Pith. sign in

REVIEW 3 major objections 6 minor 50 references

Variance-Reduced Trajectory Unravelings for GPU Noisy Quantum-Circuit Simulation: Characterization and a Qiskit-Aer Integration Gap

T0 review · 3 major / 6 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Variance-reduced projector unravelings cut the number of trajectories needed for noisy GPU quantum-circuit simulation by roughly 21× at identical accuracy, and the paper pinpoints why a leading production simulator cannot yet expose them.

desk verdict Solid trajectory-reduction result, unverified Aer integration-gap diagnosis: the statistical claim holds up, the systems claim needs proof. read the letter →

arxiv 2607.17678 v1 pith:GVSN6HSG submitted 2026-07-20 quant-ph physics.comp-ph

Chun-Yeol You This is my paper · ORCID
classification quant-phphysics.comp-ph
keywords variance-reducedunravelingquantumtrajectoriesMonteCarlowavefunctionnoisysimulationGPUdensestatevectorprojectoranalognoise-channelcanonicalization
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 establishes that the statistical bottleneck of Monte-Carlo trajectory simulation of noisy quantum circuits can be greatly reduced by choosing a different unraveling of the same noise channel. Implementing projector and analog unravelings on a GPU dense-statevector engine, the author shows they reproduce the exact density-matrix result while the projector unraveling needs about 21× fewer trajectories to hit a target standard error compared with a production GPU trajectory path. The gain is stable across 8–20 qubits, is largest when coherent dynamics is weak, and comes with a clear noise-strength regime map. The paper also reports a systems finding: the production simulator applies noise at the channel level and rebuilds a canonical Kraus decomposition at apply time, discarding user-supplied unravelings, so variance-reduced sampling cannot be delivered through its public API. A minimal three-part change is specified that would unlock the technique in production.

What carries the argument

The central object is the unraveling freedom of a fixed Lindblad generator: many different sets of jump operators reproduce the same density-matrix evolution but give estimators with very different variances. The paper exploits the projector unraveling L_± = √(2γ)Π_±, with Π_± projectors onto the ±1 eigenspaces of the dephasing operator Z; a jump collapses the state into a Z-eigenstate and pins any anticommuting observable to zero thereafter, turning the estimator into a low-variance Bernoulli variable. The analog unraveling instead applies frequent near-identity kicks e^{iθZ}, suppressing fluctuations at weak noise. On the systems side, the key mechanism is channel canonicalization: the sim

What would settle it

Implement the proposed three-part change in the production simulator and run the same benchmark (n=10, γt=1.5, target SE≤10⁻²): if the required trajectory count does not drop from roughly 998 to roughly 48, then either the canonicalization diagnosis is wrong or the proposed fix does not unlock the variance-reduced unraveling. Alternatively, inspect the simulator's apply-time code path to confirm whether user-supplied Kraus operators survive or are replaced by a reconstructed canonical set.

Watch

Extended reading notes

Core claim

Variance-reduced unraveling works on a GPU dense-statevector engine, not just CPU tensor networks: the projector unraveling reaches a target standard error with about 21× fewer trajectories than the production GPU path at strong noise (19–26× across n=8–20), and all three unravelings are unbiased against the exact density matrix (trace distance < 0.01). The projector collapses each trajectory into a Z-eigenspace, creating an absorbing window that suppresses fluctuations of anticommuting observables; analog unraveling wins below a crossover near γt=0.35. The production blocker is architectural: the simulator canonicalizes noise channels at apply time, discarding user-supplied Kraus decomposit

Load-bearing premise

The production-integration claim rests on the diagnosis that the production simulator discards any user-supplied Kraus decomposition and reconstructs its own canonical one at apply time; this was verified by code inspection, not by an implemented patch, so the promise of a 21× saving in production depends on that diagnosis and the minimal fix being sufficient.

