Pith. sign in

REVIEW 4 major objections 5 minor 24 references

Guess, SWAP, Repeat : Capturing Quantum Snapshots in Classical Memory

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper argues that an unknown pure quantum state can be reconstructed, stored classically, and recreated on demand through a guess-and-check loop whose only feedback is a SWAP-test fidelity number, with no direct measurement of the…

desk verdict A plausibly useful tomography variant buried under a non-destructive-snapshot claim that the protocol itself contradicts. read the letter →

arxiv 2504.14459 v1 pith:UJZ5IFEV submitted 2025-04-20 quant-ph cs.LG

Debarshi Kundu , Avimita Chatterjee , Swaroop Ghosh This is my paper · ORCID
classification quant-phcs.LG MSC 81P6881P45 PACS 03.67.-a
keywords quantumsnapshotSWAPtestfidelityestimationevolutionarystrategystatereconstructionmemoryQRAMNISQhardware
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that an unknown pure quantum state can be learned, and later recreated, without ever measuring the state directly. The method runs a guess-and-check loop: a classical generator proposes a candidate state, a SWAP test compares the candidate against a freshly re-prepared copy of the unknown state and returns a single fidelity number, and that number is the only feedback the generator receives as it refines its guess. Once the fidelity approaches 1, the winning candidate is stored as a classical vector and can be reloaded into any compatible circuit, which the authors present as a working form of classical quantum memory useful for debugging, circuit introspection, and QRAM-style reuse. In noiseless simulation the loop reaches average fidelity 0.999 over 100 random states, and on real superconducting qubit hardware it reconstructs single-qubit states at near-unit fidelity. The authors themselves delimit the claim: the method requires many re-preparations of the unknown state and does not extend to mixed states, because the SWAP test returns the Hilbert-Schmidt inner product rather than true fidelity.

What carries the argument

The load-bearing mechanism is the SWAP test used as a non-destructive fidelity oracle. A candidate state and a re-prepared copy of the unknown state are loaded side by side with an ancilla qubit, and the expected value $\langle Z\rangle$ on the ancilla equals the squared overlap $|\langle\psi|\varphi\rangle|^2$, so a single measurement statistic feeds the entire learning loop without ever measuring either state directly. The second mechanism is a classical generator whose output is a complex vector, normalized to unit norm and converted into a quantum circuit through a uniformly-controlled-rotation state-preparation routine, so that each classical guess can be physically instantiated and compared. The third is the evolutionary update rule of QESwap, $w \leftarrow w + \frac{\alpha}{N\sigma}\sum_i A_i z_i$, in which Gaussian noise vectors $z_i$ generate a population of candidates, the SWAP test scores each one, and standardized advantages $A_i$ reweight the update; this derivative-free loop is what keeps working under noise and on hardware.

What would settle it

Prepare an uncharacterized state that has no preparation circuit, run the loop once, and ask whether any stored vector reproduces it; it cannot, because the SWAP test needs freshly re-prepared copies every round, so no fidelity signal ever exists. A quantitative version: reconstruct a two-qubit Bell state on real superconducting hardware using fidelity-only feedback and compare the result with the claimed 0.99 threshold, since the paper's hardware evidence covers only single-qubit states and its own noisy simulations show fidelity slipping as the qubit count rises.

Watch

Extended reading notes

Core claim

On its own terms, the paper's claim is that non-destructive observation of a quantum state reduces to maximizing a black-box fidelity function. The SWAP test provides the oracle: for two states on $n$ qubits, an ancilla-based comparison yields $\langle Z\rangle = |\langle\psi|\varphi\rangle|^2$, so the expected Pauli-$Z$ value on the ancilla is exactly the overlap between the candidate and the unknown state. Feeding that single number back into a generator turns reconstruction into an optimization problem in which neither state is ever measured directly. The paper demonstrates two optimizers: a deep neural network trained with finite-difference gradients through the non-differentiable SWAP test, and a gradient-free evolutionary strategy (QESwap) that perturbs a population of candidate state vectors and updates them with standardized fidelity rewards. Both reach fidelity above 0.99 within roughly 5 to 17 epochs in noiseless simulation; under noise the evolutionary strategy is the more stable of the two, and it is the variant demonstrated on real hardware, converging on single-qubit known states to near-unit fidelity within three epochs. The paper also draws its own boundary: density-matrix reconstruction of mixed states plateaus near 0.8 fidelity because the SWAP test's Hilbert-Schmidt inner product overestimates mixed-state similarity, so pure states are the intended scope.

