Pith. sign in

REVIEW 2 major objections 4 minor 6 cited by

Scaling Quantum Algorithms via Dissipation: Avoiding Barren Plateaus

T0 review · 2 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Periodically resetting ancillary qubits during a variational quantum circuit keeps gradient variances polynomially large, so dissipative algorithms can avoid both unitary and noise-induced barren plateaus.

desk verdict Plausible construction, but the key proof step pulls all noise out as a single scalar and that is not justified; send to review, expect major revision. read the letter →

arxiv 2507.02043 v2 pith:UGZ4XBKW submitted 2025-07-02 quant-ph

Elias Zapusek , Ivan Rojkov , Florentin Reiter This is my paper
classification quant-ph PACS 03.67.Lx03.67.-a
keywords barrenplateausdissipativequantumalgorithmsvariationalnon-unitalchannelsqubitresetnoise-inducedtoriccodegradientvariance
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

Variational quantum algorithms fail to scale when the cost landscape becomes flat: gradient variances shrink exponentially with system size (unitary barren plateaus), and noise flattens the landscape entirely. This paper claims that periodically resetting a subset of ancilla qubits during the circuit—a non-unital, entropy-removing operation—prevents both failure modes. The central analytic result is that, for circuits whose reset-to-reset jump depth is logarithmic in system size, with a constant fraction of qubits reset and a local cost function, the variance of gradients near the measurement stays at least polynomially large, even under unital noise. If the claim holds, dissipative variational circuits remain trainable where unitary ones are not, and the paper demonstrates this on toric-code ground-state preparation under noise.

What carries the argument

The load-bearing object is the circuit channel $\Phi_M = \prod_k (V_k \circ N \circ U_k \circ A_{\chi(k)})$, where $A$ is an amplitude-damping reset of $n_r$ equidistant qubits applied every $L$ layers, $N$ is a tensor product of unital single-qubit noise channels, $U_k$ are $d$-dimensional brickwork two-qubit layers, and $V_k$ are single-qubit layers. The proof uses the fact that the Clifford group is a two-design to replace averages over Haar-random gates by a fixed Clifford path that shrinks the support of a Pauli observable by one site per layer until it sits on a reset qubit. The contraction is controlled by three geometric quantities: $\Delta$, the depth needed to reach a reset qubit; $\Lambda$, the volume of gates that must be fixed along the path; and $\Gamma$, the number of resets encountered. Each is at most polynomial in the observable diameter $K$ and the jump depth $L$, and independent of total depth; with $n_r = \Omega(n)$ they are also independent of $n$. The noise contributes only the factor $D_{\max}^{2K^d \Delta}$, where $D_{\max}$ is the largest contraction factor of the single-qubit noise, and each reset contributes $q^2$ (or $(1-q)^2$ in the gradient calculation), producing the lower bound.

What would settle it

Numerically simulate the backward evolution of a Pauli operator through the random two-qubit layers (without the fixed Clifford contraction) for $L=O(\log n)$, $n_r=n/2$, and depolarizing noise, and track its Pauli weight. If the weight before contraction exceeds $K^d$ by more than a constant, the inequality behind the variance bound fails. The decisive experiment is then to measure $\mathrm{Var}[\partial C]$ for the dissipative brickwork circuit at $n=10,20,30$ with $L=\log n$, $n_r=n/2$, and noise strength $p=0.1$: the central claim predicts decay no faster than $1/\mathrm{poly}(n)$, while exponential decay $b^{-n}$ would refute it.

Watch

Extended reading notes

Core claim

The paper's central claim is that inserting reset channels into a random parameterized circuit changes the scaling of the cost landscape. For a $d$-dimensional lattice of $n$ qubits, if every $L=O(\log n)$ layers a fraction $n_r/n = \Omega(1)$ of qubits is reset by amplitude damping, and the cost observable has bounded diameter $K=O(1)$, then a parameterized gate $i=O(\log n)$ layers from the measurement has gradient variance at least $\Omega(1/\mathrm{poly}(n))$. This lower bound does not depend on total circuit depth, so the landscape does not flatten as more reset jumps are added, and it holds when unital noise is present. The proof's mechanism is to backward-evolve any local Pauli string through a fixed Clifford path to a single $Z$ operator on a reset qubit; along that path the noise contributes at most the largest contraction factor $D_{\max}$ and each reset contributes a factor $q^2$, and all remaining geometric factors are polynomial in the jump depth and observable diameter. The same mechanism yields a constant lower bound for constant-depth jumps and gates near the measurement.

Load-bearing premise

The proof assumes that the only noise an observable feels while traveling backward is determined by the size it has along one specially chosen Clifford path; if the operator actually spreads through the random two-qubit gates before that path shrinks it, the noise suppression would be stronger and the promised polynomial lower bound could collapse.

