Pith. sign in

REVIEW 4 major objections 4 minor 35 references

Random matrix perspective on probabilistic error cancellation

T0 review · 4 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read The denoiser channels used in probabilistic error cancellation have spectra that are universal: they are determined by the noise strength, the circuit depth, and the universal spectral contour of random Lindblad generators.

desk verdict A parameter-free prediction connecting denoiser spectra to random Lindbladians, numerically supported but resting on an unproved shape-stability step; deserves peer review. read the letter →

arxiv 2512.01957 v1 pith:CQ64POYY submitted 2025-12-01 quant-ph cond-mat.stat-mech

Leonhard Moske , Pedro Ribeiro , Tomaž Prosen , Sergiy Denysov , Karol Życzkowski , David J. Luitz This is my paper · ORCID
classification quant-phcond-mat.stat-mech
keywords probabilisticerrorcancellationdenoiserrandomLindbladiansLindbladspectrummatrixtheoryquantummitigationlocalnoiseBCHexpansion
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 tries to establish that the unphysical 'denoiser' operator at the heart of probabilistic error cancellation is not a featureless object: its complex spectrum is inherited from the universal spectrum of random Lindblad generators, and can be predicted by a simple exponential map involving only the noise strength t, the number of noisy layers m, and the single-generator contour. If true, this means the behavior and cost of quantum error mitigation can be anticipated without knowing the microscopic details of the hardware noise. For local (few-qubit) noise, the same mechanism explains why the denoiser spectrum shows a hierarchy of decay timescales, and suggests that the locality of real hardware noise survives the scrambling action of random circuits.

What carries the argument

The denoiser D = U Λ_U^{-1} (the unphysical channel that recovers the target unitary from the noisy circuit) and the Baker–Campbell–Hausdorff linearization D^{-1} ≈ exp(t Σ L̃_i), in which the full noisy channel is replaced by the exponential of a sum of rotated Lindblad generators. The workhorse fact is that a sum of Lindbladians is itself a Lindbladian: additivity of the trace fixes the spectral center at -m, while the √m scaling of the support follows from random-matrix concentration or free probability. These two ingredients turn the known universal contour of a single random Lindbladian into the denoiser contour via g: f_L ↦ exp(-t(√m f_L - m)).

What would settle it

Take a small random circuit with known Lindblad generators, compute the exact denoiser spectrum, and compare the boundary with f_D = exp(-t(√m f_L - m)) at progressively larger t and m; the prediction fails if eigenvalues systematically fall outside the predicted contour once O(t²) commutator corrections become significant. A more targeted test: fix t and vary m, and check whether the support of the exact denoiser spectrum rescales as √m in the exponent.

Watch

Extended reading notes

Core claim

The central claim is that, for an ensemble of random noisy circuits Λ_U built from Haar-random unitaries and Markovian noise channels N_i = exp(t L_i), the denoiser D = U Λ_U^{-1} has a spectrum described by the map f_D = exp(-t(√m f_L - m)), where f_L runs over the universal 'lemon-shaped' contour of a single random Lindbladian. Using the Baker–Campbell–Hausdorff expansion, D^{-1} ≈ exp(t Σ L̃_i) to first order in t; the sum of rotated Lindblad generators is itself a Lindbladian, with trace fixing the spectral center at -m and support scaling as √m. The paper verifies this contour numerically for global and local noise, where it reproduces the hierarchy of decay rates of local random Liouvi

Load-bearing premise

The prediction assumes the noise operators effectively commute: second-order and higher commutator terms in the Baker–Campbell–Hausdorff expansion are dropped, and the spectrum of the sum of Lindbladians is assumed to have the same lemon-shaped support as a single random generator, merely shifted and rescaled.

