Pith. sign in

REVIEW 2 major objections 5 minor 57 references

Energetic Cost of Temporal Information Processing in Quantum Reservoirs

T0 review · 2 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The energy a quantum reservoir spends per input is fixed by the local response of its spins; interactions build memory and nonlinearity without changing the bill.

desk verdict A credible, useful thermodynamic analysis of QRC that separates encoding cost from processing performance, with a real but manageable caveat about the steady-state approximation. read the letter →

arxiv 2608.08279 v1 pith:25GV566K submitted 2026-08-08 quant-ph

Gabriele Cenedese , Gonzalo Manzano , Gian Luca Giorgi , Roberta Zambrini This is my paper · ORCID
classification quant-ph
keywords quantumreservoircomputingswitchingworkthermodynamicsofinformationinteractingspinlocalresponsefunctionmemorycapacityopensystemsenergy-performancetrade-off
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

A quantum reservoir computer is a physical system whose natural dynamics converts a time-ordered input into a rich set of features, with training confined to a linear readout; this paper asks what determines the energy cost of feeding each new input into such a reservoir. Restricting to a chain of weakly interacting spins coupled to a thermal bath, the authors derive an analytical expression for the average switching work $W_k$—the work done when the input instantaneously updates the Hamiltonian—and show that in the weak-interaction regime it depends only on the local fields, disorder, input amplitude, and number of spins, not on the interaction strength $J_s$. Interactions instead redistribute the locally encoded information, generating the memory and nonlinearity that computational tasks exploit, while barely changing the work. The consequence is a task-dependent energy–performance relation: linear-memory tasks (short-term memory and NARMA) improve as the local response linearizes at large field $h$ while the work drops, whereas a nonlinearity-hungry parity-check task performs best at small $h$ where the response is strongly nonlinear and the work is higher. The paper concludes that the switching work is the energetic signature of information encoding, not of information processing, so there is no universal trade-off between energy efficiency and computational performance in this setting.

What carries the argument

The central object is the switching work per input step, $W_k = \operatorname{Tr}[(H_{k+1}-H_k)\rho_k]$, evaluated with the reservoir state immediately before the Hamiltonian update. In the weakly interacting regime the authors approximate $\rho_k$ by the product of single-spin stationary states in the instantaneous rotated local basis, whose transverse polarization is controlled by the local response function $F(a,h_i)=a/\sqrt{h_i^2+a_i^2}$. Averaging $W_k$ over uniformly distributed inputs $a\in[A,2A]$ and disorder $\delta_i\in[-D,D]$ yields Eq. (8). This identity carries the argument because it isolates the local, work-relevant part of the dynamics from the interaction-induced redistribution that builds memory and nonlinearity; it is what lets the paper separate the thermodynamics of encoding from the physics of processing.

What would settle it

Record the average switching work per input as a function of the input interval $\Delta t$ with all other parameters fixed. The analytical claim is that in the stationary regime this quantity is independent of $\Delta t$, since Eq. (8) depends only on the local stationary response; if reducing $\gamma\Delta t$ from 1 to 0.1 changes the measured work by more than the stated leading-order corrections, the local-stationary-state approximation underlying Eq. (8) is falsified.

Watch

Extended reading notes

Core claim

The central claim is that information encoding and information processing in the reservoir are governed by distinct physical mechanisms. Encoding is a local property: the average switching work per input step, Eq. (8), is the input- and disorder-averaged local response of the uncoupled spins, valid for $J_s \ll \Omega_{i,k}$, and it is independent of the interaction strength. Processing is a collective property: interactions transport and redistribute the already-encoded information, creating the nonlinear memory features the readout uses, without contributing directly to the work. The same local response function that fixes the work therefore also predicts the performance trends: as the local field $h$ grows, the response $F(a,h)=a/\sqrt{h^2+a^2}$ linearizes, which raises short-term memory and NARMA capacity while lowering the work, while the parity-check capacity rises as the response becomes more nonlinear at small $h$, where the work is larger. The paper presents this as evidence that the energy–performance relation in quantum reservoir computing is not universal but task-specific.

