Pith. sign in

REVIEW 3 major objections 5 minor 5 cited by

Robust quantum computational advantage with programmable 3050-photon Gaussian boson sampling

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper reports a programmable 1024-input, 8176-mode Gaussian boson sampler reaching 3050 detected photons and projects over 10^42 years of classical simulation time on a top supercomputer.

desk verdict A genuinely large and well-validated GBS experiment whose headline speedup depends on a fragile extrapolation; the hardware is the real contribution, the 10^42-year claim is not yet established. read the letter →

arxiv 2508.09092 v3 pith:XC7DGW77 submitted 2025-08-12 quant-ph

Hua-Liang Liu , Hao Su , Si-Qiu Gong , Yi-Chao Gu , Hao-Yang Tang , Meng-Hao Jia , Qian Wei , Yukun Song
show 29 more authors
Dongzhou Wang Mingyang Zheng Faxi Chen Libo Li Siyu Ren Xuezhi Zhu Meihong Wang Yaojian Chen Yanfei Liu Longsheng Song Pengyu Yang Junshi Chen Hong An Lei Zhang Lin Gan Guangwen Yang Jia-Min Xu Yu-Ming He Hui Wang Han-Sen Zhong Ming-Cheng Chen Xiao Jiang Li Li Nai-Le Liu Yu-Hao Deng Xiao-Long Su Qiang Zhang Chao-Yang Lu Jian-Wei Pan
This is my paper · ORCID
classification quant-ph
keywords quantumcomputationaladvantageGaussianbosonsamplingphotonlossmatrixproductstatesimulationsqueezedstatesprogrammablephotonicprocessorspatial-temporalencodingtensornetwork
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 reports a Gaussian boson sampling experiment at a scale that current classical algorithms cannot reproduce: 1024 squeezed states are sent through a programmable 8176-mode photonic circuit, and the largest runs register up to 3050 photon-detection events. The central claim is that this output distribution lies beyond the reach of the strongest known classical simulation method for lossy GBS, the matrix-product-state (MPS) algorithm, so the experiment establishes quantum computational advantage in the presence of photon loss. The quantitative basis is a cost projection: on a current top supercomputer, reproducing the samples with the MPS algorithm would take more than $10^{42}$ years, whereas the quantum processor produces one sample in 25.6 microseconds—a speedup exceeding $10^{54}$. If the projection is right, photon loss no longer provides an automatic classical simulation shortcut, and the advantage is achieved with programmable, reconfigurable hardware.

What carries the argument

The argument rests on two coupled pieces. (1) The experimental machinery: a spatial-temporal hybrid encoding in which three 16-mode interferometers are connected by two delay-loop arrays, giving cubic connectivity ($16^3 = 4096$ paths per input) while physical resources grow only linearly, producing 8176 coupled output modes. (2) The classical-cost identity: the MPS simulation runtime $T_{\mathrm{MPS}} = O(M d\, \chi(\varepsilon)^2\, 2^{N_{\mathrm{eff}}/2})$, with bond dimension $\chi(\varepsilon)$ fitted to $\chi(\varepsilon)=A(\ln(1/\varepsilon))^n$ over the tractable error range and extrapolated to $\varepsilon_T=0.05$, and $N_{\mathrm{eff}}$ computed from the calibrated squeezing and int

What would settle it

Run an independent MPS implementation on a surrogate instance with the same $N_{\mathrm{eff}}\approx 113$ but fewer modes, measure the bond dimension needed for truncation errors down to $\varepsilon\approx 0.02$, and compare with the fitted $A(\ln(1/\varepsilon))^n$ curve; a deviation that makes $\chi(0.05)$ much smaller than $8\times10^{21}$ would falsify the extrapolation. A direct check is also possible with the released raw data: any classical sampler that reproduces the L1024 two- and three-order correlation functions within the experimental $\Delta K \approx 0.02$ on currently available