Load-bearing premise

The loop assumes the unknown state can be prepared over and over again from a known procedure; if only one copy exists, there is nothing to SWAP-test against, and no snapshot is possible.

Editorial extensions

If this is right

  • Intermediate quantum states in a circuit can be captured one at a time, stored as classical vectors, and re-prepared later, giving near-term devices a form of non-volatile memory bank without persistent physical qubits.
  • Because the stored state is a plain classical vector, it can be retrieved by standard QRAM access schemes and fed into downstream circuits, enabling modular program design where whole subcircuits need not be re-executed.
  • The fidelity-only feedback signal means no training dataset and no access to the unknown density matrix are required, so the loop can run directly against noisy hardware; the paper reports near-unit hardware fidelity for single-qubit states.
  • Mixed states are outside the method's scope: the SWAP-test signal saturates at a success probability of 3/4 for distinguishing non-orthogonal mixed states, so the framework is limited to pure-state snapshots.
  • The gradient-free evolutionary strategy is the recommended configuration, reaching fidelity above 0.99 in noisy simulation where the gradient-based network stalls or fails.

Reading between the lines

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

  • The snapshot is a batch property, not a single-instance capture: the loop re-prepares the unknown state many times, so a state that appears only once in a circuit cannot be snapshotted, which quietly limits the debugging story to states that can be regenerated from a known preparation procedure.
  • What is stored is ultimately a preparation recipe rather than a quantum memory in the information-theoretic sense; the classical vector is only as good as the ability to re-prepare the state it describes, an ability the method assumes but does not certify.
  • A natural testable extension is to replace the SWAP-test oracle with an overlap estimator that is well-defined for mixed states, such as randomized-measurement or shadow-based fidelity; the paper's own analysis pinpoints the Hilbert-Schmidt inner product as the blocker, so the fix is localized.
  • If the same fidelity-only loop scales to multi-qubit entangled states on real hardware, the claimed memory architecture becomes substantially more credible, but the paper's hardware evidence covers only three single-qubit states, so scaling is the open risk.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript proposes a machine-learning-driven method, named QESwap, for reconstructing unknown pure quantum states using only SWAP-test-based fidelity feedback. The authors combine a deep-neural-network generator and an evolutionary-strategy optimizer with iterative SWAP tests, and report near-unit fidelities in noiseless and noisy simulations for up to six qubits, as well as single-qubit demonstrations on IBM hardware. The central claim is that this enables non-destructive 'quantum snapshots' of states at arbitrary points in a circuit, with classical storage and later reuse, thereby functioning as a form of quantum memory. The paper also provides comparative benchmarks, an entanglement-entropy analysis of reconstructed states, and a detailed gate-level noise model.

Significance. If the central claim were correct, the ability to observe a quantum state non-destructively, store it classically, and reuse it would be a major advance with implications for quantum debugging, QRAM, and modular circuit design. The paper has some genuine strengths: it openly discusses the mixed-state limitation of the SWAP test, provides a realistic noise model, and reports reproducible-looking simulation benchmarks over many random states. However, the advertised capability is not supported by the protocol as written. The method repeatedly re-prepares the unknown state and consumes copies at every fidelity evaluation, so it is a copy-consuming, tomography-like estimation scheme rather than a single-copy non-destructive snapshot. The no-cloning theorem makes the advertised general capability impossible, and this is not a peripheral caveat but a direct contradiction of the paper's main contribution. Consequently, the significance of the reported numerical results is confined to a classical optimization loop over many identical copies, which is a much weaker and less novel claim.

