Pith. sign in

REVIEW 3 major objections 5 minor 20 references

Continuous Variable Hamiltonian Learning at Heisenberg Limit via Displacement-Random Unitary Transformation

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read This paper claims that any finite-order multi-mode bosonic Hamiltonian can be characterized from vacuum at the Heisenberg limit, with total evolution time scaling as the inverse of the target precision, using a displacement-and-random-phase

desk verdict A credible Heisenberg-limited CV Hamiltonian learning protocol with a clean core, but the load-bearing Trotter step lacks error bounds and the promised numerics are missing. read the letter →

arxiv 2510.08419 v2 pith:7DN62W23 submitted 2025-10-09 quant-ph

Xi Huang , Lixing Zhang , Di Luo This is my paper
classification quant-ph PACS 03.67.-a
keywords HamiltonianlearningcontinuousvariablesHeisenberglimitbosonicsystemsdisplacementoperatorrandomunitarytransformationrobustphaseestimationpolynomialrecovery
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

The paper introduces Displacement-Random Unitary Transformation (D-RUT), a protocol for learning the coefficients of arbitrary finite-order bosonic Hamiltonians. The central claim is that these coefficients can be recovered with total evolution time O(1/ε) for target RMSE ε — the Heisenberg limit — using only vacuum input, displacement gates, number-phase rotations, and an ancilla qubit. The key step is converting the Hamiltonian into a diagonal effective Hamiltonian whose vacuum eigenvalue is a polynomial whose coefficients are exactly the unknowns, then measuring that polynomial via robust phase estimation. For multi-mode systems, a hierarchical divide-and-conquer scheme recovers single-mode and coupling coefficients separately and provably achieves lower or equal estimation variance than a simultaneous scheme. The protocol is extended to first-quantized Hamiltonians in position and momentum, yielding physical parameters at the Heisenberg limit up to a logarithmic factor, under conditions of prior knowledge of a signal coefficient.

What carries the argument

The load-bearing object is the Displacement-Random Unitary Transformation (D-RUT): a two-step operation D(β) then phase-averaging U(θ)=e^{-iθN}, which maps any finite-order bosonic Hamiltonian to a number-conserving effective Hamiltonian with constant term C(β). This constant term is a polynomial in the complex displacement β of degree equal to the Hamiltonian order, with coefficients equal to the unknowns. The recovery pipeline consists of Chebyshev interpolation in the radial direction (to suppress Runge phenomena and keep the Vandermonde matrix well-conditioned) and an inverse discrete Fourier transform in the angular direction. The phase is read out with robust phase estimation, which pr

What would settle it

Compute or measure the phase error of the Trotterized D-RUT unitary (Eq. 20) on the vacuum state for a known single-mode Hamiltonian as a function of the number of Trotter steps L and the RPE evolution depth κ=2^j. If the minimum L needed to keep the phase error below the RPE tolerance grows faster than O(1/ε) — or if the ancilla contrast in the measured probabilities (Eqs. 24-25) decays with κ — then the claimed total evolution time O(1/ε) will fail. Concretely, an experiment with a Kerr oscillator with known coefficients could measure the recovered RMSE versus total evolution time and check

Watch

Extended reading notes

Core claim

D-RUT displaces the unknown Hamiltonian by D(β) and averages over random number-phase rotations, projecting away all off-diagonal terms. The resulting effective Hamiltonian is diagonal in the number operator, and its vacuum eigenvalue C(β) is exactly the generating polynomial C(β) = Σ g_{p,q}(β*)^p β^q whose coefficients are the Hamiltonian parameters to be learned. Measuring C(β) at Chebyshev radial nodes and a discrete set of angles, then inverting a Vandermonde system and a discrete Fourier transform, recovers all g_{p,q}. Robust phase estimation on the ancilla gives a Heisenberg-limited estimate of each C(β), and the paper proves the RMSE of the final coefficients scales as the measureme

Load-bearing premise

The protocol assumes that a finite Trotterized sequence of D-RUT operations with independently sampled random phases approximates the ideal effective-Hamiltonian evolution on the vacuum state with phase error below the robust-phase-estimation tolerance at every RPE step, without requiring a Trotter step count that grows faster than 1/ε; the paper does not provide that bound.