Editorial extensions

If this is right

  • Noisy-circuit simulation on GPU dense statevectors can reach a target standard error with roughly 20× fewer trajectories in the strong-noise regime, directly reducing the statistical cost of variational algorithms and error-mitigation pipelines.
  • The proposed three-part modification to the production simulator—a preserve-unraveling flag, Born-rule sampling over the supplied Kraus set, and a per-trajectory expectation accumulator—would make the 21× saving available in production with machinery that already exists.
  • The crossover near γt≈0.35 gives a practical policy: use analog sampling under weak noise and projector sampling under strong noise; the projector advantage grows with accumulated noise strength.
  • Because the reduction is a statistical property rather than a memory effect, it is expected to persist on data-center GPUs up to roughly 30 qubits, a regime where exact density-matrix simulation is impossible.
  • The gain is not universal: interleaved coherent rotations erode the projector's absorbing window (the speedup falls from roughly 20× to about 2×), so the method pays off most for decoherence/idle-dominated workloads such as quantum-memory characterization.

Reading between the lines

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

  • If the canonicalization diagnosis is correct, the same blocker likely affects any channel-level simulator that rebuilds Kraus decompositions internally, so the proposed fix could benefit a whole class of production tools, not only the one studied.
  • The trajectory-count reduction is engine-independent, but the paper demonstrates it in a Python prototype; a compiled CUDA implementation would be the natural next test, and the author's projected 5–10× memory-bandwidth gain suggests a total wall-clock improvement of one to two orders of magnitude over a standard baseline.
  • The absorbing-window mechanism suggests a testable extension: for sparse Pauli-Lindblad noise with multiple channels, an adaptive per-channel and per-time-window choice of unraveling—projector, analog, or standard—could be made at runtime using the measured crossover as a policy.
  • The validation stops at n=20 on an 8 GB GPU; whether the 19–26× factor persists at larger n near memory capacity is a projection, not a measurement, but since the advantage is statistical rather than memory-bound, there is no obvious n-dependence that would break it.
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 / 6 minor

Summary. The paper presents a GPU implementation of two variance-reduced trajectory unravelings (projector and analog) for noisy quantum-circuit simulation, alongside a characterization of their statistical efficiency and a claimed systems-level obstacle in Qiskit-Aer. The core result is that, for a single-qubit dephasing channel in the strong-noise regime (γt=1.5), the projector unraveling reaches a target standard error with about 20.8× fewer trajectories than Qiskit-Aer's batched_shots_gpu standard unraveling at n=10, with the factor stated to remain in the 19–26× range for n=8–20. The engine is validated against Qiskit-Aer's exact density-matrix simulation: ideal-circuit fidelity 1−2.2×10⁻¹⁶, trace-distance convergence at the 1/√N rate, and all three unravelings unbiased to trace distance below 0.01. Closed-form variance predictions (Eqs. (6)–(7)) are reproduced. The paper also reports that Qiskit-Aer canonicalizes user-supplied Kraus decompositions at apply time, discarding any user-specified unraveling, and proposes a three-step minimal change (preserve-unraveling flag, Born-rule sampling over the supplied Kraus set, per-trajectory expectation accumulator) to unlock the technique in production.

Significance. If the quantitative claims hold, the paper delivers a practically important result: a ~20× reduction in trajectory count for GPU-based noisy simulation, achieved without changing the simulated channel and verified against an external exact benchmark. The validation methodology is a strength: the estimator is unbiased against exact density matrices, the convergence rate is demonstrated, and the head-to-head control against Aer's standard unraveling (997 vs 998 trajectories, identical per-sample spread) correctly isolates the unraveling as the source of the saving. The closed-form variance predictions are derived without fitted constants and are empirically reproduced; the paper also ships reproduction scripts and raw data. However, the production-integration diagnosis is unverified — no implementation of the proposed fix demonstrates the saving inside Aer — and the analog results depend on an unspecified free parameter (the kick angle θ₀). These gaps affect central claims (the integration gap is in the title/abstract; the analog regime map is a headline result), so the manuscript is not yet in publishable form.