Editorial extensions

If this is right

  • With logarithmic-depth resets, the $O(\log n)$ parameterized gates nearest the measurement are trainable despite unital noise, so optimization needs only polynomially many measurement shots for those parameters.
  • Because the variance bound is independent of the number of reset jumps $M$, a dissipative circuit can be extended to arbitrary depth without reintroducing landscape flatness.
  • Constant-depth jumps and local observables give a constant gradient-variance lower bound for near-measurement gates, making the circuit trainable in the strongest sense.
  • The toric-code example shows a concrete task—ground-state preparation in a topologically ordered system—where unitary circuits with depth $\Omega(\sqrt{n})$ hit noise-induced barren plateaus, while constant-depth dissipative jumps do not.
  • The architectural conditions ($n_r=\Omega(n)$, $L=O(\log n)$, $K=O(1)$) are explicit design rules: a circuit builder who wants guaranteed trainability should reset a constant fraction of qubits every logarithmic-depth block and measure a local observable.

Reading between the lines

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

  • The paper does not claim, but its numerics hint, that correlated (repeated) parameters remain trainable; verifying this analytically would extend the result.
  • A direct extrapolation from the bound: gradient variance should increase with reset probability $q$ and reset fraction $n_r/n$; measuring this dependence would test the mechanism beyond the paper's examples.
  • The reset-induced trainability threshold may coincide with a measurement-induced entanglement phase transition, a connection the paper leaves unexplored.
  • The same proof strategy may apply to dissipative Gibbs-state preparation and stabilizer-code learning, where resets already play a central role; this is an extension, not a stated result.
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

2 major / 4 minor

Summary. The paper studies random parameterized quantum circuits with periodic resets of ancillary qubits, in the presence of unital noise. It claims that such dissipative circuits avoid both unitary and noise-induced barren plateaus: Proposition II.1 gives a lower bound on the variance of expectation values that is independent of total circuit depth, and Corollary II.1.1 states that gradients of gates within O(log n) layers of the measurement do not concentrate exponentially, provided the number of reset qubits scales as Ω(n), the jump depth is O(log n), and the cost is local. The proof uses a Clifford-fixing path technique, importing Pauli 2-mixing and second-moment lemmas from Ref. [36]. Numerical simulations on brickwork and QAOA ansätze, and on toric code ground-state preparation, support the qualitative claims.

Significance. If the central proof is correct, the result is significant: it identifies periodic resets as a concrete, architecturally explicit mechanism for avoiding both unitary and noise-induced barren plateaus, with a variance lower bound that is independent of the total number of jumps and written explicitly in terms of the reset probability q, the noise contraction factor D_max, the observable diameter K, and the lattice parameters. The paper makes appropriate use of recognized techniques and existing lemmas, and no fitted constants appear. The numerical section goes beyond the proof's assumptions by including correlated parameters and QAOA-inspired circuits, which strengthens the practical relevance. The main weakness is that the proof's handling of noise in the Heisenberg picture is sketched rather than rigorously derived; the central variance bound depends on a noise-accounting step that is not fully justified.

major comments (2)
  1. [Sec. IV B, Eq. (17); SI G, Eqs. (G7)-(G14)] The derivation of the noise factor is not justified. The right-hand side of Eq. (17) applies the adjoint noise channel N* once to the initial Pauli string and then evolves through a completely noiseless circuit Φ''_{[LM,1]}, where Φ'' is defined in SI F as the circuit with both single-qubit gates and noise removed. In the actual channel, however, unital noise acts after each of the LM layers; in the Heisenberg picture each layer contributes a factor D_{Q_j} for the Pauli operator Q_j present at that depth. The later factor D_max^{2K^d Δ} in Eq. (20) (and SI Eq. (G14)) is asserted in the text with the remark that 'each layer also applies a noise channel, contributing a suppression factor,' but it is not derived from Eq. (17). To make the proof valid, the authors need a lemma showing that, on the fixed-Clifford path, each of the Δ layers contributes at least D_max^{2K^d} and that the noiseless replacement in Eq. (17) does not discard noise factors from other layers. Without such a lemma, a careful extraction of one D_Q factor per layer would yield a factor that can be as small as D_min^{2LM}, exponentially small in total depth, which would invalidate Proposition II.1.
  2. [Sec. IV C, Eqs. (27)-(28); SI I, Eqs. (I16)-(I17)] The proof of the gradient bound assumes the existence of a fixed-Clifford path that backward-evolves the observable to the parameterized gate H_μ while keeping the support at most K^d ('we make sure that the support of the operator does not grow'). This assumption is load-bearing because the noise exponent in Eq. (33) is 2K^d(i−1)+2Δ. In a generic brickwork circuit, the backward light cone of a local observable over i layers has size growing with i; if the support actually grows, the exponent would grow like O(i^2), and D_max^{O(i^2)} with i=O(log n) is super-polynomially small, contradicting the claimed Ω(1/poly(n)) in Corollary II.1.1. The paper should either construct the path explicitly (for example, by SWAP routing along a fixed wire) or state and prove a lemma that such a path exists with the required volume bound. The current assertion that the path is guaranteed to exist is not sufficient for a rigorous proof.
minor comments (4)
  1. [Sec. II B, Proposition II.1] The condition 'n_r = O(n)' should be 'n_r = Ω(n)' for the lower bound to be independent of n; the SI (Corollary I.5.1) correctly uses n_r = Ω(n).
  2. [Sec. II A and SI F] The phrase 'Each reset occurs with probability q' is misleading: the reset is implemented by an amplitude-damping channel with parameter q that is applied deterministically every L layers, not by a probabilistic application of a reset operation. Please rephrase.
  3. [SI G, Eq. (G9)] The inequality D_max^{2|P|} ≥ D_max^{2K^d} is used without noting that it relies on D_max ≤ 1, and the surrounding text about per-layer noise factors is difficult to follow because Φ'' has noise removed; this step should be connected to the missing noise-extraction lemma described in the first major comment.
  4. [Sec. IV C, Eq. (24)-(25)] The notation f_j(·) is used in Eq. (24) but defined only in Eq. (25); reordering the definitions would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the analytic lower bounds are derived from stated assumptions and external lemmas, with no fitted parameters or load-bearing self-citation chain.