Editorial extensions

If this is right

  • If correct, any finite-order bosonic Hamiltonian — including strongly nonlinear terms — can be fully characterized from vacuum at the Heisenberg limit with experimentally accessible gates.
  • The hierarchical multi-mode strategy guarantees that no estimated parameter has larger variance than in the simultaneous approach, so larger coupled systems can be learned with fewer independent measurements.
  • The first-quantization extension allows extraction of physical parameters such as mass and frequency from Hamiltonians in x and p, at the Heisenberg limit up to a logarithmic factor, given a known-zero coefficient as a signal.
  • SPAM errors from imperfect displacement grow only linearly in the displacement deviation, with an amplification factor controlled by the conditioning of the recovery matrix, enabling experimental calibration of the probes.
  • Since the unknown coefficients are obtained from a small classical polynomial-recovery problem, standard numerical conditioning tools apply directly to the error analysis.

Reading between the lines

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

  • The same displacement-and-averaging trick could likely be adapted to learning couplings of bosonic lattice systems with different kinetic structures, e.g., gauge-coupling terms, by choosing a different averaging group that preserves a different symmetry.
  • In the first-quantization setting, the known-zero coefficient acts as a built-in alignment beacon; this suggests a general calibration principle: any Hamiltonian with a symmetry-forbidden coefficient can be self-calibrated without external reference.
  • A natural test of the claimed O(1/ε) scaling is to implement D-RUT on a single Kerr oscillator with known coefficients and measure the empirical RMSE as a function of total evolution time; a deviation beyond a constant factor would indicate that the Trotter step bound is too optimistic.
  • Because the protocol's SPAM robustness is linear in displacement error, D-RUT itself could be repurposed as a displacement-calibration routine for CV hardware, by learning a known Hamiltonian and using the residual error to infer gate imperfections.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces D-RUT, a protocol for learning the coefficients of finite-order multi-mode bosonic Hamiltonians of the form Eq. (1). The idea is to displace the Hamiltonian by D(β), average over random phase rotations U(θ)=e^{-iθN}, and thereby define an effective number-conserving Hamiltonian H̄(β) whose vacuum eigenvalue is the polynomial C(β)=Σ g_{p,q}(β*)^p β^q. C(β) is estimated by robust phase estimation using a Trotterized random-phase product (Eq. 20), and the coefficients are recovered by Chebyshev interpolation in the radial variable and inverse DFT in the angular variable. For multi-mode systems the protocol uses a hierarchical two-stage strategy (single-mode terms first, then couplings), and it is claimed to have lower or equal covariance than the simultaneous strategy of [17]. The paper also extends the approach to first-quantized Hamiltonians via a Bogoliubov transformation and an iterative search for the physical frame. The main theorem asserts total evolution time O(1/ε), hierarchical statistical efficiency, and robustness to small displacement SPAM errors.

Significance. If the main claims hold, the paper provides a valuable reduction: arbitrary finite-order bosonic Hamiltonian learning is reduced to polynomial inversion, with a Heisenberg-limited total evolution time and an explicit covariance comparison. The algebraic core is sound: Eq. (19) is an exact, parameter-free expression for C(β), the radial/angular recovery (Eqs. 26–30) is a clean Chebyshev/DFT inversion, and the block-inverse PSD comparison in Appendix A is correct for a fixed partitioned design. The extension to first quantization is clearly conditional on stated assumptions, and the paper honestly lists the prior-knowledge limitations in the conclusion. However, the measurement step is not rigorously controlled: the Trotterized random-phase implementation of the effective Hamiltonian is asserted without a finite-L error bound, and the hierarchical covariance claim is not fully tied to the actual experimental designs used in the protocol. These gaps directly affect the central claims, so the manuscript needs revision before the theorems can be accepted.