major comments (3)
  1. [§7.2–7.3] The central systems finding — that Qiskit-Aer discards user-supplied Kraus decompositions by reconstructing a canonical set at apply time — is inferred from source inspection and is not verified by an implementation. The proposed three-step change (preserve-unraveling flag, Born-rule sampling over the supplied set, per-trajectory accumulator) is untested. Because the title and abstract present this integration gap as a principal contribution, the claim needs direct support: e.g., a minimal patch to qiskit-aer-gpu 0.15.1 that demonstrates the 20.8× trajectory reduction through the Aer API, or at minimum a reproducible test that the supplied Kraus set is discarded. Without this, the integration-gap contribution should be reframed as a hypothesis.
  2. [Table 3, §6.3; abstract] The speedup numbers in Table 3 do not match the ratios of the listed trajectory counts. For n=16, 632/25=25.3, not 25.7; for n=20, 539/22=24.5, not 25.0. More seriously, n=12 gives 785/44=17.8, which contradicts the abstract's '19–26× across n=8–20' and the text's 'essentially flat' 19.1–25.7× range. Either recompute the ratios with error bars (the N values are themselves estimates of a threshold), or revise the stated range. As written, the headline scaling claim is inconsistent with the data.
  3. [§3.2, §4.2, §6.1, Table 2] The analog unraveling is defined by a two-point kick law θ=±θ₀ with λ sin²θ₀=γ, but the paper never states the value of θ₀ used in the experiments, nor does it analyze the sensitivity of the results to this choice. The analog variance ratios in Table 2 and the crossover at γt≈0.35 depend on this free parameter, making the regime map irreproducible. Please specify θ₀ (and the sampling distribution w) for every reported analog result, and show that the qualitative regime map is stable across reasonable θ₀ choices. If analog is intended only as an illustration, remove it from the abstract's conclusions.
minor comments (6)
  1. [§5 (Fig. 2 caption)] The figure caption states 'D·√N stays near 1.0', while the text reports values 1.42, 0.81, 1.19, 0.87, 1.02 — a spread from 0.81 to 1.42. Please harmonize the caption or provide error bars to justify 'near constant'.
  2. [§6.4] The phrase 'The earlier "no benefit for GHZ"' refers to a discussion not present in the manuscript. Either add the earlier claim with appropriate context or rephrase the sentence.
  3. [§7.2] The claim that Aer's statevector kraus sampler is 'proper Born-rule collapse' with '29.6% collapsed' is an empirical detail without a measurement description or version-stamped citation. Please state how this number was obtained and under which conditions.
  4. [§11 / Disclosure] The AI-assistance note is transparent and helpful. However, the assertion 'no performance number is AI-generated' is not verifiable from the manuscript; please phrase it as a verification procedure (e.g., 'all reported numbers were regenerated from the raw data by the author'). Also, references [49] and [50] are to one published paper and one 'under review' preprint; the latter is not publicly attributable and should be marked or removed.
  5. [References] Several entries ([14], [17], [43]) are listed as 'Authors of arXiv...' with no author names. Use the standard author list where available, or indicate the reason for anonymity.
  6. [§6.3 Table 3] The statevector memory column shows 4 KB for n=8 but the text says 16.8 MB at n=20; both are correct, but the table mixes decimal prefixes without units. Clarify with KiB/MiB if precision matters.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the variance-reduction and trajectory-saving claims are derived from first principles and validated against an external exact-density-matrix benchmark; the unverified Aer integration-gap diagnosis is a correctness risk, not a circular step.

full rationale

The central variance-reduction claim is self-contained. Equations (6)-(7) in Section 3.3 are derived directly from the Poisson/absorbing-window dynamics of the standard and projector unravelings defined in Section 3.2, with no fitted constants; Section 6.1 reproduces them (e.g. 0.0485 vs predicted 0.0473), and Section 5 validates unbiasedness against Qiskit-Aer's exact density matrix as an external benchmark. The head-to-head trajectory saving is likewise a closed-form consequence: at γt=1.5, Var_proj/Var_std = e^{-2γt}/(1+e^{-2γt}) ≈ 0.047, so the ~21× trajectory reduction is the ratio of two analytically derived variances, not a fitted parameter renamed as a prediction. The analog parameter condition λ sin²θ0 = γ is an exact map to the same channel, not a data fit. The Section 7.2 Aer-canonicalization diagnosis is an unverified code-inspection claim, but it is falsifiable and independent of the trajectory-reduction result; a wrong diagnosis would weaken the integration-gap contribution without touching the core statistical result. The only self-citations are [14] as background for the unraveling definitions (which the paper re-derives) and [49,50] for AI-development methodology; neither is load-bearing for the physics. Section 9 explicitly scopes wall-clock, scale, and regime-dependence limitations, further indicating the claims are not constructed to be unfalsifiable. No circular step is present.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central trajectory-reduction claim rests only on standard open-quantum-systems formalism and the Pauli-Lindblad noise assumption; no ad hoc physical entities are introduced. One free tuning parameter (θ0) for the analog unraveling is unspecified. The Qiskit-Aer integration claim rests on an unverified software-behavior assumption.