Watch

Extended reading notes

Core claim

Jiuzhang 4.0 implements 1024 single-mode squeezed states passing through three 16-mode interferometers connected by two fiber-delay loop arrays, forming a hybrid spatial-temporal circuit of 8176 output modes with cubic connectivity. The measured overall system efficiency is 51%, with 92% source efficiency and 93% detector efficiency, and the output photon-number distribution matches a lossy, partially distinguishable GBS ground-truth model. Against the squashed-state, greedy, independent-pairs-and-singles, and treewidth samplers, the experimental samples pass the Bayesian test and reproduce the theoretical two- and three-order correlation functions within $\Delta K \approx 0.02$. The decisiv

Load-bearing premise

The $>10^{42}$-year estimate depends on the fitted power-of-log scaling $\chi(\varepsilon)=A(\ln(1/\varepsilon))^n$ staying valid from the tractable error range down to $\varepsilon_T=0.05$, about twenty orders of magnitude beyond the fitted data; if the required bond dimension grows faster there, the projection collapses.

Editorial extensions

If this is right

  • Photon loss stops being a reliable route to classical simulability: at fixed transmission, growing the number of inputs pushes the MPS algorithm out of its efficient regime, so loss-aware spoofing cannot keep pace.
  • The demonstrated programmability means the same 8176-mode processor can be reconfigured to sample from many different unitaries, so the advantage claim is not tied to a single fixed circuit.
  • The architecture's linear-resource, cubic-connectivity scaling gives a concrete path to even larger experiments; more modes and higher squeezing should widen the gap over classical simulation.
  • The validation toolbox—subsystem Bayesian tests plus multi-order correlation distances—provides a template for future photonic advantage experiments to rule out several classical spoofing families at once.
  • If the cost projection holds, classical simulation of lossy GBS at this scale requires a new algorithmic idea rather than more supercomputer time.

Reading between the lines

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

  • The $10^{54}$ factor is a point estimate built from one extrapolated curve; even if the true bond dimension were orders of magnitude smaller than fitted, the qualitative conclusion—that this scale is out of reach for current MPS-based simulation—would probably survive. That robustness is an editorial inference, not something the paper quantifies.
  • The same hybrid spatial-temporal architecture is, in effect, a generator of large entangled qumode cluster states, so a natural testable extension is to use Jiuzhang 4.0 to prepare and verify cluster states with thousands of modes.
  • Because the paper validates against specific spoofing families (squashed, greedy, IPS, treewidth, MPS), a future classical sampler that targets only the validated moments could still evade these tests; extending the certification to higher-order correlation functions is an obvious next check.
  • One could reuse the released dataset to benchmark future classical samplers directly against the experimental samples, turning the paper's data into a persistent testbed for the next round of quantum-classical competition.
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 manuscript reports a large-scale Gaussian boson sampling (GBS) experiment, Jiuzhang 4.0, with 1024 squeezed inputs, 8176 output modes, and up to 3050 photon-detection events. The authors validate the data against classical mockups using photon-number distributions, Bayesian tests, and low-order correlation functions, and they compare against the matrix-product-state (MPS) algorithm of Ref. [23]. From truncation-error data at small scale, they fit χ(ε)=A(ln(1/ε))^n and extrapolate to ε_T=0.05, obtaining χ>8×10^21 for the largest instance. Combining this with Eq. (1), they estimate that the MPS algorithm would need >10^42 years on the El Capitan supercomputer, corresponding to a speedup >10^54 relative to their 25.6 μs sampling time. The paper claims this establishes a robust quantum computational advantage.