major comments (3)
  1. [§3.3, Eqs. (20)–(25)] The central measurement step is asserted, not proved. The unitary product in Eq. (20) is claimed to act on |vac⟩ as e^{-iC(β)κ}|vac⟩ 'at the limit of infinite L', and the text states without proof that finite L will not destroy Heisenberg scaling. For independent random θ_j, the Hamiltonians U†(θ_j)D†HD U(θ_j) do not commute; standard Trotter analysis gives a deterministic phase error O(κ²/L) and a stochastic phase fluctuation O(κ/√L). For κ≈1/ε, keeping the phase error below the RPE tolerance requires L to grow at least as ε^{-4} (fluctuation) or ε^{-3} (commutator), and the accumulated orthogonal component makes ⟨vac|U(κ)|vac⟩ have modulus V<1. The measured probabilities then take the form (1+V cos(κC+δ))/2, which is not the independent additive error ε_C used in §3.5 and Appendix A. Since Theorem 1, claim 1, and the robustness claim both rest on this step, a rigorous finite-L bound an
  2. [Appendix A, Eqs. (92)–(108) vs §4.1] The hierarchical covariance comparison is not obviously tied to the protocol's actual experimental designs. The appendix compares simultaneous and hierarchical strategies using the same matrices M1, M>1 and the same Gram blocks A=M1†M1, D=M>1†M>1. But §4.1.1 learns single-mode terms by setting all other displacement parameters to zero, while §4.1.2 learns coupling terms from full multi-mode measurements; these are different designs. The M1 matrix in Step 1 is not necessarily the same as the single-mode block of the simultaneous design. Without specifying how the experiment points are matched, or how the two designs' Gram matrices are related, the PSD inequality does not establish that the hierarchical protocol has lower variance than the simultaneous protocol. This needs to be clarified or the proof restated with distinct matrices.
  3. [§3.6, Eq. (43)] The SPAM robustness bound is ||δg_SPAM||² ≤ (L_C/σ_min(K))||δβ||², but no lower bound on σ_min(K) is provided. The statement that robustness is 'controllable through strategic selection' of {β_j} is not a proof: for the claimed linear control to be meaningful, one needs a concrete choice of r_min, r_max, d, and Chebyshev nodes with an explicit conditioning bound. As written, the bound could be vacuous if σ_min(K) is exponentially small in d or in the interval width. This directly affects Theorem 1 claim 3.
minor comments (5)
  1. [Abstract / §6] The abstract states that 'numerical experiments on single- and multi-mode nonlinear systems validate the predicted Heisenberg scaling', but the main text contains no numerical experiments, plots, or simulation results. Either include the numerical section or remove the claim.
  2. [§3.3, Eq. (21)] The notation e^{-iH̄(β)κ}|vac⟩ is used after a finite-L product. Since the product is random, the limit L→∞ needs a precise mode of convergence (e.g., in probability, in expectation, or in the sense of a Trotter–Kato limit). The current phrasing is informal.
  3. [§3.6.1, Eqs. (44)–(46)] The Lipschitz bound splits cases 'if |g_{p,q}|≤1' and 'if |g_{p,q}|≥1', but this is not exhaustive for a general bounded coefficient set. The bound with Σ|g_{p,q}| covers both cases; the presentation should be unified.
  4. [Theorem 2] The overlap condition in Theorem 2, condition 3, is written as sqrt(m0ω0/mω)+sqrt(mω/m0ω0) < 1/(2−√3), while §5.3 Eq. (63) uses u<1/(4−2√3). These are equivalent (multiply by 2), but the connection should be stated to avoid confusion.
  5. [General] Several typographical issues: 'the t th bosonic mode', 'divide-and-conque', 'stateded below', and 'Leaning'. The notation for the ordered set S and the tuple powers p_S, q_S could be defined more cleanly to avoid confusion in multi-mode expressions.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the coefficient-recovery chain is an exact algebraic inversion, and the only same-author citation is non-load-bearing provenance.

full rationale

The central derivation is self-contained. The constant term C(β)=∑ g_{p,q}(β*)^p β^q is obtained by direct algebraic identities (Eqs. 14-19), and the recovery of g_{p,q} is a parameter-free linear inversion via Chebyshev interpolation and inverse DFT (Eqs. 26-30). The measured C(β) values are not fitted to the target coefficients; they are independent RPE phase estimates, so the inversion is not circular. The Heisenberg-limit scaling is imported from the external RPE references [18,19], not from a self-citation. Reference [16] is a same-author-group paper cited for the RUT idea, but the paper proves the needed phase-averaging projection itself (Eq. 12) and also cites the independent [15], so the self-citation is provenance and not load-bearing. The first-quantization result (Theorem 2) is explicitly conditional on prior knowledge of a coefficient known to be zero and a non-zero response; this is a stated assumption of the iterative self-calibration, not a hidden circularity. The main weakness is in Sec. 3.3 (Eqs. 20-21): the convergence of the finite-L Trotterized D-RUT sequence to e^{-iC(β)κ}|vac> is asserted without a quantitative error bound, and the resulting contrast loss is not analyzed. That is a correctness/rigor gap in proving the claimed O(1/ε) evolution time, but it is not circular, because the target coefficients are not inserted as fit parameters in that convergence argument. Hence the circularity score is 0.