full rationale

The paper's central claims are analytic bounds, not fitted predictions. Proposition II.1 lower-bounds the variance using Pauli 2-mixing lemmas imported from Ref. [36] (an external group), the exact two-design property of the Clifford group, and explicit geometric bounds on the required contraction depth Delta, volume Lambda, and reset count Gamma. No parameter is fit to data and then renamed as a prediction; the lower bound is an explicit function of the given constants q, D_max, n_r, L, and K. Corollary II.1.1 follows by applying Proposition II.1 to a derived observable Q_R, not by assuming the conclusion. Self-citations appear only as context or as extensions (e.g., Refs. [29], [55], [56], [59], [60]) and are not load-bearing for the barren-plateau proofs. The one potentially fragile step is Eq. (17) / SI Eq. (G8), where all noise is effectively compressed into a single D_max factor before switching to a noiseless backward evolution; if that step is invalid the bound could fail, but that is a mathematical rigor concern, not circularity, because the claimed result is not equivalent to its inputs by construction.

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

The central claim rests on standard random-circuit assumptions (2-design gates, product unital noise, amplitude-damping resets) and on importing lemmas from Ref. [36]. No free parameters are fitted to data, and no new physical entities are postulated.

assumptions (5)
  • domain assumption Each two-qubit gate layer is drawn from a distribution forming a unitary 2-design.
    Used throughout to replace Haar averages with Clifford averages; stated in Sec. II A.
  • domain assumption The unital noise channel N is a tensor product of single-qubit contractive channels, not fully depolarizing, with D_max > 0.
    Required for a nonzero lower bound; stated in Sec. II A and SI E.
  • standard math The Clifford group is a unitary 2-design.
    Lemma D.1, used to fix specific Clifford gates in lower-bound arguments.
  • domain assumption Reset operations are amplitude damping channels with probability q, and reset qubits are equidistant on a d-dimensional lattice.
    Defines the dissipative channel A; stated in Sec. II A and App. F.
  • domain assumption The observable has bounded diameter K=O(1) and the parameterized gate lies within the inverse light cone of the observable.
    Required for the polynomial lower bound in Corollary II.1.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scaling Quantum Algorithms via Dissipation: Avoiding Barren Plateaus." pith.science (2026). https://pith.science/paper/UGZ4XBKW

@misc{pith2026250702043,
  author       = {Pith},
  title        = {Pith review of: Scaling Quantum Algorithms via Dissipation: Avoiding Barren Plateaus},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UGZ4XBKW}},
  note         = {Machine review of arXiv:2507.02043}
}
read the original abstract

Variational quantum algorithms (VQAs) have enabled a wide range of applications on near-term quantum devices. However, their scalability is fundamentally limited by barren plateaus, where the probability of encountering large gradients vanishes exponentially with system size. In addition, noise induces barren plateaus, deterministically flattening the cost landscape. Dissipative quantum algorithms that leverage nonunitary dynamics to prepare quantum states via engineered cooling offer a complementary framework with remarkable robustness to noise. We demonstrate that dissipative quantum algorithms based on non-unital channels can avoid both unitary and noise-induced barren plateaus. Periodically resetting ancillary qubits actively extracts entropy from the system, maintaining gradient magnitudes and enabling scalable optimization. We provide analytic conditions ensuring they remain trainable even in the presence of noise. Numerical simulations confirm our predictions and illustrate scenarios where unitary algorithms fail but dissipative algorithms succeed. Our framework positions dissipative quantum algorithms as a scalable, noise-resilient alternative to traditional VQAs.

Figures

Figures reproduced from arXiv: 2507.02043 by the authors.