free parameters (1)
  • analog kick angle θ0 (and rate λ) = unspecified
    Analog unraveling uses a two-point sign law θ∈{±θ0} with λ sin²θ0 = γ; θ0 is a free tuning choice that affects analog estimator variance. No value or optimality condition is given in the paper (Section 4.2), so the analog results in the regime map and crossover depend on an unstated parameter.
assumptions (4)
  • standard math Lindblad master equation governs the open-system dynamics
    Eq. (1); standard framework for Markovian noise.
  • standard math MCWF unraveling reproduces the exact density matrix in the average
    Standard result (Refs. [9-13]) assumed for all three unravelings; validated numerically in Section 5.
  • domain assumption Pauli-Lindblad noise has state-independent jump rates, enabling exact Poisson event sampling
    Section 4.3 and Section 9 limitations; the method and exact sampling are restricted to Pauli-Lindblad noise. Non-Pauli channels require norm-dependent sampling.
  • ad hoc to paper Qiskit-Aer canonicalizes user-supplied Kraus decompositions at apply time and the proposed preserve-unraveling flag would be sufficient to restore them
    Section 7.2-7.3; the diagnosis is based on inspecting qiskit-aer-gpu 0.15.1 and is not validated by an implementation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Variance-Reduced Trajectory Unravelings for GPU Noisy Quantum-Circuit Simulation: Characterization and a Qiskit-Aer Integration Gap." pith.science (2026). https://pith.science/paper/GVSN6HSG

@misc{pith2026260717678,
  author       = {Pith},
  title        = {Pith review of: Variance-Reduced Trajectory Unravelings for GPU Noisy Quantum-Circuit Simulation: Characterization and a Qiskit-Aer Integration Gap},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GVSN6HSG}},
  note         = {Machine review of arXiv:2607.17678}
}
abstract

Monte-Carlo trajectory (quantum-jump) methods are the practical route to simulating noisy quantum circuits once the exact density-matrix method is precluded by its $4^n$ memory cost. Their bottleneck is estimator variance: resolving one expectation value can demand thousands of trajectories. Recent tensor-network work shows that \emph{variance-reduced unravelings} -- projector and analog sampling -- sharply cut this variance, but only on CPU matrix-product-state backends, with no path into production tooling. We implement both unravelings on a \emph{GPU dense-statevector} trajectory engine and validate them against the exact density matrix (ideal-circuit fidelity $1-2.2\times10^{-16}$; $1/\sqrt{N}$ convergence; all unravelings unbiased to trace distance $<0.01$). On a single consumer GPU, projector unraveling reaches a target standard error with $20.8\times$ fewer trajectories than Qiskit-Aer's \texttt{batched\_shots\_gpu} at $n=10$, a factor that holds at $19$--$26\times$ across $n=8$--$20$. A regime map places analog sampling optimal at weak noise and projector at strong noise, crossing near $\gamma t\approx0.35$. We further report a systems finding: Qiskit-Aer applies noise at the \emph{channel} level and reconstructs a canonical Kraus decomposition at apply time, discarding any user-supplied unraveling, so variance-reduced unravelings cannot be delivered through its public API. Because Aer's Born-rule collapse machinery already exists, we specify a minimal change that would unlock the technique in production.

Figures

Figures reproduced from arXiv: 2607.17678 by the authors.