Editorial extensions

If this is right

  • Denoiser spectra can be predicted from the universal Lindblad contour without characterizing the noise microscopically.
  • The denoiser spectral center sits at exp(t m), so the cost of inverting noise grows exponentially with noise strength and circuit depth.
  • For local noise, the denoiser spectrum displays a hierarchy of decay timescales inherited from local random Liouvillians.
  • The persistence of locality structure through random circuit scrambling suggests effective denoisers could be built from shallow few-body circuits.
  • The first-order BCH approximation is accurate for small t and small m, with deviations growing as both increase.

Reading between the lines

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

  • If the universal contour holds beyond the exact model, the sampling overhead of probabilistic error cancellation may be estimable a priori from t and m alone, without full process tomography.
  • The √m support scaling suggests that in the local case the hierarchy gaps between decay sectors should scale in a similar way; this is testable numerically.
  • A direct experimental test would be to measure the denoiser spectrum on hardware via process tomography and compare it to Eq. (17); systematic deviations would indicate non-Markovian or correlated noise.
  • The framework could be extended to non-Markovian noise channels, where the BCH argument fails; seeing how the contour breaks would quantify the limits of error mitigation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper studies an ensemble of noisy quantum circuits in the folded superoperator picture, with Haar-random unitary layers and random Lindblad noise channels of strength t. It defines the denoiser D = U Λ_U^{-1} and shows numerically that its complex spectrum has a lemon-like support. Using a first-order Baker-Campbell-Hausdorff approximation, the denoiser is approximated by exp(-t Σ_i \tilde L_i), and the spectrum of the sum of Lindbladians is argued to be a shifted and √m-rescaled version of the universal single-Lindblad spectrum of Ref. [15]. This leads to the central prediction Eq. (17), f_D = exp(-t(√m f_L - m)), with no fitted constants. The paper also considers local noise and shows that the hierarchy of relaxation timescales known for local random Liouvillians is inherited by the denoiser spectra.

Significance. If Eq. (17) is correct, the spectral support of a denoiser can be predicted from the universal random-Lindblad contour, without characterizing the individual noise channels. This would be useful for understanding error-mitigation overhead and the structure of unphysical inverse channels. The paper's numerics for N=32 and moderate t, m support the predicted contour, and the extension to local noise is a valuable qualitative observation. The derivation has no fitted parameters and relies on the independent universal spectrum of [15], which is a strength. However, the analytical link between the sum of Lindbladians and the single-Lindblad contour is not established, and the numerical evidence is largely visual. The central claim is plausible but not proven at the level required for a general universal statement.