Figure 1
Figure 1. Comparison of unitary and dissipative algorithms. (a) Unitary circuit. For most problems the depth of the unitary circuit must scale with the size of the problem. The increase in depth exponentially suppresses the gradients of the cost function making the circuit untrainable. (b) Cost landscape. Dissipative algorithms can avoid the suppression of gradients. (c) Dissipative circuits. The algorithm works by weakly cou… view at source ↗
Figure 2
Figure 2. Circuit model. The circuit model Φ[a,b] in one dimension. The circuit consists of the channel A χ (reset￾ting equidistant qubits), two-qubit unitaries U, noise N and single-qubit unitaries V. Since χ(b) = 1 and χ(a) = 0, a reset is applied in the first displayed layer but not in the last. The jump depth—defined as the number of layers between con￾secutive resets—is L = 2. The fraction of qubits that is reset is desc… view at source ↗
Figure 2
Figure 2. It consists of two-qubit bricks of the form: [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Noise-induced barren plateaus. The presence of noise leads to exponential concentration of expectation val￾ues in the depth of the circuit. In (a) we plot the variance of the gradient as a function of qubit number for the circuits that prepare toric code ground states.…
Figure 5
Figure 5. Figure 5: Operator transformation example A circuit il￾lustrating how two-qubit Clifford gates can be fixed to trans￾form observables. Specifically, the circuit maps the observable X ⊗4 to I ⊗ I ⊗ X ⊗ I. The required gate depth ∆ and the volume Λ of fixed gates enabling this tra…
Figure 6
Figure 6. Figure 6: Numerical simulation of the entropy of noisy quantum circuits. We numerically explore how the entropy of the output state scales compared to the fully mixed state for different system sizes n and unitary and dissipative ans¨atze preparing toric code ground states. Unit…
Figure 7
Figure 7. Figure 7: Entropy of noisy quantum circuits. In quantum algorithms subject to depolarizing noise, circuit depth scaling with system size leads to exponentially increasing entropy, quickly making computations unusable. Error correction is the solution, but it requires significant…
Figure 8
Figure 8. Figure 8: Dissipative quantum computation. In the ideal noiseless setting the overlap with the computational output state |ψT ⟩ scales as 1/(T + 1). Noise spreads the population across 2T states leading to the exponential suppression of the population in the desired output state…
Figure 9
Figure 9. Figure 9: Circuit model The circuit model in one dimension. The circuit consists of channel A resetting equidistant qubits, two-qubit unitaries U, noise N and single-qubit unitaries V. We show Φ[a,b] . Here χ(b) = 1, and χ(a) = 0 therefore we apply a reset in the first shown lay…
Figure 10
Figure 10. Figure 10: Operator transformation example A circuit illustrating how two-qubit Clifford gates can be fixed to transform observables. Specifically, the circuit maps the observable X ⊗4 to I ⊗ I ⊗ X ⊗ I where the X acts on a reset qubit. In every circuit layer the support can be …
Figure 11
Figure 11. Figure 11: Dissipative perceptron-based quantum neural networks. (a) Global perceptrons. Dissipative quantum neural networks that rely on deep global perceptrons, analogous to fully connected classical networks, suffer from barren plateaus. (b) Geometrically local perceptrons. B…
Figure 12
Figure 12. Figure 12: Variance of the brickwork circuit We plot the gradient variance for a one-dimensional brickwork ansatz with 40 layers. The dissipative ansatz resets every second qubit after 5 layers. In (a), all layer parameters are random, while in (b), they repeat every five layers…
Figure 13
Figure 13. Figure 13: Variance of the QAOA ansatz We plot the gradient variance for a one-dimensional QAOA ansatz with 40 layers. The dissipative ansatz resets every second qubit after 5 layers. In (a), all layer parameters are random, while in (b), they repeat every five layers. The gradi…
Figure 14
Figure 14. Figure 14: Circuits for toric code ground state preparation a. The unitary circuit is initialized in the product state [PITH_FULL_IMAGE:figures/full_fig_p032_14.png]
Figure 15
Figure 15. Figure 15: Noise-induced variance concentration The unitary circuit is initialized in the state |+⟩ ⊗n and then subjected to n/2 layers of unitaries, interleaved with depolarizing noise at a rate of 0.1. In contrast, the dissipative circuit starts in the fully mixed state and un…
Figure 16
Figure 16. Figure 16: Noise-induced deterministic concentration In the presence of noise the expectation values of any unitary circuits exponentially approach that of the fully mixed state in the depth of the circuit. Dissipative circuits can avoid this concentration. In Panel (a) we compa…
Figure 17
Figure 17. Figure 17: Infidelity of layered simulation with steady-state. We plot the infidelity of the exact steady-state with a state that consists of m layers. The layered simulation exponentially approximates the exact steady-state. At around 60 layers this decay plateaus with an infid…

Discussion (0). Sign in to comment.

Forward citations

Cited by 6 Pith papers

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

  1. Characterizing and Benchmarking Dynamic Quantum Circuits

    quant-ph 2026-04 unverdicted novelty 7.0 of 10

    Dynamarq is a new scalable benchmarking framework that defines structural features for dynamic quantum circuits and uses statistical models to predict hardware fidelity with transferable parameters.

  2. Challenges in Barren Plateau Mitigation with Dynamic Parameterized Quantum Circuits

    quant-ph 2026-06 unverdicted novelty 6.0 of 10

    Dynamic parameterized quantum circuits still leave a significant fraction of parameters untrainable despite cost anti-concentration, implying BP mitigation via DPQCs is at least as hard as designing BP-free unitaries.

  3. Challenges in Barren Plateau Mitigation with Dynamic Parameterized Quantum Circuits

    quant-ph 2026-06 conditional novelty 6.0 of 10

    Adding non-unitary gadgets to variational quantum circuits can create an illusion of trainability — the cost varies, but most parameters remain exponentially hard to train.