Load-bearing premise

Eq. (8) assumes that just before each switch the reservoir state is well approximated by the product of the single-spin stationary states in the rotated basis of the current input, even though the input interval is set comparable to the local relaxation time ($\gamma\Delta t=1$), so relaxation is incomplete.

Editorial extensions

If this is right

  • In the weak-interaction regime, the average switching work per input can be predicted from single-spin local response data alone; no knowledge of the interaction couplings is required.
  • Raising the local field $h$ linearizes the encoding, so short-term memory and NARMA capacities rise while the work per input falls, making energy efficiency and linear-memory performance compatible.
  • For parity-check tasks the opposite holds: the best nonlinear performance occurs at small $h$, where the local response is most nonlinear and the switching work is highest, so efficiency and nonlinear performance conflict.
  • Interactions modify the computational capacities but leave the switching work essentially unchanged until $J_s/h$ approaches order one; beyond that crossover both the work and the capacities change qualitatively.
  • Varying the disorder strength $D$ or the input amplitude $A$ changes the work and the capacities through the same local response function, so the encoding-versus-processing separation holds away from the homogeneous, fixed-amplitude case.

Reading between the lines

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

  • If the paper is right, energy-aware design of quantum reservoirs should be task-specific: a designer would tune the local field toward the linear or nonlinear regime depending on whether the target task is memory-dominated or nonlinearity-dominated, rather than minimizing switching work in all cases.
  • If the paper is right, the switching work could serve as a direct experimental probe of the local response during operation; measuring the transverse magnetization at switching instants would give an on-chip estimate of the encoding cost without solving the many-body dynamics.
  • If the paper is right, the same separation should appear for other input-encoding operators (for instance coupling the input to $\sigma^z_i$ or to a collective mode), with the local response function replaced by the appropriate susceptibility, although the present derivation does not test those cases.
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 / 5 minor

Summary. The paper studies the thermodynamics of information injection in a quantum reservoir computer. For an interacting spin reservoir driven by piecewise-constant inputs, it defines the switching work W_k = Tr[(H_{k+1} - H_k) rho_k] and derives, in the weakly interacting regime, an analytical formula (Eq. (8)) for the input- and disorder-averaged work that depends only on the local single-spin response F(a,h) = a/sqrt(h^2+a^2) and is independent of the interaction strength. Numerical simulations of the full GKLS dynamics for N=5 spins confirm the work formula in the claimed perturbative regime, and benchmark tasks (STM, parity-check, NARMA) are used to show that STM and NARMA capacities improve with a more linear local encoding at low work, while the parity-check capacity improves with a more nonlinear local encoding at higher work. The paper concludes that interactions mainly redistribute encoded information, so the energetic cost of encoding is set by the local response while interactions provide memory and nonlinearity.

Significance. The claimed separation between a local, input-driven energetic cost and an interaction-driven computational processing is conceptually attractive and, if correct, would provide a useful design principle for quantum reservoir computers. The analytical formula contains no fitted parameters and makes falsifiable, sign-specific predictions, such as opposite energetic-performance correlations for linear versus nonlinear tasks, and the numerical study is extensive, with 100 realizations and multiple benchmarks. The main weakness is that the central quantitative prediction relies on a local stationary-state approximation that is uncontrolled at the operating point gamma Delta t = 1; the 'essentially exact' agreement claims therefore need further quantitative support. With that support, the work would be a solid contribution to the thermodynamics of quantum reservoir computing.