Significance. If the central quantitative claim is fully established, this would be a landmark experiment: it would be the largest GBS implementation by a wide margin, with high-efficiency sources and a programmable hybrid spatial-temporal interferometer, and the first to confront the MPS loss-based simulation algorithm at scale. The validation machinery (Bayesian tests, correlation benchmarking, and direct MPS simulations for smaller instances) is thoughtful and mostly convincing. Strengths include the open data statement, the detailed characterization of loss, indistinguishability, and phase stability, and the two-pronged MPS exclusion strategy. However, the headline speedup is contingent on an extrapolation over roughly 17 orders of magnitude in bond dimension with no uncertainty quantification. The significance as stated is therefore not yet at the level the paper claims; the result would be better described as a strong evidence claim conditional on an assumed scaling law, rather than a categorical exclusion of all classical simulation possibilities.

major comments (3)
  1. [Eq. (1), Fig. 4a] The lower bound χ>8×10^21 is obtained by fitting χ(ε)=A(ln(1/ε))^n to data in Fig. S19, then extrapolating from the tractable regime (ε≈0.07–0.16 for S64/M256; ε≈1 for L1024) to ε_T=0.05. The paper gives no error bars on A and n, no bootstrap, and no test of alternative functional forms (e.g., χ∝ε^{-c}). Because Eq. (1) scales as χ^2, a modest change in the fitted exponent changes the projected runtime by many orders of magnitude. Please provide uncertainty bands, a sensitivity analysis, or a conservative upper bound that does not rely on the assumed functional form.
  2. [Fig. 3c and 'Targeting ideal ground truth distribution'] For the L1024 group, the accessible truncation errors remain near unity (ε=0.9999 at χ=10^4; still close to 1 at χ=2×10^5). The scaling law is therefore effectively unmeasured in the regime relevant to the claim and is transferred from S64/M256, which have different sizes, effective photon numbers, and connectivity. The paper should justify this transfer, for example by collapsing data across groups in terms of a predicted scaling variable, or explicitly label the L1024 extrapolation as an assumption rather than a fit.
  3. [End of MPS section (after Fig. 3f)] The statement 'These results exclude all possibility that the MPS algorithm could reproduce the ground truth distribution better than our experiment' is stronger than the evidence. The first branch is an extrapolated estimate for an ideal-targeting MPS; the second branch tests only one interpolated family with reduced N_eff at fixed χ=10^4. This does not exclude other MPS configurations, algorithmic improvements, or a possible crossover in the scaling law. I recommend replacing 'exclude all possibility' with a statement limited to the tested and extrapolated scenarios, or to the specific MPS variant of Ref. [23] used here.
minor comments (5)
  1. [Abstract / Introduction] Typo: 'favourable scaling low' should be 'favourable scaling law'.
  2. [Fig. 1a caption] Typo: 'spacial modes' should be 'spatial modes'.
  3. [Section on runtime estimate] Grammar: 'Choosing a conservative truncation error of εT = 0.05 yields' should read 'Choosing ... yields' or 'With ... yields'.
  4. [General] There are several awkward hyphenated line breaks in the text (e.g., 'indistinguisha-bility'); a copyedit pass would improve readability.
  5. [Fig. 2b caption] The caption says 'photon number distribution of the experimental results, the ground-truth theory and classical mockups'; please specify the sample sizes and normalization to help reproducibility.

Circularity Check

1 steps flagged · score 6.0 of 10

The advertised 10^42-year MPS cost is obtained by evaluating a least-squares fit χ(ε)=A(ln(1/ε))^n at ε_T=0.05, so the headline QCA 'prediction' reduces by construction to the fitted curve.

  1. fitted input called prediction [Section 'Finally, we benchmark the QCA...' following Eq. (1), Fig. 4a]
    "Following Ref. [23] which shows that χ(ε) = O(polylog(1/ε)), we find our data to be well described by χ(ε) = A(ln(1/ε))^n (Fig. S19), with coefficients A, ndetermined by least-squares fitting. Choosing a conservative truncation error ofεT = 0.05 yields a lower bound of the required χ > 8 × 10^21 for our largest L1024 data set"

    The 'predicted' required bond dimension χ>8×10^21 is not computed from an independent physical bound; it is the fitted function χ(ε)=A(ln(1/ε))^n evaluated at ε_T=0.05. The constants A and n are least-squares fits to data in the tractable regime (ε≈0.07–0.16 for S64/M256), and the paper itself states 'we extrapolate from the tractable regime to estimate the required χ.' Thus the central QCA runtime estimate inherits the fitted coefficients by construction, and any breakdown of the assumed polylog scaling directly becomes an error in the claimed 10^42-year cost.