major comments (4)
  1. [Section III-A and III-B2(c)] The protocol contradicts the paper's central claim of non-destructive observation. Section III-A states that the authors 'repeatedly reinitialize the unknown quantum state, thereby requiring multiple copies, an assumption aligned with the principles of quantum state tomography,' and Section III-B2(c) says each fidelity evaluation uses 'a re-prepared instance of the unknown target state.' The SWAP test itself disturbs or consumes the copy used for the overlap estimate. Therefore the method does not observe a state without destroying it; it estimates a state from many fresh copies, exactly like quantum state tomography. This invalidates the headline claims of 'non-destructive snapshot' and 'single-copy, non-destructive quantum learning' made in Sections I and III-C.
  2. [Algorithm 2 and Section III-C] The copy consumption is quantitatively severe. Algorithm 2 evaluates a population of N=50 candidates per iteration, and each evaluation requires a separate freshly prepared instance of the unknown state. With up to 100 iterations, a single reconstruction consumes thousands of copies of the target state. The reader should not mistake this for a resource-light observation: this is a classical optimization loop over multiple copies, and the claimed equivalence to a non-destructive 'classical quantum memory' is not established by the reported experiments.
  3. [Section III-D] The claim of observing states 'at different points within a single circuit, one at a time' is not supported. Section III-D acknowledges that 'states are re-prepared in each iteration.' A snapshot at a circuit point would require the state at that point to be preserved or at least accessible without re-preparation; instead, the method produces a statistical estimate from many independent runs. This does not enable mid-circuit introspection of a single evolving computation, and the QRAM, debugging, and modular-program applications in Section V-A therefore rest on an unsupported capability.
  4. [Section IV, Real Hardware Results] The hardware validation is limited to three known single-qubit states (|0>, |1>, and (|0>+|1>)/sqrt(2)) on ibm_sherbrooke. No unknown or hidden states are used, no multi-qubit states are tested, and the fidelity on hardware is not compared against any baseline. This is a useful sanity check for the simulation pipeline, but it does not support the paper's broad claims about capturing arbitrary quantum states or about scalability to the 6-qubit systems reported in simulation.
minor comments (5)
  1. [Section III-B1 and Algorithm 1] The notation is inconsistent: Section III-B1 defines the unknown target as |ψ> while Algorithm 1 and the surrounding text call it |ϕ_u>. Please pick one symbol and use it consistently.
  2. [Appendix A] The text says the bit-flip probability of 2.003e-04 represents a '0.1% chance of bit flips,' but 2.003e-04 is 0.02%. Please correct the percentage or the value.
  3. [Section V-B] The bound P_success ≤ 3/4 for SWAP-test-based discrimination of two non-orthogonal mixed states is stated without a proof or reference. Please provide a citation or a derivation.
  4. [Table III] Several table entries are marked 'NA' for noisy simulations (e.g., gradient-based unitary matrix, 2 and 3 qubits). The figure caption and text interpret these as cases that failed to converge, but the table itself does not say so explicitly. Please add a footnote defining 'NA' consistently.
  5. [Section IV, Benchmarking] The final fidelity for all simulated experiments is computed classically via the inner product with the known target state. This is fine as a validation step, but the text should be clearer that this evaluation uses the classically known target vector, not the SWAP-test estimate, so the reported fidelity is not the noisy hardware fidelity in general.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity found; reported fidelities are optimized objectives and the sole self-citation is peripheral.

full rationale

Walked the claimed derivation chain: the paper samples a target pure state (Section III-B3), generates candidate states from a classical NN/ES parameterization (Algorithms 1 and 2), estimates overlap with the target through the standard SWAP-test identity F = |⟨ψ|ϕu⟩|² = ⟨Z⟩ (Eq. 1), and then uses this same F as the training loss (L = 1−F) or ES reward. The final reported fidelities (0.999 simulation, ~1.0 hardware) are therefore the values of the optimization objective at convergence, not out-of-sample predictions. This is a validation weakness: the success metric and the learning signal coincide, so the numbers do not independently certify the advertised non-destructive snapshot capability. It is not, however, a circular derivation: the target states are externally sampled from the complex unit sphere (or known hardware-prepared states), the optimizer can and does fail in several reported configurations (e.g., NN noisy cases, mixed-state plateau ~0.8), and the reported numbers are empirical convergence results rather than quantities defined to equal their own inputs. The only self-citation is [14] (Phalak, Chatterjee, Ghosh) for QRAM retrieval algorithms; it appears in the QRAM application discussion and is not load-bearing for the reconstruction method. The paper also states the multiple-copy assumption ('We repeatedly reinitialize the unknown quantum state, thereby requiring multiple copies') and later confines claims to pure states because the SWAP test estimates the Hilbert-Schmidt inner product rather than fidelity for mixed states (Section V-B). These admissions are consistency limitations, not circular steps. Verdict: no significant circularity; score 2 reflects one minor non-load-bearing self-citation and the shared training/evaluation objective.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rides on the repeated-copies assumption (axiom 1) and the purity assumption (axiom 2). The free parameters are standard ML hyperparameters; their values are not derived from physics. No new physical entities are introduced.