Figure 1
Figure 1. Where this work intervenes in the Qiskit-Aer stack. Left: the simulation pipeline. A quantum circuit and a NoiseModel carrying measured device physics (T1, T2, Pauli-Lindblad rates, crosstalk) are passed to the Qiskit-Aer backend (dashed box), which (i) represents the noise as a channel (Kraus/Lindblad), (ii) unravels that channel into sampled quantum-jump trajectories, and (iii) evolves each trajectory on a GPU den… view at source ↗
Figure 2
Figure 2. Correctness of the GPU trajectory engine. We reconstruct the density matrix ρ = 1 N P i |ψ (i) ⟩⟨ψ (i) | from N standard-unraveling trajectories and plot its trace distance to Qiskit-Aer’s exact density matrix result, on log–log axes; the x-axis is the number of trajectories N (2 × 102 to 6 × 104 ) and the y-axis is the trace distance (lower is more accurate). Blue circles are the measured error; the grey dashed lin… view at source ↗
Figure 3
Figure 3. Regime map of unraveling efficiency. The x-axis is the accumulated noise strength γt (larger = stronger noise); the y-axis is the per-trajectory variance of the estimator relative to the standard unraveling, so lower is better and the black dashed line at 1 is the standard baseline. Blue circles are projector, orange squares are analog (single dephasing channel, n = 12, N = 2 × 103 ). Both curves lie entirely below … view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Scaling of the trajectory-reduction factor. The x-axis is the qubit count n (state dimension 2 n, i.e. 256 to ∼ 106 amplitudes); the y-axis is how many times fewer trajectories the projector unraveling needs than standard to reach the same target standard error (SE ≤ 1…
Figure 5
Figure 5. Figure 5: Applicability map: coherent dynamics erodes the projector advantage. The x-axis is the angle θ of RY (θ) rotations interleaved between noise windows—a proxy for how much coherent “compute” the circuit performs relative to idle decoherence; the y-axis is the projector t…
Figure 6
Figure 6. Figure 6: Head-to-head against Qiskit-Aer. Bar height is the number of trajectories (equivalently shots) needed to reach the target standard error SE ≤ 10−2 on the observable 1 n P i Xi ; lower is better. Conditions: n = 10, strong noise γt = 1.5, exact mean e −2γt = 0.0498, all…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 4 linked inside Pith

  1. [1]

    Quantum computing in the nisq era and beyond.Quantum, 2:79, 2018

    John Preskill. Quantum computing in the nisq era and beyond.Quantum, 2:79, 2018

    2018

  2. [2]

    Thomas H¨ aner and Damian S. Steiger. 0.5 petabyte simulation of a 45-qubit quantum circuit. InProc. Int. Conf. for High Performance Computing, Networking, Storage and Analysis (SC), 2017

    2017

  3. [3]

    Benjamin

    Tyson Jones, Anna Brown, Ian Bush, and Simon C. Benjamin. Quest and high performance simulation of quantum computers.Scientific Reports, 9:10736, 2019

    2019

  4. [4]

    Gian Giacomo Guerreschi, Justin Hogaboam, Fabio Baruffa, and Nicolas P. D. Sawaya. Intel quantum simulator: a cloud-ready high-performance simulator of quantum circuits.Quantum Science and Technology, 5(3):034007, 2020

    2020

  5. [5]

    cuquantum sdk: A high-performance library for accelerating quantum science

    Harun Bayraktar et al. cuquantum sdk: A high-performance library for accelerating quantum science. InIEEE Int. Conf. on Quantum Computing and Engineering (QCE), 2023

    2023

  6. [6]

    Qiskit aer: A high performance simulator for quantum circuits, 2024

    Qiskit contributors. Qiskit aer: A high performance simulator for quantum circuits, 2024. https://github.com/Qiskit/qiskit-aer

    2024

  7. [7]

    Gambetta

    Kristan Temme, Sergey Bravyi, and Jay M. Gambetta. Error mitigation for short-depth quantum circuits.Physical Review Letters, 119(18):180509, 2017. 17

    2017

  8. [8]

    Minev, Abhinav Kandala, and Kristan Temme

    Ewout van den Berg, Zlatko K. Minev, Abhinav Kandala, and Kristan Temme. Probabilistic error cancellation with sparse pauli-lindblad models on noisy quantum processors.Nature Physics, 19:1116–1121, 2023

    2023

Show all 50 references
  1. [9]

    Wave-function approach to dissipative pro- cesses in quantum optics.Physical Review Letters, 68(5):580, 1992