  4. Mitigating Noise-Induced Barren Plateaus Using a Non-Unitary Ansatz: Application to Molecular Electronic Transport

    quant-ph 2026-05 unverdicted novelty 6.0 of 10

    Non-unitary variational ansatze restore finite gradients under noise in VQAs for open quantum steady states, shown on an infinite-range dissipative Ising model and applied to first-principles electron transport in OPE-SMe.

  5. A Variational Dissipative Framework for Quantum Algorithms

    quant-ph 2026-05 unverdicted novelty 6.0 of 10

    Proposes integrating parameterized dissipative blocks via ancilla coupling and reset into variational quantum circuits for broader mixed-state preparation in optimization and recovery applications.

  6. Beyond Unitary Quantum Simulation: Open-System Approaches to Quantum Chemistry toward Quantum Advantage

    quant-ph 2026-05 unverdicted novelty 4.0 of 10

    Review arguing that open-system approaches integrating dissipation into quantum chemistry simulations on fault-tolerant computers offer practical advantages for robustness and potential quantum advantage over purely u...

Reference graph

Works this paper leans on

83 extracted references · 56 canonical work pages · cited by 5 Pith papers

  1. [36]

    A. A. Mele, A. Angrisani, S. Ghosh, S. Khatri, J. Eis- ert, D. S. Fran¸ ca, and Y. Quek, Noise-induced shal- low circuits and absence of barren plateaus (2024), arXiv:2403.13927

    arXiv 2024

  2. [1]

    Then, E U∼E Tr OU BU† 2 = X P∈{I,X,Y,Z} ⊗n a2 P E U∼E Tr P U BU† 2 .(D11)

    LetO= P P∈{I,X,Y,Z} ⊗n aP P, wherea P ∈Rfor allP. Then, E U∼E Tr OU BU† 2 = X P∈{I,X,Y,Z} ⊗n a2 P E U∼E Tr P U BU† 2 .(D11)

    work page

  3. [2]

    For anyP∈ {I, X, Y, Z}⊗n, E U∼E Tr P U BU† 2 = 1 3|P| X supp(Q)=supp(P) Q∈{I,X,Y,Z} ⊗n Tr(QB)2.(D12) Proof.We aim to calculateE U∼E [Tr OU BU† 2 ]. First, we rewrite the expression as E U∼E [Tr OU BU† 2 ] =E U∼E [Tr O⊗2U ⊗2B⊗2(U †)⊗2 ].(D13) ExpandingOin the Pauli basis asO= P P∈{I,X,Y,Z} ⊗n aP P, this becomes X P,Q∈{I,X,Y,Z} ⊗n aP aQ E U∼E [Tr (P⊗Q)U ⊗2B...

    work page

  4. [34]

    M. J. Kastoryano, F. Reiter, and A. S. Sørensen, Phys. Rev. Lett.106, 090502 (2011)

    2011

  5. [35]

    Morigi, J

    G. Morigi, J. Eschner, C. Cormick, Y. Lin, D. Leibfried, and D. J. Wineland, Phys. Rev. Lett.115, 200502 (2015)

    2015

  6. [37]

    J. C. Halimeh and P. Hauke, Stabilizing Gauge Theories in Quantum Simulators: A Brief Review (2022)

    2022

  7. [38]

    Deshpande, M

    A. Deshpande, M. Hinsche, S. Najafi, K. Sharma, R. Sweke, and C. Zoufal, Dynamic parameterized quan- tum circuits: Expressive and barren-plateau free (2024), arXiv:2411.05760

    arXiv 2024

  8. [39]

    Dankert, R

    C. Dankert, R. Cleve, J. Emerson, and E. Livine, Phys. Rev. A80, 012304 (2009)

    work page 2009

Show all 83 references
  1. [40]

    A. W. Harrow and S. Mehraban, Communications in Mathematical Physics401, 1531 (2023)

    2023

  2. [41]

    Farhi, J

    E. Farhi, J. Goldstone, and S. Gutmann, A Quantum Approximate Optimization Algorithm (2014), version Number: 1

    2014

  3. [42]

    Verdon, M

    G. Verdon, M. Broughton, and J. Biamonte, arXiv:1712.05304 (2019)

    2019 arXiv

  4. [43]

    Dennis, A

    E. Dennis, A. Kitaev, A. Landahl, and J. Preskill, J. Math. Phys.43, 4452 (2002), arXiv:quant-ph/0110143

    2002 arXiv

  5. [44]

    Bravyi, M

    S. Bravyi, M. B. Hastings, and F. Verstraete, Phys. Rev. Lett.97, 050401 (2006)

    2006

  6. [45]

    K¨ onig and F

    R. K¨ onig and F. Pastawski, Phys. Rev. B90, 045101 (2014)

    2014

  7. [46]

    J. T. Barreiro, M. M¨ uller, P. Schindler, D. Nigg, T. Monz, M. Chwalla, M. Hennrich, C. F. Roos, P. Zoller, and R. Blatt, Nature470, 486 (2011)

    2011

  8. [47]

    Raghunandan, F

    M. Raghunandan, F. Wolf, C. Ospelkaus, P. O. Schmidt, and H. Weimer, Sci. Adv.6, eaaw9268 (2020)

    2020

  9. [48]