free parameters (6)
  • ES population size N = 50
    Chosen by hand for QESwap; affects convergence and cost.
  • ES noise scale sigma = 0.1
    Chosen by hand in Algorithm 2; controls exploration.
  • ES learning rate alpha = 0.05
    Chosen by hand in Algorithm 2; controls update step.
  • NN learning rate = 1e-4
    Adam optimizer learning rate for the neural generator.
  • Manual gradient scaling_factor = not specified
    All gradients are manually scaled by a constant scaling_factor (Section III-B7); value not given.
  • Latent dimension 256 = 256
    Generator input dimension chosen without stated justification.
assumptions (5)
  • domain assumption Access to multiple identical copies (or re-preparation) of the unknown quantum state
    Section III-A: 'We repeatedly reinitialize the unknown quantum state, thereby requiring multiple copies.' Without this, the SWAP test cannot be iterated; this assumption contradicts the paper's non-destructive framing.
  • domain assumption Target states are pure
    Section V-B: SWAP test measures Hilbert-Schmidt inner product for mixed states, overestimating similarity and lacking operational meaning; method is inapplicable to mixed states.
  • domain assumption Known state preparation circuit for the target
    The 'unknown' states are generated by the authors via Mottonen state preparation (qml.StatePrep), meaning the experimenter knows the preparation procedure and can re-prepare. In a real 'unknown' scenario, such a procedure may not be available.
  • domain assumption SWAP-test fidelity can be estimated accurately from a finite number of shots
    The paper does not report shot counts or confidence intervals for the fidelity estimates used as learning signals.
  • ad hoc to paper The neural generator architecture and GELU activations are sufficient to represent any 2^n-dimensional state
    No theoretical guarantee; the architecture is chosen heuristically.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Guess, SWAP, Repeat : Capturing Quantum Snapshots in Classical Memory." pith.science (2026). https://pith.science/paper/UJZ5IFEV

@misc{pith2026250414459,
  author       = {Pith},
  title        = {Pith review of: Guess, SWAP, Repeat : Capturing Quantum Snapshots in Classical Memory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UJZ5IFEV}},
  note         = {Machine review of arXiv:2504.14459}
}
read the original abstract

We introduce a novel technique that enables observation of quantum states without direct measurement, preserving them for reuse. Our method allows multiple quantum states to be observed at different points within a single circuit, one at a time, and saved into classical memory without destruction. These saved states can be accessed on demand by downstream applications, introducing a dynamic and programmable notion of quantum memory that supports modular, non-destructive quantum workflows. We propose a hardware-agnostic, machine learning-driven framework to capture non-destructive estimates, or "snapshots," of quantum states at arbitrary points within a circuit, enabling classical storage and later reconstruction, similar to memory operations in classical computing. This capability is essential for debugging, introspection, and persistent memory in quantum systems, yet remains difficult due to the no-cloning theorem and destructive measurements. Our guess-and-check approach uses fidelity estimation via the SWAP test to guide state reconstruction. We explore both gradient-based deep neural networks and gradient-free evolutionary strategies to estimate quantum states using only fidelity as the learning signal. We demonstrate a key component of our framework on IBM quantum hardware, achieving high-fidelity (approximately 1.0) reconstructions for Hadamard and other known states. In simulation, our models achieve an average fidelity of 0.999 across 100 random quantum states. This provides a pathway toward non-volatile quantum memory, enabling long-term storage and reuse of quantum information, and laying groundwork for future quantum memory architectures.

Figures

Figures reproduced from arXiv: 2504.14459 by the authors.