    Jean Dalibard, Yvan Castin, and Klaus Mølmer. Wave-function approach to dissipative pro- cesses in quantum optics.Physical Review Letters, 68(5):580, 1992

    1992

  2. [10]

    Monte carlo wave-function method in quan- tum optics.Journal of the Optical Society of America B, 10(3):524–538, 1993

    Klaus Mølmer, Yvan Castin, and Jean Dalibard. Monte carlo wave-function method in quan- tum optics.Journal of the Optical Society of America B, 10(3):524–538, 1993

    1993

  3. [11]

    The quantum-jump approach to dissipative dynamics in quantum optics.Reviews of Modern Physics, 70(1):101, 1998

    Martin B Plenio and Peter L Knight. The quantum-jump approach to dissipative dynamics in quantum optics.Reviews of Modern Physics, 70(1):101, 1998

    1998

  4. [12]

    Oxford University Press, 2002

    Heinz-Peter Breuer and Francesco Petruccione.The Theory of Open Quantum Systems. Oxford University Press, 2002

    2002

  5. [13]

    Andrew J. Daley. Quantum trajectories and open many-body quantum systems.Advances in Physics, 63(2):77–149, 2014

    2014

  6. [14]

    Noisy quantum circuit simulation with the tensor jump method

    Authors of arXiv:2607.01323. Noisy quantum circuit simulation with the tensor jump method. arXiv preprint arXiv:2607.01323, 2026. Variance-aware tensor-network framework: projector and analog unravelings on MPS with local TDVP

    2026 arXiv

  7. [15]

    Efficient classical simulation of slightly entangled quantum computations.Phys- ical Review Letters, 91(14):147902, 2003

    Guifr´ e Vidal. Efficient classical simulation of slightly entangled quantum computations.Phys- ical Review Letters, 91(14):147902, 2003

    2003

  8. [16]

    Ignacio Cirac, David P´ erez-Garc ´ ıa, Norbert Schuch, and Frank Verstraete

    J. Ignacio Cirac, David P´ erez-Garc ´ ıa, Norbert Schuch, and Frank Verstraete. Matrix prod- uct states and projected entangled pair states: Concepts, symmetries, theorems.Reviews of Modern Physics, 93(4):045003, 2021

    2021

  9. [17]

    Locally purified density operators for noisy quantum circuits

    Authors of arXiv:2312.02854. Locally purified density operators for noisy quantum circuits. arXiv preprint arXiv:2312.02854, 2023

    2023 arXiv

  10. [18]

    Simulating noisy quantum circuits with matrix product density operators

    Song Cheng et al. Simulating noisy quantum circuits with matrix product density operators. Physical Review Research, 3:023005, 2021

    2021

  11. [19]

    Physical Review A, 110:032604, 2024

    Tomography-assisted noisy quantum circuit simulator using matrix product density operators. Physical Review A, 110:032604, 2024

    2024

  12. [20]

    Efficient classical simulation of noisy random quantum circuits in one dimension.Quantum, 4:318, 2020

    Kyungjoo Noh, Liang Jiang, and Bill Fefferman. Efficient classical simulation of noisy random quantum circuits in one dimension.Quantum, 4:318, 2020

    2020

  13. [21]

    Atlas: Hierarchical partitioning for quantum circuit simulation on gpus

    Mingkuan Xu et al. Atlas: Hierarchical partitioning for quantum circuit simulation on gpus. InProc. Int. Conf. for High Performance Computing, Networking, Storage and Analysis (SC), 2024

    2024

  14. [22]

    questab: Towards scalable quantum circuit simulation on multi-gpu using an extended stabilizer formalism

    quEStab authors. questab: Towards scalable quantum circuit simulation on multi-gpu using an extended stabilizer formalism. InProc. ACM Int. Conf. on Supercomputing (ICS), 2025

    2025

  15. [23]

    Wood, Jake Lishman, Julien Gacon, Simon Martiel, Paul D

    Ali Javadi-Abhari, Matthew Treinish, Kevin Krsulich, Christopher J. Wood, Jake Lishman, Julien Gacon, Simon Martiel, Paul D. Nation, Lev S. Bishop, Andrew W. Cross, Blake R. John- son, and Jay M. Gambetta. Quantum computing with Qiskit.arXiv preprint arXiv:2405.08810, 2024. 18