    Aharonov, M

    D. Aharonov, M. Ben-Or, R. Impagliazzo, and N. Nisan, Limitations of Noisy Reversible Computation (1996), arXiv:quant-ph/9611028

    1996 arXiv

  10. [49]

    Aharonov, X

    D. Aharonov, X. Gao, Z. Landau, Y. Liu, and U. Vazi- rani, inProceedings of the 55th Annual ACM Sympo- sium on Theory of Computing(ACM, Orlando FL USA,

  11. [50]

    Fontana, M

    E. Fontana, M. S. Rudolph, R. Duncan, I. Rungger, and C. C ˆ ırstoiu, Classical simulations of noisy variational quantum circuits (2023), arXiv:2306.05400

    2023 arXiv

  12. [51]

    Gonz´ alez-Garc ´ ıa, J

    G. Gonz´ alez-Garc ´ ıa, J. I. Cirac, and R. Trivedi, Quan- tum9, 1730 (2025)

    2025

  13. [52]

    Fefferman, S

    B. Fefferman, S. Ghosh, M. Gullans, K. Kuroiwa, and K. Sharma, PRX Quantum5, 030317 (2024)

    2024

  14. [53]

    Ben-Or, D

    M. Ben-Or, D. Gottesman, and A. Hassidim, Quantum Refrigerator (2013), arxiv:1301.1995

    2013 arXiv

  15. [54]

    Shtanko and K

    O. Shtanko and K. Sharma, Complexity of Lo- cal Quantum Circuits under Nonunital Noise (2024), arXiv:2411.04819

    2024 arXiv

  16. [55]

    Zapusek, A

    E. Zapusek, A. Javadi, and F. Reiter, Quantum Sci. Technol.8, 015001 (2023)

    2023

  17. [56]

    Van Mourik, E

    M. Van Mourik, E. Zapusek, P. Hrmo, L. Gerster, R. Blatt, T. Monz, P. Schindler, and F. Reiter, Phys. Rev. Lett.132, 040602 (2024)

    2024

  18. [57]

    Reiter, A

    F. Reiter, A. S. Sørensen, P. Zoller, and C. A. Muschik, Nat. Commun.8, 1822 (2017)

    2017

  19. [58]

    I. Cong, S. Choi, and M. D. Lukin, Nat. Phys.15, 1273 (2019)

    2019

  20. [59]

    Rojkov, M

    I. Rojkov, M. Simoni, E. Zapusek, F. Reiter, and J. Home, Stabilization of cat-state manifolds using non- linear reservoir engineering (2024)

    2024

  21. [60]

    Rojkov, E

    I. Rojkov, E. Zapusek, and F. Reiter, Scalable dissipa- tive quantum error correction for discrete-variable codes (2025), arXiv:2507.12534

    2025

  22. [61]

    Banchi, J

    L. Banchi, J. Pereira, and S. Pirandola, PRX Quantum 2, 040321 (2021)

    2021

  23. [62]

    Muser, E

    T. Muser, E. Zapusek, V. Belis, and F. Reiter, Phys. Rev. A110, 032434 (2024)

    2024

  24. [63]

    F. Hu, S. A. Khan, N. T. Bronn, G. Angelatos, G. E. Rowlands, G. J. Ribeill, and H. E. T¨ ureci, Nat. Com- mun.15, 7491 (2024). 11

    2024

  25. [64]

    King and M

    C. King and M. Ruskai, IEEE Trans. Inf. Theory47, 192 (2001)

    2001

  26. [65]

    Verstraete and H

    F. Verstraete and H. Verschelde, On quantum channels (2002), version Number: 2

    2002

  27. [66]

    Collins and P

    B. Collins and P. ´Sniady, Communications in Mathe- matical Physics264, 773 (2006)

    2006

  28. [67]

    Christandl, The Structure of Bipartite Quantum States - Insights from Group Theory and Cryptography (2006), arXiv:quant-ph/0604183

    M. Christandl, The Structure of Bipartite Quantum States - Insights from Group Theory and Cryptography (2006), arXiv:quant-ph/0604183

    2006 arXiv

  29. [68]

    A. A. Mele, Introduction to Haar Measure Tools in Quantum Information: A Beginner’s Tutorial (2023), arXiv:2307.08956

    2023 arXiv

  30. [69]

    Gross, K

    D. Gross, K. Audenaert, and J. Eisert, J. Math. Phys. 48, 052104 (2007)

    2007

  31. [70]

    Webb, Quantum Inf

    Z. Webb, Quantum Inf. Comput.16, 1379 (2016)

    2016

  32. [71]

    T. H. Cormen, C. E. Leiserson, R. L. Rivest, and C. Stein,Introduction to algorithms, fourth edition ed. (The MIT Press, Cambridge, Massachusetts London, 2022)

    2022

  33. [72]

    M¨ uller-Hermes, D

    A. M¨ uller-Hermes, D. Stilck Fran¸ ca, and M. M. Wolf, J. Math. Phys.57, 022202 (2016)

    2016

  34. [73]

    A. M. Dalzell, N. Hunter-Jones, and F. G. S. L. Brand˜ ao, Random quantum circuits transform local noise into global white noise (2021)