major comments (2)
  1. [SM, 'Work estimate with jump operators in the rotated local basis'; main text, 'Numerical testing'] Equation (S31) replaces the pre-switch reservoir state rho_k by the product of single-spin stationary states in the rotated basis of the current input a_k. Since gamma Delta t = 1, relaxation within each input interval is incomplete: transverse dressed components decay as e^{-gamma Delta t / 2} = e^{-1/2} and longitudinal components as e^{-1}. Because W_k in Eq. (1) depends linearly on rho_k, averaging over independent stationary inputs does not remove the resulting error; a minimal scalar relaxation model would give E[W_k] = (1-lambda) Cov(a, F) rather than Cov(a, F). The manuscript acknowledges the leading-order nature of the formula but then states that Eq. (8) is 'essentially exact' for J_s = 10^{-1} and an 'excellent asymptotic approximation' up to J_s/h ~ 1; as presented, this could be error cancellation rather than evidence for the local-stationary assumption. I request a controlled test of this load-bearing step, for example by comparing Eq. (8) with numerics at gamma Delta t >> 1 (near-full relaxation) and at gamma Delta t = 1, or by deriving and benchmarking the leading relaxation correction to Eq. (8).
  2. [Main text, 'Model and analytical results' and 'Numerical testing'] The explanation that weak interactions 'promote mixing' and thereby improve agreement with the local analytical formula is asserted without a quantitative argument. Since Eq. (8) has no J_s dependence, the paper should either show within a perturbative expansion how interactions accelerate relaxation toward the local stationary state, or restrict the 'essentially exact' claim to the fully relaxed limit. Without this, the claim that interactions leave the switching work unaffected is supported only by numerical data in a narrow window under the same uncontrolled approximation.
minor comments (5)
  1. [Main text, 'Model and analytical results'] The main text sets D = 0 for the numerical scans, but Eq. (8) is written for D > 0 and has a 1/D prefactor; please state explicitly that the D -> 0 limit is used for Figs. 1-3 and provide the limiting expression to avoid ambiguity.
  2. [Abstract and Figs. 1-3] The statement that interactions 'generate memory and nonlinear features' is stronger than the numerical evidence, where increasing J_s leaves STM/NARMA capacities roughly flat until J_s/h ~ 1 and generally lowers the PC capacity; consider qualifying this claim to 'redistribute' rather than 'generate'.
  3. [Fig. 4 caption] The normalization of capacity and switching work to the interval [0,1] 'per task' is not fully specified; state whether min-max normalization over the scanned h values is used, and note that the displayed slopes are not quantitative.
  4. [SM, 'Microscopic description of the thermal GKLS generator'] The sentence 'flat bath spectral density, J(omega) = gamma, independent of frequency, , yielding gamma_|omega| = gamma' contains a stray double comma; also, the notation gamma for both the coupling strength and the frequency-dependent rates is confusing and should be disambiguated.
  5. [Main text, 'Numerical testing'] The term 'validate' is slightly overstrong because the numerics implement the same global GKLS master equation whose local limit was used to derive Eq. (8); the agreement is an internal consistency check, and the manuscript could state this more carefully.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Eq. (8) is a parameter-free analytical consequence of the stated model; the self-citations are contextual, not load-bearing.

full rationale

The derivation chain is self-contained. The switching work is defined in Eq. (1) from the Hamiltonian update and the pre-switch state; the model Hamiltonian is given in Eq. (3); the local stationary polarizations are derived in Eqs. (4)-(6); and Eq. (7) approximates the pre-switch state by the local stationary state, leading to the averaged expression Eq. (8). No parameter is fitted to the quantity being predicted, and no external result is imported to force the conclusion. The numerical simulations integrate the same GKLS model and therefore provide a self-consistency check rather than an independent benchmark, but that is not circularity. The paper explicitly acknowledges the main approximation: in the main text it states that because the reservoir does not fully relax within one input interval, the analytical expressions 'should therefore be interpreted as leading-order approximations rather than exact stationary results,' and in the Supplemental Material it notes that 'the estimates above assume that the system has enough time to approach the local rotated-basis stationary state during each input interval, and they neglect interaction-induced corrections.' This is a validity limitation, not a reduction of the prediction to its own inputs. The self-citations (e.g., Refs. [5,27,29,32,49]) are used for context or for previously established QRC properties; the central energetic result does not depend on their correctness. No self-definitional, fitted-input, self-citation-load-bearing, uniqueness-imported, ansatz-smuggled, or renaming circularity is present.