full rationale

The paper's main QCA claim—that EI Capitan would need >10^42 years to construct the MPS tensor network—is computed from Eq. (1) with Neff exact and χ obtained by extrapolating the fitted law χ(ε)=A(ln(1/ε))^n to ε_T=0.05. This is a fitted input called a prediction: the output χ>8×10^21 is the fitted curve evaluated at the target, not an independently established quantity. The categorical statement 'These results exclude all possibility that the MPS algorithm could reproduce the ground truth distribution better than our experiment' rests on this extrapolation, which is not validated in the intractable regime and has no reported error bars or alternative-form checks. This is the load-bearing circular element. The rest of the validation—Bayesian tests and correlation benchmarks against squashed-state, greedy, IPS, and treewidth samplers—is self-contained and does not reduce to the fitted parameters, and the ground-truth model calibration with measured loss and indistinguishability is standard practice rather than circular. Self-citations appear (e.g., Ref. [41] for the runtime prefactor) but the main issue is the extrapolated bond-dimension fit. Score 6 reflects that the central QCA speedup number reduces by construction to the fitted scaling law.

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

The central cost projection hangs on the MPS algorithm's runtime model and a fitted extrapolation, while the device validation relies on a calibrated lossy-GBS ground truth. No new physical entities are introduced.

free parameters (5)
  • A (prefactor in χ-ε scaling law) = least-squares fit (value in Fig. S19)
    Used to extrapolate required bond dimension at ε_T=0.05; central to runtime estimate.
  • n (exponent in χ-ε scaling law) = least-squares fit (value in Fig. S19)
    Same as above.
  • ε_T (target truncation error) = 0.05
    Chosen as 'conservative' target; directly sets the required χ and the 10^42-year estimate.
  • transmission scaling factor (Fig. 3d) = not stated
    Scanned to yield favorable truncation error 0.01 for the adapted MPS; post hoc selection.
  • squeezing parameters r (0.9-1.8) = up to 1.8
    Experimental controls that set N_eff; calibrated but not fitted to claims.
assumptions (5)
  • domain assumption The MPS algorithm of Ref. [23] correctly computes lossy GBS statistics and its runtime follows Eq. (1) with O(χ^2) scaling.
    The entire cost projection rests on this algorithm's correctness and complexity bound.
  • domain assumption The scaling χ(ε)=O(polylog(1/ε)) from Ref. [23] remains valid, and the specific functional form A(ln(1/ε))^n describes the data in the extrapolated regime.
    Invoked to extrapolate χ from fits at large ε to the target ε_T.
  • domain assumption The experiment's ground truth is a lossy, partially distinguishable GBS model with calibrated efficiency and indistinguishability.
    Validation against this model assumes it accurately represents the device and that no unmodeled effects (e.g., detector crosstalk, phase drift) alter the distribution.
  • domain assumption The classical component W in the loss decomposition can be sampled efficiently.
    Standard result of Ref. [23], needed for the MPS simulator's speed.
  • domain assumption No classical algorithm more efficient than the MPS algorithm exists for simulating this experiment.
    The paper claims to outperform 'all classical spoofing algorithms' but only tests five known samplers; the absence of a faster algorithm is an assumption, not a proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Robust quantum computational advantage with programmable 3050-photon Gaussian boson sampling." pith.science (2026). https://pith.science/paper/XC7DGW77

@misc{pith2026250809092,
  author       = {Pith},
  title        = {Pith review of: Robust quantum computational advantage with programmable 3050-photon Gaussian boson sampling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XC7DGW77}},
  note         = {Machine review of arXiv:2508.09092}
}
abstract