    2021

  35. [74]

    Bultrini, S

    D. Bultrini, S. Wang, P. Czarnik, M. H. Gordon, M. Cerezo, P. J. Coles, and L. Cincio, Quantum7, 1060 (2023)

    2023

  36. [75]

    Bravyi, D

    S. Bravyi, D. Poulin, and B. Terhal, Phys. Rev. Lett. 104, 050503 (2010)

    2010

  37. [76]

    Watrous,The theory of quantum information(Cam- bridge university press, 2018)

    J. Watrous,The theory of quantum information(Cam- bridge university press, 2018)

    2018

  38. [77]

    J. M. Renes, R. Blume-Kohout, A. J. Scott, and C. M. Caves, J. Math. Phys.45, 2171 (2004)

    2004

  39. [78]

    Gottesman, The Heisenberg Representation of Quan- tum Computers (1998), arXiv:quant-ph/9807006

    D. Gottesman, The Heisenberg Representation of Quan- tum Computers (1998), arXiv:quant-ph/9807006

    1998 arXiv

  40. [79]

    Gottesman,Surviving as a Quantum Computer in a Classical World(Textbook manuscript preprint, 2016)

    D. Gottesman,Surviving as a Quantum Computer in a Classical World(Textbook manuscript preprint, 2016)

    2016

  41. [80]

    Kraus, A

    K. Kraus, A. B¨ ohm, J. D. Dollard, and W. H. Wootters, inStates, Effects, and Operations Fundamental Notions of Quantum Theory, Vol. 190 (Springer Berlin Heidel- berg, Berlin, Heidelberg, 1983) pp. 1–12, series Title: Lecture Notes in Physics

    1983

  42. [81]

    Beth Ruskai, S

    M. Beth Ruskai, S. Szarek, and E. Werner, Linear Al- gebra and its Applications347, 159 (2002)

    2002

  43. [82]

    D. P. Divincenzo, inMesoscopic Electron Transport, edited by L. L. Sohn, L. P. Kouwenhoven, and G. Sch¨ on (Springer Netherlands, Dordrecht, 1997) pp. 657–677

    1997

  44. [83]

    Kastler, J

    A. Kastler, J. Phys. Radium11, 255 (1950)

    1950

  45. [84]

    Wineland, W

    D. Wineland, W. Itano, J. Bergquist, J. Bollinger, and J. Prestage, Ann. Phys.10, 737 (1985)

    1985

  46. [85]

    Magnard, P

    P. Magnard, P. Kurpiers, B. Royer, T. Walter, J.-C. Besse, S. Gasparinetti, M. Pechal, J. Heinsoo, S. Storz, A. Blais, and A. Wallraff, Phys. Rev. Lett.121, 060502 (2018)

    2018

  47. [86]

    F. G. S. L. Brand˜ ao, A. W. Harrow, and M. Horodecki, Communications in Mathematical Physics346, 397 (2016)

    2016

  48. [87]

    Ortiz Marrero, M

    C. Ortiz Marrero, M. Kieferov´ a, and N. Wiebe, PRX Quantum2, 040316 (2021)

    2021

  49. [88]

    Ragone, B

    M. Ragone, B. N. Bakalov, F. Sauvage, A. F. Kem- per, C. Ortiz Marrero, M. Larocca, and M. Cerezo, Nat. Commun.15, 7172 (2024)

    2024

  50. [89]

    Fontana, D

    E. Fontana, D. Herman, S. Chakrabarti, N. Kumar, R. Yalovetzky, J. Heredge, S. H. Sureshbabu, and M. Pistoia, Nat. Commun.15, 7171 (2024)

    2024

  51. [90]

    Binkowski, G

    L. Binkowski, G. Koßmann, T. J. Osborne, R. Schwon- nek, and T. Ziegler, From barren plateaus through fer- tile valleys: Conic extensions of parameterised quantum circuits (2023), version Number: 2

    2023

  52. [91]

    D. N. Page, Physical Review Letters71, 1291 (1993)

    1993

  53. [92]

    Skinner, J

    B. Skinner, J. Ruhman, and A. Nahum, Phys. Rev. X 9, 031009 (2019)

    2019

  54. [93]

    Arrasmith, Z

    A. Arrasmith, Z. Holmes, M. Cerezo, and P. J. Coles, Quantum Sci. Technol.7, 045015 (2022)

    2022

  55. [94]

    Miao and T

    Q. Miao and T. Barthel, Quantum Sci. Technol.9, 045039 (2024)

    2024

  56. [95]

    M. V. Altaisky, Quantum neural network (2001), arXiv:quant-ph/0107012

    2001 arXiv

  57. [96]

    Sagheer and M

    A. Sagheer and M. Zidan, Autonomous Quantum Per- ceptron Neural Network (2013), version Number: 1

    2013

  58. [97]

    Schuld, I

    M. Schuld, I. Sinayskiy, and F. Petruccione, Quantum Inf. Process.13, 2567 (2014)

    2014

  59. [98]

    Siomau, Quantum Inf

    M. Siomau, Quantum Inf. Process.13, 1211 (2014), arXiv:1210.6626

    2014 arXiv

  60. [99]