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

The analytical result rests on the weak-coupling GKLS framework and the local stationary-state approximation. No free parameters are fitted to data; model parameters (A, h, J_s, D, beta, gamma Delta t) are inputs. No new entities are introduced.

assumptions (5)
  • domain assumption The weak-coupling Born-Markov-secular (Davies) GKLS master equation accurately describes the open system dynamics.
    Standard open quantum systems framework invoked in 'Thermodynamic framework' and SM 'Microscopic description of the thermal GKLS generator'.
  • domain assumption The input sequence is i.i.d. from a stationary distribution and, after washout, the reservoir is in a statistically stationary regime with <Delta U> = 0 and <Delta S> = 0.
    Used to equate the average dissipated work with the average switching work; stated in 'Thermodynamic framework'.
  • ad hoc to paper At each switching event the reservoir state is approximately the product of local stationary states in the rotated basis of the current input (SM Eq. S31).
    This is the key approximation enabling the analytical work formula (Eq. 8); it is a leading-order approximation that ignores incomplete relaxation and interaction corrections.
  • domain assumption The bath satisfies the KMS condition with local detailed balance, gamma_omega = gamma_{-omega} e^{beta omega}.
    Standard thermal bath assumption; used to define the dissipative dynamics and stationary state.
  • domain assumption The analytical work formula uses the zero-temperature limit of the local stationary polarization, with beta Omega much greater than 1.
    The derivation in SM sets tanh(beta Omega) approximately 1; numerically beta = 10 is chosen so this holds over the scanned parameter range.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Energetic Cost of Temporal Information Processing in Quantum Reservoirs." pith.science (2026). https://pith.science/paper/25GV566K

@misc{pith2026260808279,
  author       = {Pith},
  title        = {Pith review of: Energetic Cost of Temporal Information Processing in Quantum Reservoirs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/25GV566K}},
  note         = {Machine review of arXiv:2608.08279}
}
read the original abstract

Quantum reservoir computing offers a promising route toward energy-efficient machine learning by processing temporal information with minimal training overhead. Yet, the physical principles linking its energetic cost to computational performance remain largely unexplored. Here we show that, in an interacting spin reservoir, information encoding and information processing are governed by distinct physical mechanisms. In the weak interacting regime, we derive an analytical expression for the average (switching) work, showing that the energetic cost of encoding new inputs is determined by the local response of the reservoir units. In contrast, interactions primarily redistribute the encoded information, generating memory and nonlinear features while only weakly affecting the work. This separation produces opposite correlations between energetic cost and performance for representative linear and nonlinear benchmark tasks. Our results identify the switching work as the energetic signature of information encoding and clarify when energetic efficiency and computational performance are compatible.

Figures

Figures reproduced from arXiv: 2608.08279 by the authors.

Figure 3
Figure 3. FIG. 3. Same as Fig. 1(c,d) for the fifth-order NARMA capac [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Same as Fig. 1 for the fourth-order PC capacity, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. FIG. 4. Normalized task capacity versus normalized aver [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (2 more)
Figure 7
Figure 7. Figure 7: FIG. 7. (a) Average switching work as a function of the dis [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Average capacity for the third- and tenth-order [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 29 canonical work pages

  1. [1]

    Maass, T

    W. Maass, T. Natschl¨ ager, and H. Markram, Real-time computing without stable states: A new framework for neural computation based on perturbations, Neural Com- putation14, 2531 (2002)

    2002

  2. [2]

    Lukoˇ seviˇ cius and H

    M. Lukoˇ seviˇ cius and H. Jaeger, Reservoir computing ap- proaches to recurrent neural network training, Computer Science Review3, 127 (2009)

    work page 2009

  3. [3]

    Fujii and K

    K. Fujii and K. Nakajima, Harnessing disordered- ensemble quantum dynamics for machine learning, Phys- ical Review Applied8, 024030 (2017)

    2017

  4. [4]

    Mujal, R

    P. Mujal, R. Mart ´ ınez-Pe˜ na, J. Nokkala, J. Garc ´ ıa-Beni, G. L. Giorgi, M. C. Soriano, and R. Zambrini, Oppor- tunities in quantum reservoir computing and extreme learning machines, Advanced Quantum Technologies4, 2100027 (2021), arXiv:2102.11831 [quant-ph]

    arXiv 2021

  5. [5]