The creation of large-scale, high-fidelity quantum computers is not only a fundamental scientific endeavour in itself, but also provides increasingly robust proofs of quantum computational advantage (QCA) in the presence of unavoidable noise and the dynamic competition with classical algorithm improvements. To overcome the biggest challenge of photon-based QCA experiments, photon loss, we report new Gaussian boson sampling (GBS) experiments with 1024 high-efficiency squeezed states injected into a hybrid spatial-temporal encoded, 8176-mode, programmable photonic quantum processor, Jiuzhang 4.0, which produces up to 3050 photon detection events. Our experimental results outperform all classical spoofing algorithms, particularly the matrix product state (MPS) method, which was recently proposed to utilise photon loss to reduce the classical simulation complexity of GBS. Using the state-of-the-art MPS algorithm on the most powerful supercomputer EI Capitan, it would take > $10^{42}$ years to construct the required tensor network for simulation, while our Jiuzhang 4.0 quantum computer takes 25.6 $\mu$s to produce a sample. This work establishes a new frontier of QCA and paves the way to fault-tolerant photonic quantum computing hardware.

Figures

Figures reproduced from arXiv: 2508.09092 by the authors.

Figure 1
Figure 1. b traces evolution of the photon population inside the circuit. Our cascaded architecture harnesses both spatial and temporal interference, so that information encoded in one degree of freedom can be transformed with the other as the photons propagate. Step 1: After U1 the photons are delo￾calized over all 16 sptial modes. Step 2: The subsequent array of delay loops L1 causes the photons to populate the nearest 16 t… view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Proof of Hiding Conjecture in Gaussian Boson Sampling

    quant-ph 2025-08 conditional novelty 8.0 of 10

    The top-left N by N submatrix of an M by M circular orthogonal ensemble random matrix, scaled by sqrt(M), converges to a complex symmetric Gaussian matrix in total variation distance for N much smaller than sqrt(M).

  2. Boosting Gaussian Boson Sampling using Optical Parametric Amplification Networks

    quant-ph 2025-11 conditional novelty 6.0 of 10

    An OPA-based nonlinear interferometer is proposed that keeps GBS entanglement linear in the number of modes under realistic photon loss, which the authors argue prevents efficient classical simulation.

  3. Matrix product state approach to lossy boson sampling and noisy IQP sampling

    quant-ph 2025-10 accept novelty 6.0 of 10

    Lossy boson sampling and noisy IQP sampling are classically simulable with matrix product states, with the same known noise thresholds and accuracy controlled by bond dimension.

  4. Quantum latent distributions in deep generative models

    cs.LG 2025-08 conditional novelty 6.0 of 10

    Quantum latent distributions from boson samplers are shown in theory to expand the output distribution class of invertible Lipschitz generators, and in GAN benchmarks on QM9 to beat Gaussian, Bernoulli, and distinguis...

  5. Configurable photonic simulator for quantum field dynamics

    quant-ph 2025-06 conditional novelty 6.0 of 10

    The Optical Time Algorithm maps free scalar quantum field Hamiltonians onto a fixed optical circuit in which time evolution is carried entirely by a single phase-shift layer, and reduces to Gaussian boson sampling for...

Reference graph

Works this paper leans on

42 extracted references · 36 canonical work pages · cited by 5 Pith papers

  1. [23]

    C. Oh, M. Liu, Y . Alexeev, B. Fefferman, and L. Jiang, Classi- cal algorithm for simulating experimental gaussian boson sam- pling, Nature Physics 20, 1461 (2024)

    work page 2024

  2. [1]

    Arute, K

    F. Arute, K. Arya, R. Babbush, D. Bacon, J. C. Bardin, R. Barends, R. Biswas, S. Boixo, F. G. Brandao, D. A. Buell, et al. , Quantum supremacy using a programmable supercon- ducting processor, Nature 574, 505 (2019)

    work page 2019

  3. [2]