major comments (4)
  1. [§IV, Eq. (15) and Appendix C] The first-order BCH truncation D^{-1} ≈ exp(t Σ_i \tilde L_i) is load-bearing. The paper states in Appendix C that 'the commutators of random Lindblad generators are small' but gives no proof, no bound, and no citation. The numerical agreement in Fig. 11 and Fig. 5 is shown only for small t and m (e.g., t=0.1, m=2; t=0.5, m=5). Since the central prediction Eq. (17) is built on this truncation, the omitted commutator estimate is not a technical footnote; it is required to justify the step from the exact denoiser to the sum-of-Lindbladians form.
  2. [§IV, Eq. (17)] The passage from the sum S_m = Σ_i \tilde L_i to the contour f_D = exp(-t(√m f_L - m)) is not derived. The text establishes only (i) the trace center at -m (Eq. (16), Appendix B) and (ii) an expected spectral-norm growth √m via Refs. [27,28]. Neither fixes the full boundary of the eigenvalue support. Appendix D computes the support endpoints of a sum of Kossakowski matrices, not the spectral shape of a sum of Lindbladians, and it gives only the endpoints m+1±2√m, not the lemon contour. For independent random Lindbladians, the m-fold free convolution of the universal measure is not obviously the same lemon shape; indeed, for large m the centered/rescaled sum may converge to a different law (e.g., the circular law). The paper needs either a proof of shape stability under free convolution or a direct computation of the S_m spectrum with a quantitative comparison to f_L.
  3. [Fig. 6 and Appendix D] The numerical evidence for the central prediction is visual-only and for a single parameter set (N=32, t=0.5, m=10). No quantitative metric is provided for how well the empirical boundary matches the predicted contour, and there is no systematic study in m of the shape of the rescaled sum-of-Lindbladians spectrum. Since the paper claims universality, it should show that the agreement persists for a range of N, m, and t, and ideally test the shape directly by comparing the spectral density of S_m to f_L after shifting and rescaling. Without such a test, the observed agreement in Fig. 6 could be a finite-size or small-m effect.
  4. [§V, local noise] The local-noise section demonstrates qualitatively that the hierarchy of timescales from local random Liouvillians appears in denoiser spectra, and Appendix C shows that the sum-of-Lindbladians approximation works in the local case. However, Eq. (17) is not tested for local noise: no predicted contour is overlaid on Fig. 7 or Fig. 12. Since the paper's title and abstract emphasize the inheritance of structure from random Lindbladians, the local case should either be connected to the quantitative prediction or clearly stated to be a separate qualitative result.
minor comments (4)
  1. [Eq. (17)] The definition of f_L is vague: the text says 'points {f_{L,i}} on the contour of the Lindblad spectrum with center at 0 and extent 2/N', but the actual universal contour of [15] is not written explicitly in the main text. A reader cannot reproduce Eq. (17) without going to the cited literature. Please give the explicit parametrization or at least the defining equation.
  2. [Fig. 5 and Fig. 8] The histograms plot 'minimal distance' between spectra but the axes are not labeled in the text. It would help to state the norm used (Euclidean in the complex plane) and to indicate the sample size used for the histogram.
  3. [Appendix D] The R-transform calculation has a potential inconsistency: the text writes R(z) = 1/(1-z) for a single Kossakowski matrix and then R(z) = m/(1-z) for the sum. This is correct for free convolution, but the derivation should state more explicitly that the Kossakowski matrices are assumed free and that the Lindblad generators inherit this property. The reader is left to infer the free-probability setup.
  4. [General notation] The symbol N is used both for the Hilbert-space dimension and for the number of circuit layers in Appendix D (where 'n' is used for the latter). This is confusing; please use distinct notation for layer count.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the denoiser contour prediction is a parameter-free map from an external Lindblad-spectrum result, not a fit or self-referential reduction.

full rationale

The derivation chain is: exact denoiser D = U Λ^{-1}; first-order BCH truncation D^{-1} ≈ exp(t Σ ilde L_i) (Eq. 15); observation that Σ ilde L_i is itself a Lindbladian; trace calculation fixing the spectral center at −m (Eq. 16 and Appendix B); and support scaling √m from independent matrix-concentration/free-probability results (Refs. [27,28] and Appendix D). The final contour map of Eq. 17, f_D = exp(−t(√m f_L − m)), transforms the external single-Lindblad contour f_L of [15] and is then compared with exact numerics (Figs. 5, 6, 11) without fitting any parameter to the denoiser spectra. The load-bearing citations [15] and [16] are independent published results on random Lindblad and Liouvillian spectra, not restatements of the present denoiser model; the author overlap therefore does not constitute circular evidence under the stated independence rule. The genuine weakness of the paper — that Eq. 17 assumes, rather than proves, that the full contour of a sum of m Lindbladians is the same lemon-shaped contour as a single Lindbladian, merely shifted and rescaled — is an unproven assumption and a formal gap, not a reduction of the prediction to its inputs. No fitted parameter is renamed as a prediction, and no equation is equivalent by construction to the quantity it is said to predict. Hence no specific circular step can be exhibited.