    Mart ´ ınez-Pe˜ na, G

    R. Mart ´ ınez-Pe˜ na, G. L. Giorgi, J. Nokkala, M. C. So- riano, and R. Zambrini, Dynamical phase transitions in quantum reservoir computing, Physical Review Letters 127, 100502 (2021), arXiv:2103.05348 [quant-ph]

    arXiv 2021

  6. [6]

    Kornjaˇ ca, H.-Y

    M. Kornjaˇ ca, H.-Y. Hu, C. Zhao, J. Wurtz, P. Wein- berg, M. Hamdan, A. Zhdanov, S. H. Cantu, H. Zhou, R. A. Bravo,et al., Large-scale quantum reservoir learn- ing with an analog quantum computer, arXiv preprint arXiv:2407.02553 (2024)

    arXiv 2024

  7. [7]

    Ghosh, A

    S. Ghosh, A. Opala, M. Matuszewski, T. Paterek, and T. C. Liew, Quantum reservoir processing, npj Quantum Information5, 35 (2019)

    work page 2019

  8. [8]

    Kobayashi and Y

    K. Kobayashi and Y. Motome, Edge of many-body quan- tum chaos in quantum reservoir computing, Physical Re- view Letters136, 040602 (2026)

    work page 2026

Show all 57 references
  1. [9]

    Llodr` a, C

    G. Llodr` a, C. Charalambous, G. L. Giorgi, and R. Zam- brini, Benchmarking the role of particle statistics in quantum reservoir computing, Advanced Quantum Tech- nologies6, 2200100 (2023)

    2023

  2. [10]

    Ghosh, A

    S. Ghosh, A. Opala, M. Matuszewski, T. Paterek, and T. C. Liew, Reconstructing quantum states with quan- tum reservoir networks, IEEE Transactions on Neural Networks and Learning Systems32, 3148 (2020)

    2020

  3. [11]

    Nokkala, R

    J. Nokkala, R. Mart ´ ınez-Pe˜ na, G. L. Giorgi, V. Pa- rigi, M. C. Soriano, and R. Zambrini, Gaussian states of continuous-variable quantum systems provide univer- sal and versatile reservoir computing, Communications Physics4, 53 (2021)

    2021

  4. [12]

    L. C. G. Govia, G. J. Ribeill, G. E. Rowlands, H. K. Krovi, and T. A. Ohki, Quantum reservoir computing with a single nonlinear oscillator, Phys. Rev. Res.3 (2021)

    2021

  5. [13]

    W. D. Kalfus, G. J. Ribeill, G. E. Rowlands, H. K. Krovi, T. A. Ohki, and L. C. G. Govia, Hilbert space as a com- putational resource in reservoir computing, Phys. Rev. Res.4(2022)

    2022

  6. [14]

    Garc ´ ıa-Beni, G

    J. Garc ´ ıa-Beni, G. L. Giorgi, M. C. Soriano, and R. Zam- brini, Scalable photonic platform for real-time quantum reservoir computing, Physical Review Applied20, 014051 (2023), arXiv:2207.14031 [quant-ph]

    2023 arXiv

  7. [15]

    Paparelle, J

    I. Paparelle, J. Henaff, J. Garc ´ ıa-Beni,´E. Gillet, D. Mon- tesinos, G. L. Giorgi, M. C. Soriano, R. Zambrini, and V. Parigi, Experimental memory control in continuous- variable optical quantum reservoir computing, Nat. Pho- tonics20, 413 (2026)

    2026

  8. [16]

    Nerenberg, O

    S. Nerenberg, O. D. Neill, G. Marcucci, and D. Faccio, Photon number-resolving quantum reservoir computing, Optica Quantum3, 201 (2025)

    2025

  9. [17]

    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 com- puting using a gaussian boson sampler, NPJ Quantum Inf.12(2026)

    2026

  10. [18]

    Suzuki, Q

    Y. Suzuki, Q. Gao, K. C. Pradel, K. Yasuoka, and N. Ya- mamoto, Natural quantum reservoir computing for tem- poral information processing, Scientific Reports12, 1353 (2022)

    2022

  11. [19]

    J. Chen, H. I. Nurdin, and N. Yamamoto, Temporal infor- mation processing on noisy quantum computers, Physical Review Applied14, 024065 (2020)