    Zhong, H

    H.-S. Zhong, H. Wang, Y .-H. Deng, M.-C. Chen, L.-C. Peng, Y .-H. Luo, J. Qin, D. Wu, X. Ding, Y . Hu,et al., Quantum com- putational advantage using photons, Science 370, 1460 (2020)

    work page 2020

  4. [3]

    Wu, W.-S

    Y . Wu, W.-S. Bao, S. Cao, F. Chen, M.-C. Chen, X. Chen, T.-H. Chung, H. Deng, Y . Du, et al., Strong quantum computational advantage using a superconducting quantum processor, Physi- cal Review Letters 127, 180501 (2021)

    work page 2021

  5. [4]

    Zhong, Y .-H

    H.-S. Zhong, Y .-H. Deng, J. Qin, H. Wang, M.-C. Chen, L.-C. Peng, Y .-H. Luo, D. Wu, S.-Q. Gong, H. Su, et al. , Phase-programmable gaussian boson sampling using stimu- lated squeezed light, Physical review letters 127, 180502 (2021)

    work page 2021

  6. [5]

    L. S. Madsen, F. Laudenbach, M. F. Askarani, F. Rortais, T. Vincent, J. F. Bulmer, F. M. Miatto, L. Neuhaus, L. G. Helt, M. J. Collins, et al., Quantum computational advantage with a programmable photonic processor, Nature 606, 75 (2022). 7

    work page 2022

  7. [6]

    Deng, Y .-C

    Y .-H. Deng, Y .-C. Gu, H.-L. Liu, S.-Q. Gong, H. Su, Z.- J. Zhang, H.-Y . Tang, M.-H. Jia, J.-M. Xu, M.-C. Chen, et al., Gaussian boson sampling with pseudo-photon-number- resolving detectors and quantum computational advantage, Physical review letters 131, 150601 (2023)

    work page 2023

  8. [7]

    Morvan, B

    A. Morvan, B. Villalonga, X. Mi, S. Mandr `a, A. Bengtsson, P. Klimov, Z. Chen, S. Hong, C. Erickson, I. Drozdov, et al., Phase transitions in random circuit sampling, Nature 634, 328 (2024)

    work page 2024

Show all 42 references
  1. [8]

    D. Gao, D. Fan, C. Zha, J. Bei, G. Cai, J. Cai, S. Cao, F. Chen, J. Chen, K. Chen, X. Chen, X. Chen, Z. Chen, Z. Chen, Z. Chen, W. Chu, H. Deng, Z. Deng, P. Ding, X. Ding, Z. Ding, S. Dong, et al. , Establishing a New Benchmark in Quantum Compu- tational Advantage with 105-qub...

    2025

  2. [9]

    D. A. Abanin, R. Acharya, L. Aghababaie-Beni, G. Aigeldinger, A. Ajoy, R. Alcaraz, I. Aleiner, T. I. Andersen, M. Ansmann, F. Arute, K. Arya, A. Asfaw, N. As- trakhantsev, J. Atalaya, et al., Constructive interference at the edge of quantum ergodic dynamics (2025), arXiv:2506....

    2025 arXiv

  3. [10]

    Aaronson and A

    S. Aaronson and A. Arkhipov, The Computational Complexity of Linear Optics, Theory of Computing 9, 143 (2013)

    2013

  4. [11]

    Bouland, B

    A. Bouland, B. Fefferman, C. Nirkhe, and U. Vazirani, On the complexity and verification of quantum random circuit sam- pling, Nature Physics 15, 159 (2019)

    2019

  5. [12]

    Pan and P

    F. Pan and P. Zhang, Simulation of Quantum Circuits Using the Big-Batch Tensor Network Method, Physical Review Letters 128, 030501 (2022)

    2022

  6. [13]

    F. Pan, K. Chen, and P. Zhang, Solving the Sampling Problem of the Sycamore Quantum Circuits, Physical Review Letters 129, 090502 (2022)

    2022

  7. [14]

    R. Fu, Z. Su, H.-S. Zhong, X. Zhao, J. Zhang, F. Pan, P. Zhang, X. Zhao, M.-C. Chen, C.-Y . Lu, J.-W. Pan, Z. Pei, X. Zhang, and W. Ouyang, Surpassing sycamore: Achieving energetic su- periority through system-level circuit simulation, in Proceed- ings of the International Con...