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

The model relies on standard random-matrix assumptions (Haar unitaries, Wishart Lindbladians) and two ad hoc approximations: first-order BCH truncation and contour-shape preservation of sums of Lindbladians. No free parameters are fitted to the denoiser spectra.

assumptions (6)
  • domain assumption Circuit layers U_i are i.i.d. Haar-random unitaries and noise channels N_i = exp(tL_i) with L_i independent random Lindbladians from a Wishart Kossakowski matrix.
    Model of unstructured noisy circuits defined in §II; the Haar/Wishart choices define the ensemble.
  • domain assumption Noise channels are Markovian CPTP maps generated by Lindbladians (Eq. 3).
    Stated in §II as justified for transmon platforms.
  • ad hoc to paper Higher-order BCH terms (commutators of rotated Lindbladians) are negligible for the denoiser spectrum.
    Eq. (15) drops O(t^2) terms; Appendix C claims commutators are small without proof or citation.
  • ad hoc to paper The spectrum of a sum of m independent Lindbladians has the same contour shape as a single Lindbladian, with mean shifted to -m and support scaled by √m.
    Used to derive Eq. (17); only the boundary scaling is argued via [27,28] and Appendix D.
  • domain assumption The denoiser is well-defined, i.e., Λ_U is invertible for the ensemble.
    Stated in §II; holds generically for finite t for these random channels.
  • domain assumption Local hardware noise is represented by kmax-local Lindbladians with Haar-rotated Kossakowski matrices on the local subspace.
    Model in §V; reproduces spectra of [16].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Random matrix perspective on probabilistic error cancellation." pith.science (2026). https://pith.science/paper/CQ64POYY

@misc{pith2026251201957,
  author       = {Pith},
  title        = {Pith review of: Random matrix perspective on probabilistic error cancellation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CQ64POYY}},
  note         = {Machine review of arXiv:2512.01957}
}
read the original abstract

Probabilistic error cancellation is an attempt to reverse the effect of dissipative noise channels on quantum computers by applying unphysical channels after the execution of a quantum algorithm on noisy hardware. We investigate on general grounds the properties of such unphysical quantum channels by considering a random matrix ensemble modeling noisy quantum algorithms. We show that the complex spectra of denoiser channels inherit their structure from random Lindbladians. Additional structure imposed by the locality of noise channels of the quantum computer emerges in terms of a hierarchy of timescales.

Figures

Figures reproduced from arXiv: 2512.01957 by the authors.

Figure 1
Figure 1. FIG. 1. Diagrammatic representation of eq. ( [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. FIG. 3. denoiser spectra for different system sizes [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Spectra of Λ [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: FIG. 4. denoiser spectra for different [PITH_FULL_IMAGE:figures/full_fig_p003_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Minimal distance between eigenvalues of the denoiser [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Comparison of the denoiser spectrum and the bound [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 8
Figure 8. Figure 8: FIG. 8. Minimal distance between eigenvalues of the denoiser [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9. Left: Spectra of Lindbladians, for [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Rescaled and shifted spectra of Lindbladians, for different system sizes [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11. Comparison of the denoiser spectrum (x) and the spectrum of exp [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12. Denoiser spectra for different localities [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 2 linked inside Pith

  1. [15]

    Denisov, T

    S. Denisov, T. Laptyeva, W. Tarnowski, D. Chru´ sci´ nski, and K. ˙Zyczkowski, Universal spectra of random Lind- blad operators, Physical Review Letters123, 140403 (2019)

    2019

  2. [1]

    Preskill, Quantum computing in the NISQ era and beyond, Quantum2, 79 (2018)

    J. Preskill, Quantum computing in the NISQ era and beyond, Quantum2, 79 (2018)

    2018

  3. [2]

    Dinca, D

    M. Dinca, D. J. Luitz, and M. Debertolis, Quantum process tomography of a compressed time evolution circuit on superconducting quantum processors (2025), 2509.25342

    arXiv 2025

  4. [3]

    Temme, S

    K. Temme, S. Bravyi, and J. M. Gambetta, Error mitiga- tion for short-depth quantum circuits, Physical Review Letters119, 180509 (2017)