Assumptions & free parameters 4 free parameters · 7 assumptions · 0 invented entities

The central claim rests on the measurement model (unitary access and gate interleaving), the exact-degree-d polynomial model, the unproven convergence of the random-phase Trotterized product, the equal-variance/independence noise model, and — for Theorem 2 — three explicit prior-knowledge conditions. No fundamentally new physical entities are postulated; H̄(β) is a mathematical average, not a new interaction. The main free choices are the Trotter count L, the displacement-radius interval, the initial reference frame, and the per-measurement precision allocation.

free parameters (4)
  • Trotter step count L = unspecified ('large enough')
    The D-RUT sequence (Eq. 20) replaces the ideal averaged Hamiltonian with L discrete random-phase steps; the replacement error scales like O(κ²/L) but L is never bounded. The Heisenberg-limit claim implicitly treats L as a free knob chosen large enough, and the gate overhead depends on it.
  • Displacement radius interval [r_min, r_max] and Chebyshev nodes {r_μ} = unspecified
    The recovery error (MSE proportional to Tr[(L†L)^{-1}], §3.5.1) depends on the conditioning of the Chebyshev-Vandermonde matrix, which depends on the chosen interval; the paper never specifies how to choose r_min, r_max, leaving the constant factor in the O(1/ε) scaling uncontrolled.
  • Reference frame (m_0, ω_0) and initial squeezing guess R' = constrained by u < 1/(4-2√3) ≈ 1.866
    In the first-quantization protocol, the initial guess (m_0, ω_0) is a free input that must be sufficiently close to the truth; the answer is refined via bisection, but the starting interval and its size are free choices (Theorem 2 condition 3, §5.3).
  • Per-measurement RPE precision allocation ε_C = ε_C proportional to ε/cond(K)
    Each C(β) is estimated to precision ε_C; the relation between ε_C and the final target ε is set by the conditioning of the recovery maps, and the allocation of evolution time across measurement points is a protocol choice determining the total time constant.
assumptions (7)
  • domain assumption Unitary access to e^{-iHt} for arbitrary times t, plus ability to interleave displacement D(β) and number-rotation U(θ) gates
    The whole protocol is built on applying the unknown evolution e^{-iH(κ/L)} between gates (Eq. 20); standard in Hamiltonian learning but physically nontrivial for unbounded CV operators.
  • domain assumption Hamiltonian is exactly finite order d (Eq. 1), with no higher-order contamination
    C(β) is exactly a degree-d polynomial in β; any higher-order term biases every recovered coefficient. This is the model the learner assumes, stated in the problem definition.
  • ad hoc to paper Random-phase Trotterized product converges to e^{-iH̄(β)κ} with negligible phase error and preserved contrast for large L
    Asserted in §3.3 (Eqs. 20-21) without a bound on L or an analysis of dephasing in the ancilla measurement (Eqs. 24-25); the Heisenberg-limit measurement relies on this convergence.
  • standard math RPE overlap condition p_0 > 4-2√3 (from [19])
    Used in §5.3 to constrain the allowed reference-frame mismatch; taken from the cited robust-phase-estimation literature.
  • domain assumption Independence and equal variance ε_C² of the C(β_j) measurements
    The error-propagation analysis (§3.5) and the PSD comparison (Appendix A) assume independent equal-variance measurement noise; this is a statistical model, not a proven property of RPE.
  • ad hoc to paper First-quantization: existence of a coefficient known to be ideally zero, with non-zero response to ΔR
    The iterative basis-search (§5.3, §6.1) needs this prior knowledge (Theorem 2, conditions 1-2); without it there is no signal function and no convergence. The paper acknowledges this as an open limitation in §7.
  • domain assumption Chebyshev-node Vandermonde system is well-conditioned (σ_min(K) not too small)
    The SPAM bound (Eq. 43) has amplification factor 1/σ_min(K); the paper asserts Chebyshev nodes keep this under control but does not quantify σ_min for the full K (Chebyshev + DFT) matrix.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Continuous Variable Hamiltonian Learning at Heisenberg Limit via Displacement-Random Unitary Transformation." pith.science (2026). https://pith.science/paper/7DN62W23

@misc{pith2026251008419,
  author       = {Pith},
  title        = {Pith review of: Continuous Variable Hamiltonian Learning at Heisenberg Limit via Displacement-Random Unitary Transformation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7DN62W23}},
  note         = {Machine review of arXiv:2510.08419}
}
read the original abstract