    2024

  8. [15]

    F. Pan, H. Gu, L. Kuang, B. Liu, and P. Zhang, Efficient Quan- tum Circuit Simulation by Tensor Network Methods on Mod- ern GPUs, ACM Transactions on Quantum Computing 5, 26:1 (2024)

    2024

  9. [16]

    Zhao, H.-S

    X.-H. Zhao, H.-S. Zhong, F. Pan, Z.-H. Chen, R. Fu, Z. Su, X. Xie, C. Zhao, P. Zhang, W. Ouyang, C.-Y . Lu, J.-W. Pan, and M.-C. Chen, Leapfrogging Sycamore: Harnessing 1432 GPUs for 7× faster quantum random circuit sampling, National Sci- ence Review 12, nwae317 (2025)

    2025

  10. [17]

    H. Qi, D. J. Brod, N. Quesada, and R. Garc ´ıa-Patr´on, Regimes of classical simulability for noisy gaussian boson sampling, Physical Review Letters 124, 100502 (2020)

    2020

  11. [18]

    Villalonga, M

    B. Villalonga, M. Y . Niu, L. Li, H. Neven, J. C. Platt, V . N. Smelyanskiy, and S. Boixo, Efficient approximation of experimental gaussian boson sampling, arXiv preprint arXiv:2109.11525 (2021)

    2021 arXiv

  12. [19]

    Quesada, R

    N. Quesada, R. S. Chadwick, B. A. Bell, J. M. Arrazola, T. Vin- cent, H. Qi, and R. Garc´ıa-Patr´on, Quadratic speed-up for sim- ulating gaussian boson sampling, PRX Quantum 3, 010306 (2022)

    2022

  13. [20]

    J. F. Bulmer, B. A. Bell, R. S. Chadwick, A. E. Jones, D. Moise, A. Rigazzi, J. Thorbecke, U.-U. Haus, T. Van Vaerenbergh, R. B. Patel,et al., The boundary for quantum advantage in gaus- sian boson sampling, Science advances 8, eabl9236 (2022)

    2022

  14. [21]

    C. Oh, Y . Lim, B. Fefferman, and L. Jiang, Classical simulation of boson sampling based on graph structure, Physical Review Letters 128, 190501 (2022)

    2022

  15. [22]

    C. Oh, L. Jiang, and B. Fefferman, Spoofing cross-entropy mea- sure in boson sampling, Physical Review Letters 131, 010401 (2023)

    2023

  16. [24]

    C. S. Hamilton, R. Kruse, L. Sansoni, S. Barkhofen, C. Sil- berhorn, and I. Jex, Gaussian boson sampling, Physical review letters 119, 170501 (2017)

    2017

  17. [25]

    Quesada, J

    N. Quesada, J. M. Arrazola, and N. Killoran, Gaussian bo- son sampling using threshold detectors, Physical Review A 98, 062322 (2018)

    2018

  18. [26]

    J. M. Arrazola, T. R. Bromley, and P. Rebentrost, Quantum ap- proximate optimization with Gaussian boson sampling, Physi- cal Review A 98, 012322 (2018)

    2018

  19. [27]

    J. M. Arrazola and T. R. Bromley, Using Gaussian Boson Sam- pling to Find Dense Subgraphs, Physical Review Letters 121, 030503 (2018)

    2018

  20. [28]

    Banchi, M

    L. Banchi, M. Fingerhuth, T. Babej, C. Ing, and J. M. Arra- zola, Molecular docking with Gaussian Boson Sampling, Sci- ence Advances 6, eaax1950 (2020)

    2020

  21. [29]

    J. Huh, G. G. Guerreschi, B. Peropadre, J. R. McClean, and A. Aspuru-Guzik, Boson sampling for molecular vibronic spec- tra, Nature Photonics 9, 615 (2015)