    2020

  12. [20]

    F. Hu, S. A. Khan, N. T. Bronn, G. Angelatos, G. E. Rowlands, G. J. Ribeill, and H. E. T¨ ureci, Overcoming the coherence time barrier in quantum machine learn- ing on temporal data, Nature communications15, 7491 (2024)

    2024

  13. [21]

    Dudas, B

    J. Dudas, B. Carles, E. Plouet, F. A. Mizrahi, J. Grol- lier, and D. Markovi´ c, Quantum reservoir computing im- plementation on coherently coupled quantum oscillators, npj Quantum Information9, 64 (2023)

    2023

  14. [22]

    Senanian, S

    A. Senanian, S. Prabhu, V. Kremenetski, S. Roy, Y. Cao, J. Kline, T. Onodera, L. G. Wright, X. Wu, V. Fatemi, et al., Microwave signal processing using an analog quan- tum reservoir computer, Nature Communications15, 7490 (2024)

    2024

  15. [23]

    S. Das, G. L. Giorgi, and R. Zambrini, Quantum reservoir computing in jaynes-cummings models: Nonlinear mem- ory and time-series prediction, Physical Review Research 8, 023148 (2026)

    2026

  16. [24]

    Negoro, K

    M. Negoro, K. Mitarai, K. Fujii, K. Nakajima, and M. Kitagawa, Machine learning with controllable quan- tum dynamics of a nuclear spin ensemble in a solid, arXiv preprint arXiv:1806.10910 (2018)

    2018 arXiv

  17. [25]

    Y. Hou, J. Hua, Z. Wu, W. Xia, Y. Chen, X. Li, Z. Li, X. Peng, and J. Du, High-accuracy temporal prediction via experimental quantum reservoir computing in corre- lated spins, Physical Review Letters136, 120602 (2026)

    2026

  18. [26]

    Araiza Bravo, K

    R. Araiza Bravo, K. Najafi, X. Gao, and S. F. Yelin, Quantum reservoir computing using arrays of rydberg atoms, PRX Quantum3, 030325 (2022), arXiv:2111.10956 [quant-ph]

    2022 arXiv

  19. [27]

    Llodr` a, P

    G. Llodr` a, P. Mujal, R. Zambrini, and G. L. Giorgi, Quantum reservoir computing in atomic lattices, Chaos, Solitons & Fractals195, 116289 (2025)

    2025

  20. [28]

    Wringe, M

    C. Wringe, M. Trefzer, and S. Stepney, Reservoir com- puting benchmarks: a tutorial review and critique, Inter- national Journal of Parallel, Emergent and Distributed Systems40, 313 (2025)

    2025

  21. [29]

    Palacios, R

    A. Palacios, R. Mart ´ ınez-Pe˜ na, M. C. Soriano, G. L. Giorgi, and R. Zambrini, Role of coherence in many-body quantum reservoir computing, Communications Physics 7, 369 (2024), arXiv:2409.17734 [quant-ph]. 7

    2024 arXiv

  22. [30]

    Mujal, R

    P. Mujal, R. Mart ´ ınez-Pe˜ na, G. L. Giorgi, M. C. Soriano, and R. Zambrini, Time-series quantum reservoir comput- ing with weak and projective measurements, npj Quan- tum Information9, 16 (2023), arXiv:2205.06809 [quant- ph]

    2023 arXiv

  23. [31]

    Morgu ´ ı-Sancho, G

    O. Morgu ´ ı-Sancho, G. Manzano, G. L. Giorgi, and R. Zambrini, General theory of monitored quantum reser- voir computing, arXiv preprint arXiv:2607.14311 (2026), arXiv:2607.14311 [quant-ph]

    2026 arXiv

  24. [32]

    Sannia, R

    A. Sannia, R. Mart ´ ınez-Pe˜ na, M. C. Soriano, G. L. Giorgi, and R. Zambrini, Dissipation as a resource for quantum reservoir computing, Quantum8, 1291 (2024), arXiv:2212.12078 [quant-ph]

    2024 arXiv

  25. [33]