    Torrontegui and J

    E. Torrontegui and J. J. Garcia-Ripoll, EPL (Euro- physics Letters)125, 30004 (2019)

    2019

  61. [100]

    Tacchino, C

    F. Tacchino, C. Macchiavello, D. Gerace, and D. Bajoni, npj Quantum Inf.5, 26 (2019)

    2019

  62. [101]

    K. Beer, D. Bondarenko, T. Farrelly, T. J. Osborne, R. Salzmann, D. Scheiermann, and R. Wolf, Nat. Com- mun.11, 808 (2020)

    2020

  63. [102]

    S. A. Wilkinson and M. J. Hartmann, Evaluating the performance of sigmoid quantum perceptrons in quan- tum neural networks (2022), version Number: 1

    2022

  64. [103]

    Wecker, M

    D. Wecker, M. B. Hastings, and M. Troyer, Phys. Rev. A92, 042303 (2015)

    2015

  65. [104]

    Bergholm, J

    V. Bergholm, J. Izaac, M. Schuld, C. Gogolin, S. Ahmed, V. Ajith, M. S. Alam, G. Alonso-Linaje, B. AkashNarayanan, A. Asadi,et al., PennyLane: Auto- matic differentiation of hybrid quantum-classical com- putations (2018), version Number: 4

    2018

  66. [105]

    Huang and C

    Q. Huang and C. B. Mendl, Phys. Rev. A105, 022409 (2022). 12 Appendix A: Definitions and notation The following definitions and notational conventions are used throughout the supplemental material. •Forn∈Nwe use [n] to refer to the set of integers{1, . . . , n}. •B(H) denotes ...

    2022

  67. [106]

    Additionally, the basis vectors should be orthonormal with respect to the inner product

    The coherence vector picture On the space of bounded linear operatorsB(H) we can choose a basis{F j}d2−1 j=0 that satisfiesF 0 =Iand Tr(F j) = 0∀j >0. Additionally, the basis vectors should be orthonormal with respect to the inner product. Fj =F † j ,⟨F k, Fj⟩= 1 d Tr (FkFj) =...

  68. [107]

    Then any density matrix can be represented as ρ= 1/2(F 0 +v·F) withv∈R 3 and∥v∥ ≤1

    Single-qubit channels and normal form We consider a single-qubit system withF 0 =IandF={X, Y, Z}. Then any density matrix can be represented as ρ= 1/2(F 0 +v·F) withv∈R 3 and∥v∥ ≤1. We define thePauli transfer matrixas M= 1 0 cM (E19) 21 where in accordance with Eq. (E13) and ...

  69. [108]

    No qubit in the path of the Pauli is reset

  70. [109]

    In the first case, no qubit in supp(P) is reset and we need to establish a connection to the closest reset qubit

    One or multiple qubits are reset. In the first case, no qubit in supp(P) is reset and we need to establish a connection to the closest reset qubit. We choose any direction ˆei. In this direction in the worst case the operator is positioned symmetrically between two reset qubit...

  71. [110]

    Specifically, they show that for local cost functions the lastO(log(n)) layers remain trainable

    proved that in the presence of single qubit non-unital noise parameterized quantum circuits avoid both noise induced and unitary barren plateaus. Specifically, they show that for local cost functions the lastO(log(n)) layers remain trainable. This counterintuitive result highl...

  72. [111]

    We have E[(f(O)) 2] = X P∈{I,X,Y,Z} ⊗n a2 P E[(f(P)) 2].(I8)

    LetO:= P P∈{I,X,Y,Z} ⊗n aP P, wherea P ∈Rfor anyP∈ {I, X, Y, Z}⊗n. We have E[(f(O)) 2] = X P∈{I,X,Y,Z} ⊗n a2 P E[(f(P)) 2].(I8)

  73. [112]

    Moreover, for anyP∈ {I, X, Y, Z}⊗n, we have E[(f(P)) 2] = 1 3|P| X Q∈{I,X,Y,Z} ⊗n| supp(Q)=supp(P) E[(f(Q)) 2].(I9) For proof see Ref. [36]. Lemma I.4.(Gradient zero outside the light cone) IfH µ is outside the inverse light cone of the observableOthe gradient vanishes∂ µC= 0....

  74. [113]

    The circuit ansatz we consider is built from two-qubit bricks of the form: U(θ) =R Y (θ1)⊗R Y (θ2) CNOTRX (θ3)⊗R X (θ4),(J1) denoting rotations generated byAasR A(θ) =e − i 2 θA

    Unitary barren plateaus We simulate random parameterized quantum circuits to numerically investigate the presence of unitary barren plateaus. The circuit ansatz we consider is built from two-qubit bricks of the form: U(θ) =R Y (θ1)⊗R Y (θ2) CNOTRX (θ3)⊗R X (θ4),(J1) denoting r...

  75. [114]

    We consider ground state preparation of the toric code Hamiltonian

    Noise induced barren plateaus Here we consider the effect of noise numerically and present a problem that can only be solved through the use of dissipation. We consider ground state preparation of the toric code Hamiltonian. The toric code is a type of quantum error-correcting...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.