    2015

  22. [30]

    Jahangiri, J

    S. Jahangiri, J. Miguel Arrazola, N. Quesada, and A. Delgado, Quantum algorithm for simulating molecular vibrational excita- tions, Physical Chemistry Chemical Physics 22, 25528 (2020)

    2020

  23. [31]

    Jahangiri, J

    S. Jahangiri, J. M. Arrazola, and A. Delgado, Quantum Algo- rithm for Simulating Single-Molecule Electron Transport, The Journal of Physical Chemistry Letters 12, 1256 (2021)

    2021

  24. [32]

    Deng, S.-Q

    Y .-H. Deng, S.-Q. Gong, Y .-C. Gu, Z.-J. Zhang, H.-L. Liu, H. Su, H.-Y . Tang, J.-M. Xu, M.-H. Jia, M.-C. Chen, H.-S. Zhong, H. Wang, J. Yan, Y . Hu, J. Huang, W.-J. Zhang, H. Li, X. Jiang, L. You, Z. Wang, L. Li, N.-L. Liu, C.-Y . Lu, and J.-W. Pan, Solving Graph Problems Us...

    2023 doi

  25. [33]

    Shang, H.-S

    Z.-X. Shang, H.-S. Zhong, Y .-K. Zhang, C.-C. Yu, X. Yuan, C.- Y . Lu, J.-W. Pan, and M.-C. Chen, Boson sampling enhanced quantum chemistry (2024), arXiv:2403.16698 [quant-ph]

    2024 arXiv

  26. [34]

    Cimini, M

    V . Cimini, M. M. Sohoni, F. Presutti, B. K. Malia, S.-Y . Ma, R. Yanagimoto, T. Wang, T. Onodera, L. G. Wright, and P. L. McMahon, Large-scale quantum reservoir computing using a gaussian boson sampler (2025), arXiv:2505.13695 [quant-ph]

    2025 arXiv

  27. [35]

    Gong, M.-C

    S.-Q. Gong, M.-C. Chen, H.-L. Liu, H. Su, Y .-C. Gu, H.- Y . Tang, M.-H. Jia, Y .-H. Deng, Q. Wei, H. Wang, H.-S. Zhong, X. Jiang, L. Li, N.-L. Liu, C.-Y . Lu, and J.-W. Pan, Enhanced image recognition using gaussian boson sampling (2025), arXiv:2506.19707 [quant-ph]

    2025 arXiv

  28. [36]

    Tzitrin, J

    I. Tzitrin, J. E. Bourassa, N. C. Menicucci, and K. K. Saba- pathy, Progress towards practical qubit computation using ap- proximate gottesman-kitaev-preskill codes, Phys. Rev. A 101, 032315 (2020)

    2020

  29. [37]

    Larsen, J

    M. Larsen, J. Bourassa, S. Kocsis, J. Tasker, R. Chadwick, C. Gonz ´alez-Arciniegas, J. Hastrup, C. Lopetegui-Gonz ´alez, F. Miatto, A. Motamedi, et al., Integrated photonic source of gottesman–kitaev–preskill qubits, Nature , 1 (2025)

    2025

  30. [38]

    Chinese BosonSampling experiment: The gloves are off, https://scottaaronson.blog/?p=5159 (2020)

    2020

  31. [39]

    Xai-colossus, https://x.ai/colossus

  32. [40]

    November 2024 | TOP500, https://www.top500.org/lists/top500/2024/11/. 8

    2024

  33. [41]

    Chen, S.-Q

    Y . Chen, S.-Q. Gong, L. Gan, Y . Liu, A. Yang, C.-Y . Lu, and G. Yang, FastMPS: Breaking the Scaling Barrier of MPS-based Gaussian Boson Sampling Simulation

  34. [42]

    Raw data of jiuzhang 4.0, https://quantum.ustc.edu.cn/web/node/1227

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.