αg(0)I ∗ 1(t)e−iϕε +α ∗ g(0)I1(t)e+iϕε − i 2 |ε|2 (I2(t)− I ∗ 2(t)) (D37a) ϕe(t) = |ε| 2 αe(0)I1(t)e−iϕε +α ∗ e(0)I ∗ 1(t)e+iϕε − i 2 |ε|2 (I ∗ 2(t)− I 2(t)) (D37b) Eqs. (D33a), (D33b) and Eqs. (D37a), (D37b) provide analytic solutions for the drive-induced displacements...

    work page

  7. [7]

    Due to the analog nature of the pulse, this operation imparts a small conditional displacement, orthogonal to the direction of the unconditional displacement

    Sensing operation We implement the sensing displacement using a pulseε s(t), of the pulse shape used in the calibration of the AWG amplitude to displacement (see Appendix C 1), with the amplitude and phase determined by the dataset. Due to the analog nature of the pulse, this operation imparts a small conditional displacement, orthogonal to the direction ...

    work page 2048

  8. [8]

    Overview In this Appendix section we provide details of various displacement sensing baselines considered. Table III sum- marizes the benefits and drawbacks of these displacement sensing baselines, along with our QCDS protocol, when used for the task of binary classification of displacements. When the quantum computational displacement sensor achieves a b...

    work page

  9. [9]

    Directly outputs the class prediction

    work page

  10. [10]

    Performance increases with circuit depthN

    work page

  11. [11]

    Quantum circuit requires to be trained for each task

    work page

  12. [12]

    Sensitive to small displacements

    Performance limited by qubit-readout fidelity Cat-state sensor 1. Sensitive to small displacements

    work page

  13. [13]

    Simple to implement in experiment

    work page

  14. [15]

    Sensitivity – dynamic range tradeoff

    work page

  15. [16]

    Performance limited by qubit-readout fidelity Compass-state sensor

    work page

  16. [19]

    Limited expressive capacity

    work page

  17. [20]

    Performance limited by qubit-readout fidelity Phase-preserving amplifier and classical postprocessing backend

    work page

  18. [23]

    Limited by the fundamental noise introduced by the amplifier

    work page

  19. [24]

    Classical postprocessing required to be trained for each task Squeezed probe-state, phase-sensitive amplifier and classical postprocessing backend

    work page

  20. [25]

    Sensitive to small displacements

    work page

  21. [26]

    High expressive capacity of the single-component displacement

    work page

  22. [27]

    Sensitive to single-component of displacement

    work page

  23. [28]

    Classical postprocessing required to be trained for each task GKP probe-state, quantum phase estimation and classical postprocessing backend

    work page

  24. [30]

    Sensitive to both components of displacement

    work page

  25. [31]

    Limited dynamic range

    work page

  26. [32]

    Classical postprocessing required to be trained for each task Two-mode squeezed probe-state, phase-sensitive amplifiers and classical postprocessing backend

    work page

  27. [33]

    High expressive capacity

    work page

  28. [34]

    Performance increases with two-mode squeezing

    work page

  29. [35]

    Hard to implement in experiment with high efficiency

    work page

  30. [36]

    For clarity, we do not include the axis labels on the 2D plots of the function of the qubit-excitation probability

    Classical postprocessing required to be trained for each task TABLE III.Benefits and drawbacks of different protocols for classifying displacements.The first row is the quantum computational displacement sensor introduced in this work, while the rest are examples of conventional displacement sensing protocols In Fig, 19, we plot the response of various pr...

    work page

  31. [37]

    Quantum computational displacement sensor In this subsection, we discuss the implementation of our quantum computational sensor for this task. In Fig. 20a, we plot the experimental classification accuracy of the protocols as a function of the circuit depthN. The best performance, acrossN, is plotted in Fig. 4. For our experiment, this is usually theN= 12 ...

    work page

  32. [38]

    This protocol can be realized for aN= 1 version of the QCDS protocol

    Cat-state sensor The cat-state sensing protocol measures a single component of the sensed displacement by mapping it onto the qubit phase [10]. This protocol can be realized for aN= 1 version of the QCDS protocol. In general, this proceeds by preparing the qubit in the|+⟩state. Then, an echoed conditional displacement gate entangles the qubit with the cav...

    work page

  33. [39]

    The compass-state sensing protocol relies on the fact that a compass state that is displaced byD(α) is quasiorthogonal to the undisplaced compass state on a scale set by 1/|α|[31]

    Compass-state sensor Another approach for conventional quantum sensing of displacements is using compass states [31], which belong to the family of generalized Schr¨ odinger cat-states. The compass-state sensing protocol relies on the fact that a compass state that is displaced byD(α) is quasiorthogonal to the undisplaced compass state on a scale set by 1...

    work page

  34. [40]

    The advantage of this technique is its simplicity: based on the estimate, further postprocessing tasks can be performed, as illustrated in Fig

    Phase-preserving amplifier and classical postprocessing backend A conventional sensing strategy is to perform heterodyne measurement of the oscillator, which provides unbiased estimates of its position and momentum (eαx,eαp) in a single shot [18]. The advantage of this technique is its simplicity: based on the estimate, further postprocessing tasks can be...

    work page

  35. [41]

    Squeezed probe-state, phase-sensitive amplifier and classical postprocessing backend Another conventional sensing strategy which we use as a benchmark is one which only measures one component of the displacement. This protocol was implemented in the second phase of the HAYSTAC experiment, which provided a sensing advantage since the axis of the displaceme...

    work page

  36. [42]

    In this subsection, we simulate the performance of a quantum sensing protocol based on using the GKP state

    Gottesman–Kitaev–Preskill (GKP) probe-state, quantum phase estimation and classical postprocessing backend A protocol can estimate both position and momentum beyond the limit set by the uncertainty principle by instead measuring their modular counterparts. In this subsection, we simulate the performance of a quantum sensing protocol based on using the GKP...

    work page

  37. [43]

    multi-mode conditional displacement gate

    Two-mode squeezed probe state, phase-sensitive amplifiers and classical postprocessing backend Two-mode squeezing (TMS) provides a resource to enable sub-vacuum-noise estimation of non-commuting quadra- tures of a sensed displacement [39, 40]. In this section we provide theoretical details of this TMS scheme for displace- ment sensing, and of the classica...

    work page