    2017

  5. [4]

    R. S. Gupta, E. v. d. Berg, M. Takita, D. Riste, K. Temme, and A. Kandala, Probabilistic error cancel- lation for dynamic quantum circuits, Phys. Rev. A109, 062617 (2024)

    2024

  6. [5]

    E. v. d. Berg, Z. K. Minev, A. Kandala, and K. Temme, Probabilistic error cancellation with sparse Pauli-Lindblad models on noisy quantum processors, Na- ture Physics19, 1116 (2023)

    2023

  7. [6]

    Kandala, K

    A. Kandala, K. Temme, A. D. C´ orcoles, A. Mezzacapo, J. M. Chow, and J. M. Gambetta, Error mitigation ex- tends the computational reach of a noisy quantum pro- cessor, Nature567, 491–495 (2019)

    2019

  8. [7]

    Scheiber, P

    T. Scheiber, P. Haubenwallner, and M. Heller, Reduced sampling overhead for probabilistic error cancellation by Pauli error propagation, Quantum9, 1840 (2025)

    2025

Show all 35 references
  1. [8]

    Ferracin, A

    S. Ferracin, A. Hashim, J.-L. Ville, R. Naik, A. Carignan- Dugas, H. Qassim, A. Morvan, D. I. Santiago, I. Siddiqi, and J. J. Wallman, Efficiently improving the performance of noisy quantum computers, Quantum8, 1410 (2024)

    2024

  2. [9]

    Takagi, S

    R. Takagi, S. Endo, S. Minagawa, and M. Gu, Funda- mental limits of quantum error mitigation, npj Quantum Information8, 114 (2022)

    2022

  3. [10]

    Y. Kim, A. Eddins, S. Anand, K. X. Wei, E. van den Berg, S. Rosenblatt, H. Nayfeh, Y. Wu, M. Zaletel, K. Temme, and A. Kandala, Evidence for the utility of 7 quantum computing before fault tolerance, Nature618, 500 (2023)

    2023

  4. [11]

    Aharonov and M

    D. Aharonov and M. Ben-Or, Fault-tolerant quantum computation with constant error rate, SIAM Journal on Computing38, 1207 (2008)

    2008

  5. [12]

    A. Y. Kitaev, Quantum computations: algorithms and error correction, Russian Mathematical Surveys52, 1191 (1997)

    1997

  6. [13]

    P. W. Shor, Scheme for reducing decoherence in quantum computer memory, Physical Review A52, R2493 (1995)

    1995

  7. [14]

    M. S. J. Tepaske and D. J. Luitz, Compressed quan- tum error mitigation, Physical Review B107, L201114 (2023)

    2023

  8. [16]

    K. Wang, F. Piazza, and D. J. Luitz, Hierarchy of relax- ation timescales in local random Liouvillians, Physical Review Letters124, 100604 (2020)

    2020

  9. [17]

    L. S´ a, P. Ribeiro, T. Can, and T. Prosen, Spectral transi- tions and universal steady states in random Kraus maps and circuits, Physical Review B102, 134310 (2020)

    2020

  10. [18]

    Timm, Random transition-rate matrices for the mas- ter equation, Physical Review E80, 021140 (2009)

    C. Timm, Random transition-rate matrices for the mas- ter equation, Physical Review E80, 021140 (2009)

    2009

  11. [19]

    Lange and C

    S. Lange and C. Timm, Random-matrix theory for the Lindblad master equation, Chaos: An Interdisciplinary Journal of Nonlinear Science31, 023101 (2021)

    2021

  12. [20]

    O. E. Sommer, F. Piazza, and D. J. Luitz, Many-body hi- erarchy of dissipative timescales in a quantum computer, Physical Review Research3, 023190 (2021)

    2021

  13. [21]

    Kossakowski, On the general form of the generator of a dynamical semi-group for the spin 1/2 system, Bull