Figure 1
Figure 1. Methodology for Non-Destructive Quantum State Observation (‘Snapshot’): This figure illustrates the methodology behind the three primary processes for observing a quantum state with n qubits. The approach begins with an unknown target circuit (marked in light blue) and a neural network model that generates an output representing either a state vector (Method 1), a unitary (Method 2), or a density matrix (Method 3). … view at source ↗
Figure 2
Figure 2. Progression of reconstructed states towards target states for noiseless and noisy conditions: The left figure shows the noiseless case where the target unknown state remains fixed, allowing smooth convergence. In contrast, the right figure illustrates the noisy case, where the target state keeps shifting due to hardware noise, making convergence significantly more challenging. III. METHODS A. The Pivotal Design Prin… view at source ↗
Figure 3
Figure 3. Entanglement entropy analysis of target and reconstructed states: (a) Distribution of target entanglement entropy across qubit counts. (b) Comparison of target and reconstructed entanglement entropy, sorted by circuit index, showing close alignment across different system sizes. are reloaded into quantum circuits using state preparation routines. IV. RESULTS AND ANALYSIS We have tested our methodology in real quantu… view at source ↗
Figures from the paper (3 more)
Figure 8
Figure 8. Figure 8: Epochs required to reach fidelity thresholds (> 0.95 and > 0.99) for state vector reconstruction across different qubit counts, under both noiseless and noisy conditions. Results are shown for gradient-based (NN) and gradient-free (ES) estimation strategies. The gradie…
Figure 6
Figure 6. Figure 6: Fidelity distributions for density matrix reconstruction under noiseless conditions using the gradient-based (NN) method. (a) Trace fidelity for 2-qubit mixed states; (b) Uhlmann fidelity across larger number of qubits. confirming that our optimization techniques are e…
Figure 7
Figure 7. Figure 7: Comparison of generated and target (unknown) quantum states in the state vector representation. The amplitudes of each basis state (|00⟩, |01⟩, |10⟩, |11⟩) are shown across multiple samples. Close alignment of amplitudes indicates accurate reconstruction. 1 2 3 4 5 6 Q…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 6 canonical work pages

  1. [1]

    Quantum memories: emerging applications and recent advances,

    K. Heshami, D. G. England, P. C. Humphreys, P. J. Bustard, V . M. Acosta, J. Nunn, and B. J. Sussman, “Quantum memories: emerging applications and recent advances,” Journal of modern optics , vol. 63, no. 20, pp. 2005–2028, 2016

    2005

  2. [2]

    Feedback control of a solid-state qubit using high-fidelity projective measurement,

    D. Rist `e, C. Bultink, K. W. Lehnert, and L. DiCarlo, “Feedback control of a solid-state qubit using high-fidelity projective measurement,” Physical review letters, vol. 109, no. 24, p. 240502, 2012

    2012

  3. [3]

    Character- izing large-scale quantum computers via cycle benchmarking,

    A. Erhard, J. J. Wallman, L. Postler, M. Meth, R. Stricker, E. A. Martinez, P. Schindler, T. Monz, J. Emerson, and R. Blatt, “Character- izing large-scale quantum computers via cycle benchmarking,” Nature communications, vol. 10, no. 1, p. 5347, 2019

    2019

  4. [4]

    Real-time dynamics of open quantum spin systems driven by dissipative processes,

    F. Hebenstreit, D. Banerjee, M. Hornung, F.-J. Jiang, F. Schranz, and U.-J. Wiese, “Real-time dynamics of open quantum spin systems driven by dissipative processes,” Physical Review B, vol. 92, no. 3, p. 035116, 2015

    2015

  5. [5]

    Efficient Quantum State Sample Tomography with Basis-Dependent Neural Networks,

    A. W. R. Smith, J. Gray, and M. S. Kim, “Efficient Quantum State Sample Tomography with Basis-Dependent Neural Networks,” PRX Quantum, vol. 2, p. 020348, June 2021. Publisher: American Physical Society

    2021

  6. [6]

    Parity measurement in the strong dispersive regime of circuit quantum acoustodynamics,

    U. von L ¨upke, Y . Yang, M. Bild, L. Michaud, M. Fadel, and Y . Chu, “Parity measurement in the strong dispersive regime of circuit quantum acoustodynamics,” Nature Physics , vol. 18, pp. 794–799, July 2022. Publisher: Nature Publishing Group

    work page 2022

  7. [7]

    Quantum decoherence,

    M. Schlosshauer, “Quantum decoherence,” Physics Reports , vol. 831, pp. 1–57, 2019

    2019

  8. [8]

    What are the ultimate limits of photonic quantum memories?,

    M. G ¨undo˘gan and D. K. Oi, “What are the ultimate limits of photonic quantum memories?,” Research Directions: Quantum Technologies , vol. 1, p. e8, 2023

    2023

Show all 24 references
  1. [9]

    The quantum internet,

    H. J. Kimble, “The quantum internet,” Nature, vol. 453, no. 7198, pp. 1023–1030, 2008

    2008

  2. [10]