Characterizing continuous-variable (CV) Hamiltonians can be formulated as Hamiltonian learning under quantum measurement constraints: finite operator coefficients are inferred from noisy measurement outcomes obtained by probing an infinite-dimensional system. Existing Heisenberg-limited CV protocols are often limited to low-order structures, vulnerable to noise, or unresolved for generic multi-mode settings. We introduce Displacement-Random Unitary Transformation (D-RUT), an active data acquisition protocol with pre-specified probes and number-preserving transformations that reduce finite-order bosonic Hamiltonian learning to polynomial recovery. We prove Heisenberg-limited total evolution time with robustness to state preparation and measurement (SPAM) errors, and develop hierarchical multi-mode coefficient recovery with better statistical efficiency than simultaneous estimation. We also extend D-RUT to first-quantized Hamiltonian coefficient learning, and numerical experiments on single- and multi-mode nonlinear systems validate the predicted Heisenberg scaling.

Figures

Figures reproduced from arXiv: 2510.08419 by the authors.

Figure 1
Figure 1. The D-RUT based learning algorithm for single mode coefficients. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

20 extracted references · 6 linked inside Pith

  1. [17]

    Heisenberg-limited hamiltonian learning continuous variable systems via engineered dissipa- tion.arXiv preprint arXiv:2506.00606, 2025

    Tim M¨ obus, Andreas Bluhm, Tuvia Gefen, Yu Tong, Albert H Werner, and Cambyse Rouz´ e. Heisenberg-limited hamiltonian learning continuous variable systems via engineered dissipa- tion.arXiv preprint arXiv:2506.00606, 2025

    arXiv 2025

  2. [1]

    Towards practical characterization of quantum systems with quan- tum hamiltonian learning

    R Santagati, J Wang, S Paesani, S Knauer, AA Gentile, N Wiebe, M Petruzzella, JL O’Brien, JG Rarity, A Laing, et al. Towards practical characterization of quantum systems with quan- tum hamiltonian learning. InFrontiers in Optics, pages FTh3E–7. Optica Publishing Group, 2017

    2017

  3. [2]

    Hamiltonian learning for 300 trapped ion qubits with long-range couplings.Science Advances, 11(5):eadt4713, 2025

    Shi-An Guo, Yu-Kai Wu, Jing Ye, Lin Zhang, Ye Wang, Wen-Qian Lian, Rui Yao, Yu-Lin Xu, Chi Zhang, Yu-Zi Xu, et al. Hamiltonian learning for 300 trapped ion qubits with long-range couplings.Science Advances, 11(5):eadt4713, 2025

    2025

  4. [3]

    Active learning of quantum system hamiltonians yields query advantage.URL https://arxiv

    Arkopal Dutt, Edwin Pednault, Chai Wah Wu, Sarah Sheldon, John Smolin, Lev Bishop, and Isaac L Chuang. Active learning of quantum system hamiltonians yields query advantage.URL https://arxiv. org/abs/2112.14553, 2021

    arXiv 2021

  5. [4]

    Learning many-body hamiltonians with heisenberg-limited scaling.Physical Review Letters, 130(20):200403, 2023

    Hsin-Yuan Huang, Yu Tong, Di Fang, and Yuan Su. Learning many-body hamiltonians with heisenberg-limited scaling.Physical Review Letters, 130(20):200403, 2023

    2023

  6. [5]

    Structure learning of hamiltonians from real-time evolution (2024).URL https://arxiv

    Ainesh Bakshi, Allen Liu, Ankur Moitra, and Ewin Tang. Structure learning of hamiltonians from real-time evolution (2024).URL https://arxiv. org/abs/2405.00082

    arXiv 2024

  7. [6]

    Ansatz-free hamiltonian learning with heisenberg-limited scaling.arXiv preprint arXiv:2502.11900, 2025

    Hong-Ye Hu, Muzhou Ma, Weiyuan Gong, Qi Ye, Yu Tong, Steven T Flammia, and Su- sanne F Yelin. Ansatz-free hamiltonian learning with heisenberg-limited scaling.arXiv preprint arXiv:2502.11900, 2025

    arXiv 2025

  8. [7]

    Learning interacting fermionic hamiltonians at the heisenberg limit (2024).arXiv preprint arXiv:2403.00069

    A Mirani and P Hayden. Learning interacting fermionic hamiltonians at the heisenberg limit (2024).arXiv preprint arXiv:2403.00069

    arXiv 2024

Show all 20 references
  1. [8]

    Effi- cient quantum transduction using antiferromagnetic topological insulators.Physical Review B, 110(8):085136, 2024

    Haowei Xu, Changhao Li, Guoqing Wang, Hao Tang, Paola Cappellaro, and Ju Li. Effi- cient quantum transduction using antiferromagnetic topological insulators.Physical Review B, 110(8):085136, 2024

    2024

  2. [9]

    Metropolitan-scale heralded entanglement of solid-state qubits.Science advances, 10(44):eadp6442, 2024

    Arian J Stolk, Kian L van der Enden, Marie-Christine Slater, Ingmar te Raa-Derckx, Pieter Botma, Joris van Rantwijk, JJ Benjamin Biemond, Ronald AJ Hagen, Rodolf W Herfst, Wouter D Koek, et al. Metropolitan-scale heralded entanglement of solid-state qubits.Science advances, 10...