    A. Kossakowski, On the general form of the generator of a dynamical semi-group for the spin 1/2 system, Bull. Acad. Pol. Sci., Ser. Sci. Math. Astron. Phys.21, 649 (1973)

    1973

  14. [22]

    Lindblad, On the generators of quantum dynamical semigroups, Communications in Mathematical Physics 48, 119 (1976)

    G. Lindblad, On the generators of quantum dynamical semigroups, Communications in Mathematical Physics 48, 119 (1976)

    1976

  15. [23]

    ˙Zyczkowski and H.-J

    K. ˙Zyczkowski and H.-J. Sommers, Induced measures in the space of mixed quantum states, Journal of Physics A: Mathematical and General34, 7111–7125 (2001)

    2001

  16. [24]

    Fasi and L

    M. Fasi and L. Robol, Sampling the eigenvalues of ran- dom orthogonal and unitary matrices, Linear Algebra and its Applications620, 297 (2021)

    2021

  17. [25]

    Mezzadri, How to generate random matrices from the classical compact groups (2007), math-ph/0609050

    F. Mezzadri, How to generate random matrices from the classical compact groups (2007), math-ph/0609050

    2007 arXiv

  18. [26]

    Choi, Completely positive linear maps on complex matrices, Linear Algebra and its Applications10, 285 (1975)

    M.-D. Choi, Completely positive linear maps on complex matrices, Linear Algebra and its Applications10, 285 (1975)

    1975

  19. [27]

    J. A. Tropp, The expected norm of a sum of independent random matrices: An elementary approach, inHigh Di- mensional Probability VII, edited by C. Houdr´ e, D. M. Mason, P. Reynaud-Bouret, and J. Rosi´ nski (Springer International Publishing, Cham, 2016) pp. 173–202

    2016

  20. [28]

    L. S´ a, P. Ribeiro, and T. Prosen, Spectral and steady- state properties of random liouvillians, Journal of Physics A: Mathematical and Theoretical53, 305303 (2020)

    2020

  21. [29]

    Mangini, M

    S. Mangini, M. Cattaneo, D. Cavalcanti, S. Filippov, M. A. C. Rossi, and G. Garc ´ ıa-P´ erez, Tensor network noise characterization for near-term quantum computers, Phys. Rev. Res.6, 033217 (2024)

    2024

  22. [30]

    Brand, I

    D. Brand, I. Sinayskiy, and F. Petruccione, Markovian noise modelling and parameter extraction framework for quantum devices, Scientific Reports14, 4769 (2024)

    2024

  23. [31]

    Di Bartolomeo, M

    G. Di Bartolomeo, M. Vischi, F. Cesa, R. Wixinger, M. Grossi, S. Donadi, and A. Bassi, Noisy gates for sim- ulating quantum computers, Phys. Rev. Res.5, 043210 (2023)

    2023

  24. [32]

    Speicher, Free probability and random matrices (2014), arXiv:1404.3393

    R. Speicher, Free probability and random matrices (2014), arXiv:1404.3393

    2014 arXiv

  25. [33]

    M. A. Nowak and W. Tarnowski, Spectra of large time- lagged correlation matrices from random matrix theory, Journal of Statistical Mechanics: Theory and Experi- ment2017, 063405 (2017)

    2017

  26. [34]

    D. V. Voiculescu, K. J. Dykema, and A. Nica,Free ran- dom variables, Vol. 1 (American Mathematical Society, Providence, RI, 1992)

    1992

  27. [35]

    J.-P. Lund, R. S. Strichartz, and J. P. Vinson, Cauchy transforms of self-similar measures, Experimental Math- ematics7, 177 (1998). 8 A. APPENDIX A: SPECTRA OF LINDBLADIANS Here we show the spectra of the Lindblad generators used in the main text. FIG. 9. Left: Spectra of Lin...

    1998

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.