    2024 arXiv

  16. [24]

    IBM’s Qiskit tool chain: Working with and developing for real quantum computers

    Robert Wille, Rod Van Meter, and Yehuda Naveh. IBM’s Qiskit tool chain: Working with and developing for real quantum computers. InDesign, Automation & Test in Europe Conference (DATE), pages 1234–1240, 2019

    2019

  17. [25]

    Modeling and simulating the noisy behavior of near-term quantum computers.Physical Review A, 104(6):062432, 2021

    Konstantinos Georgopoulos, Clive Emary, and Paolo Zuliani. Modeling and simulating the noisy behavior of near-term quantum computers.Physical Review A, 104(6):062432, 2021

    2021

  18. [26]

    Nation and Matthew Treinish

    Paul D. Nation and Matthew Treinish. Suppressing quantum circuit errors due to system variability.PRX Quantum, 4(1):010327, 2023

    2023

  19. [27]

    Quantum supremacy using a programmable superconducting processor.Nature, 574:505–510, 2019

    Frank Arute, Kunal Arya, Ryan Babbush, et al. Quantum supremacy using a programmable superconducting processor.Nature, 574:505–510, 2019

    2019

  20. [28]

    Solving the sampling problem of the Sycamore quantum circuits.Physical Review Letters, 129(9):090502, 2022

    Feng Pan, Keyang Chen, and Pan Zhang. Solving the sampling problem of the Sycamore quantum circuits.Physical Review Letters, 129(9):090502, 2022

    2022

  21. [29]

    Markov and Yaoyun Shi

    Igor L. Markov and Yaoyun Shi. Simulating quantum computation by contracting tensor networks.SIAM Journal on Computing, 38(3):963–981, 2008

    2008

  22. [30]

    Hyper-optimized tensor network contraction.Quantum, 5:410, 2021

    Johnnie Gray and Stefanos Kourtis. Hyper-optimized tensor network contraction.Quantum, 5:410, 2021

    2021

  23. [31]

    C´ orcoles, Antonio Mezzacapo, Jerry M

    Abhinav Kandala, Kristan Temme, Antonio D. C´ orcoles, Antonio Mezzacapo, Jerry M. Chow, and Jay M. Gambetta. Error mitigation extends the computational reach of a noisy quantum processor.Nature, 567:491–495, 2019

    2019

  24. [32]

    Benjamin, and Xiao Yuan

    Suguru Endo, Zhenyu Cai, Simon C. Benjamin, and Xiao Yuan. Hybrid quantum-classical algo- rithms and quantum error mitigation.Journal of the Physical Society of Japan, 90(3):032001, 2021

    2021

  25. [33]

    Evidence for the utility of quantum computing before fault tolerance.Nature, 618:500– 505, 2023

    Youngseok Kim, Andrew Eddins, Sajant Anand, Ken Xuan Wei, Ewout van den Berg, Sami Rosenblatt, Hasan Nayfeh, Yantao Wu, Michael Zaletel, Kristan Temme, and Abhinav Kan- dala. Evidence for the utility of quantum computing before fault tolerance.Nature, 618:500– 505, 2023

    2023

  26. [34]

    Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C

    M. Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C. Benjamin, Suguru Endo, Keisuke Fujii, Jarrod R. McClean, Kosuke Mitarai, Xiao Yuan, Lukasz Cincio, and Patrick J. Coles. Variational quantum algorithms.Nature Reviews Physics, 3:625–644, 2021

    2021

  27. [35]

    Kottmann, Tim Menke, et al

    Kishor Bharti, Alba Cervera-Lierta, Thi Ha Kyaw, Tobias Haug, Sumner Alperin-Lea, Abhinav Anand, Matthias Degroote, Hermanni Heimonen, Jakob S. Kottmann, Tim Menke, et al. Noisy intermediate-scale quantum algorithms.Reviews of Modern Physics, 94(1):015004, 2022

    2022

  28. [36]

    Improved simulation of stabilizer circuits.Physical Review A, 70(5):052328, 2004

    Scott Aaronson and Daniel Gottesman. Improved simulation of stabilizer circuits.Physical Review A, 70(5):052328, 2004