    2024

  3. [10]

    Logical states for fault-tolerant quantum computation with propagating light.Science, 383(6680):289– 293, 2024

    Shunya Konno, Warit Asavanant, Fumiya Hanamura, Hironari Nagayoshi, Kosuke Fukui, At- sushi Sakaguchi, Ryuhoh Ide, Fumihiro China, Masahiro Yabuno, Shigehito Miki, et al. Logical states for fault-tolerant quantum computation with propagating light.Science, 383(6680):289– 293, 2024

    2024

  4. [11]

    Fault-tolerant quantum computation by hybrid qubits with bosonic cat code and single pho- tons.PRX Quantum, 5(3):030322, 2024

    Jaehak Lee, Nuri Kang, Seok-Hyung Lee, Hyunseok Jeong, Liang Jiang, and Seung-Woo Lee. Fault-tolerant quantum computation by hybrid qubits with bosonic cat code and single pho- tons.PRX Quantum, 5(3):030322, 2024

    2024

  5. [12]

    Toward hybrid quantum simu- lations with qubits and qumodes on trapped-ion platforms.arXiv preprint arXiv:2410.07346, 2024

    Jack Y Araz, Matt Grau, Jake Montgomery, and Felix Ringer. Toward hybrid quantum simu- lations with qubits and qumodes on trapped-ion platforms.arXiv preprint arXiv:2410.07346, 2024. 23

    2024 arXiv

  6. [13]

    Quantum metrology with a continuous- variable system.Reports on Progress in Physics, 2024

    Matteo Fadel, Noah Roux, and Manuel Gessner. Quantum metrology with a continuous- variable system.Reports on Progress in Physics, 2024

    2024

  7. [14]

    Quan- tum metrological power of continuous-variable quantum networks.Physical Review Letters, 128(18):180503, 2022

    Hyukgun Kwon, Youngrong Lim, Liang Jiang, Hyunseok Jeong, and Changhun Oh. Quan- tum metrological power of continuous-variable quantum networks.Physical Review Letters, 128(18):180503, 2022

    2022

  8. [15]

    Heisenberg-limited hamil- tonian learning for interacting bosons.npj Quantum Information, 10(1):83, 2024

    Haoya Li, Yu Tong, Tuvia Gefen, Hongkang Ni, and Lexing Ying. Heisenberg-limited hamil- tonian learning for interacting bosons.npj Quantum Information, 10(1):83, 2024

    2024

  9. [16]

    Hamiltonian learning at heisenberg limit for hybrid quantum systems.arXiv preprint arXiv:2502.20373, 2025

    Lixing Zhang, Ze-Xun Lin, Prineha Narang, and Di Luo. Hamiltonian learning at heisenberg limit for hybrid quantum systems.arXiv preprint arXiv:2502.20373, 2025

    2025 arXiv

  10. [18]

    Robust calibration of a universal single-qubit gate set via robust phase estimation.Physical Review A, 92(6):062315, 2015

    Shelby Kimmel, Guang Hao Low, and Theodore J Yoder. Robust calibration of a universal single-qubit gate set via robust phase estimation.Physical Review A, 92(6):062315, 2015

    2015

  11. [19]

    On low-depth algorithms for quantum phase esti- mation.Quantum, 7:1165, 2023

    Hongkang Ni, Haoya Li, and Lexing Ying. On low-depth algorithms for quantum phase esti- mation.Quantum, 7:1165, 2023

    2023

  12. [20]

    Simulating arbitrary gaussian circuits with linear optics.Physical Review A, 98(6):062314, 2018

    Levon Chakhmakhchyan and Nicolas J Cerf. Simulating arbitrary gaussian circuits with linear optics.Physical Review A, 98(6):062314, 2018. 24 A Comparative Analysis of Statistical Efficiency We now rigorously prove that our hierarchical strategy is statistically more robust and...

    2018

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.