    Quantum repeaters based on atomic ensembles and linear optics,

    N. Sangouard, C. Simon, H. De Riedmatten, and N. Gisin, “Quantum repeaters based on atomic ensembles and linear optics,” Reviews of Modern Physics, vol. 83, no. 1, pp. 33–80, 2011

    2011

  3. [11]

    Quantum internet: A vision for the road ahead,

    S. Wehner, D. Elkouss, and R. Hanson, “Quantum internet: A vision for the road ahead,” Science, vol. 362, no. 6412, p. eaam9288, 2018

    2018

  4. [12]

    What is quantum-centric supercomput- ing?,

    J. Schneider and I. Smalley, “What is quantum-centric supercomput- ing?,” 2024. Accessed: Apr. 12, 2025

    2024

  5. [13]

    Quantum random access memory,

    V . Giovannetti, S. Lloyd, and L. Maccone, “Quantum random access memory,” Physical Review Letters , vol. 100, p. 160501, Apr. 2008. arXiv:0708.1879 [quant-ph]

    2008 arXiv

  6. [14]

    Quantum Random Access Memory For Dummies,

    K. Phalak, A. Chatterjee, and S. Ghosh, “Quantum Random Access Memory For Dummies,” May 2023. arXiv:2305.01178 [quant-ph]

    2023 arXiv

  7. [15]

    Variational quantum circuits for quantum state tomography,

    Y . Liu, D. Wang, S. Xue, A. Huang, X. Fu, X. Qiang, P. Xu, H.-L. Huang, M. Deng, C. Guo, X. Yang, and J. Wu, “Variational quantum circuits for quantum state tomography,” Physical Review A , vol. 101, p. 052316, May 2020. Publisher: American Physical Society

    2020

  8. [16]

    Quantum State Tomography with Conditional Generative Adversarial Networks,

    S. Ahmed, C. S ´anchez Mu ˜noz, F. Nori, and A. F. Kockum, “Quantum State Tomography with Conditional Generative Adversarial Networks,” Physical Review Letters , vol. 127, p. 140502, Sept. 2021

    2021

  9. [17]

    Adaptive Quantum State Tomography with Neural Networks,

    Y . Quek, S. Fort, and H. K. Ng, “Adaptive Quantum State Tomography with Neural Networks,” Dec. 2018. arXiv:1812.06693 [quant-ph]

    2018 arXiv

  10. [18]

    Universal Quantum Tomogra- phy With Deep Neural Networks,

    N. T. Luu, T. C. Truong, and D. T. Luu, “Universal Quantum Tomogra- phy With Deep Neural Networks,” Sept. 2024. arXiv:2407.01734 [quant- ph]

    2024 arXiv

  11. [19]

    Adaptive Quantum State Tomography with Active Learn- ing,

    H. Lange, M. Kebri ˇc, M. Buser, U. Schollw ¨ock, F. Grusdt, and A. Bohrdt, “Adaptive Quantum State Tomography with Active Learn- ing,” Quantum, vol. 7, p. 1129, Oct. 2023. arXiv:2203.15719 [quant-ph]

    2023 arXiv

  12. [20]

    Gradient- descent methods for fast quantum state tomography,

    A. Gaikwad, M. S. Torres, S. Ahmed, and A. F. Kockum, “Gradient- descent methods for fast quantum state tomography,” Mar. 2025. arXiv:2503.04526 [quant-ph]

    2025

  13. [21]

    Transformation of quantum states using uniformly controlled rotations,

    M. Mottonen, J. J. Vartiainen, V . Bergholm, and M. M. Salomaa, “Transformation of quantum states using uniformly controlled rotations,” July 2004. arXiv:quant-ph/0407010

    2004 arXiv

  14. [22]

    Pennylane: Automatic differentiation of hybrid quantum-classical computations,

    V . Bergholm et al. , “Pennylane: Automatic differentiation of hybrid quantum-classical computations,” 2022

    2022

  15. [23]

    Evolution Strategies as a Scalable Alternative to Reinforcement Learning,

    T. Salimans, J. Ho, X. Chen, S. Sidor, and I. Sutskever, “Evolution Strategies as a Scalable Alternative to Reinforcement Learning,” Sept

  16. [2017]

    arXiv:1703.03864 [stat]. APPENDIX A NOISE MODEL To realistically simulate the behavior of near-term noisy quantum processors, we incorporate a detailed gate-level noise model derived from an IBM GenericBackendV2 for the required number of qubits. This model is constructed usin...

    arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.