    2004

  29. [37]

    Improved classical simulation of quantum circuits dominated by Clifford gates.Physical Review Letters, 116(25):250501, 2016

    Sergey Bravyi and David Gosset. Improved classical simulation of quantum circuits dominated by Clifford gates.Physical Review Letters, 116(25):250501, 2016

    2016

  30. [38]

    The density-matrix renormalization group in the age of matrix product states.Annals of Physics, 326(1):96–192, 2011

    Ulrich Schollw¨ ock. The density-matrix renormalization group in the age of matrix product states.Annals of Physics, 326(1):96–192, 2011

    2011

  31. [39]

    Miles Stoudenmire, and Xavier Waintal

    Yiqing Zhou, E. Miles Stoudenmire, and Xavier Waintal. What limits the simulation of quan- tum computers?Physical Review X, 10(4):041038, 2020. 19

    2020

  32. [40]

    Ignacio Cirac

    Frank Verstraete, Juan Jos´ e Garc ´ ıa-Ripoll, and J. Ignacio Cirac. Matrix product density operators: Simulation of finite-temperature and dissipative systems.Physical Review Letters, 93(20):207204, 2004

    2004

  33. [41]

    Mixed-state dynamics in one-dimensional quantum lattice systems: A time-dependent superoperator renormalization algorithm.Physical Review Letters, 93(20):207205, 2004

    Michael Zwolak and Guifr´ e Vidal. Mixed-state dynamics in one-dimensional quantum lattice systems: A time-dependent superoperator renormalization algorithm.Physical Review Letters, 93(20):207205, 2004

    2004

  34. [42]

    Cache blocking technique to large scale quantum computing simulation on supercomputers

    Jun Doi and Hiroshi Horii. Cache blocking technique to large scale quantum computing simulation on supercomputers. InIEEE Int. Conf. on Quantum Computing and Engineering (QCE), 2020

    2020

  35. [43]

    Efficient techniques to gpu accelerations of multi-shot quantum computing simulations.arXiv preprint arXiv:2308.03399, 2023

    Authors of arXiv:2308.03399. Efficient techniques to gpu accelerations of multi-shot quantum computing simulations.arXiv preprint arXiv:2308.03399, 2023

    2023 arXiv

  36. [44]

    Running qiskit on rocm platform.EPJ Web of Conferences, 295:11022, 2024

    Yujiang Bi, Shun Xu, and Yunheng Ma. Running qiskit on rocm platform.EPJ Web of Conferences, 295:11022, 2024

    2024

  37. [45]

    Multi-gpu quantum circuit simulation and the impact of network performance

    Brown et al. Multi-gpu quantum circuit simulation and the impact of network performance. arXiv preprint arXiv:2511.14664, 2025

    2025

  38. [46]

    On the generators of quantum dynamical semigroups.Communications in Mathematical Physics, 48(2):119–130, 1976

    G¨ oran Lindblad. On the generators of quantum dynamical semigroups.Communications in Mathematical Physics, 48(2):119–130, 1976

    1976

  39. [47]

    Vittorio Gorini, Andrzej Kossakowski, and E. C. G. Sudarshan. Completely positive dynamical semigroups ofn-level systems.Journal of Mathematical Physics, 17(5):821–825, 1976

    1976

  40. [48]

    Nielsen and Isaac L

    Michael A. Nielsen and Isaac L. Chuang.Quantum Computation and Quantum Information. Cambridge University Press, 10th anniversary edition, 2010

    2010

  41. [49]

    Optimization of MuMax3 by using Claude Code: A CUDA-graph-based case study in AI-assisted performance engineering.Journal of Magnetics, 31(2):204–213, 2026

    Chun-Yeol You. Optimization of MuMax3 by using Claude Code: A CUDA-graph-based case study in AI-assisted performance engineering.Journal of Magnetics, 31(2):204–213, 2026

    2026

  42. [50]

    AI-assisted systematic optimization of the LAMMPS molecular dynamics simulator via large language models

    Chun-Yeol You. AI-assisted systematic optimization of the LAMMPS molecular dynamics simulator via large language models. 2026. under review. 20

    2026

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.