    Ding and X

    L. Ding and X. Qiu, Thermodynamics of quantum reser- voir computing, arXiv preprint arXiv:2607.02157 (2026)

    2026 arXiv

  26. [34]

    Gorini, A

    V. Gorini, A. Kossakowski, and E. C. G. Sudarshan, Completely positive dynamical semigroups ofn-level sys- tems, Journal of Mathematical Physics17, 821 (1976)

    1976

  27. [35]

    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

  28. [36]

    Breuer and F

    H.-P. Breuer and F. Petruccione,The Theory of Open Quantum Systems(Oxford University Press, Oxford, 2002)

    2002

  29. [37]

    Alicki, The quantum open system as a model of the heat engine, Journal of Physics A: Mathematical and General12, L103 (1979)

    R. Alicki, The quantum open system as a model of the heat engine, Journal of Physics A: Mathematical and General12, L103 (1979)

    1979

  30. [38]

    Kosloff, Quantum thermodynamics: A dynamical viewpoint, Entropy15, 2100 (2013)

    R. Kosloff, Quantum thermodynamics: A dynamical viewpoint, Entropy15, 2100 (2013)

    2013

  31. [39]

    Vinjanampathy and J

    S. Vinjanampathy and J. Anders, Quantum ther- modynamics, Contemporary Physics57, 545 (2016), https://doi.org/10.1080/00107514.2016.1201896

    2016

  32. [40]

    E. B. Davies, Markovian master equations, Communica- tions in Mathematical Physics39, 91 (1974)

    1974

  33. [41]

    Spohn and J

    H. Spohn and J. L. Lebowitz, Irreversible thermodynam- ics for quantum systems weakly coupled to thermal reser- voirs, Advances in Chemical Physics: For Ilya Prigogine , 109 (1978)

    1978

  34. [42]

    Spohn, Entropy production for quantum dynamical semigroups, Journal of Mathematical Physics19, 1227 (1978)

    H. Spohn, Entropy production for quantum dynamical semigroups, Journal of Mathematical Physics19, 1227 (1978)

    1978

  35. [43]

    J. M. Parrondo, J. M. Horowitz, and T. Sagawa, Thermo- dynamics of information, Nature physics11, 131 (2015)

    2015

  36. [44]

    A. S. Trushechkin and I. V. Volovich, Perturbative treat- ment of inter-site couplings in the local description of open quantum networks, Europhysics Letters113, 30005 (2016)

    2016

  37. [45]

    J. O. Gonz´ alez, L. A. Correa, G. Nocerino, J. P. Palao, D. Alonso, and G. Adesso, Testing the validity of the local and global GKLS master equations on an exactly solvable model, Open Systems & Information Dynamics 24, 1740010 (2017)

    2017

  38. [46]

    Jaeger,Short Term Memory in Echo State Networks, Tech

    H. Jaeger,Short Term Memory in Echo State Networks, Tech. Rep. GMD Report 152 (German National Research Center for Information Technology, 2002)

    2002

  39. [47]

    A. F. Atiya and A. G. Parlos, New results on recurrent network training: Unifying the algorithms and accelerat- ing convergence, IEEE Transactions on Neural Networks 11, 697 (2000)

    2000

  40. [48]

    Dambre, D

    J. Dambre, D. Verstraeten, B. Schrauwen, and S. Massar, Information processing capacity of dynamical systems, Scientific Reports2, 514 (2012)

    2012

  41. [49]

    Mart ´ ınez-Pe˜ na, J

    R. Mart ´ ınez-Pe˜ na, J. Nokkala, G. L. Giorgi, R. Zambrini, and M. C. Soriano, Information processing capacity of spin-based quantum reservoir computing systems, Cog- nitive Computation15, 1440 (2023), arXiv:2010.06369 [quant-ph]

    2023 arXiv

  42. [50]

    Sagawa and M

    T. Sagawa and M. Ueda, Minimal energy cost for thermo- dynamic information processing: Measurement and in- formation erasure, Phys. Rev. Lett.102, 250602 (2009)

    2009

  43. [51]

    Guryanova, N

    Y. Guryanova, N. Friis, and M. Huber, Ideal Projective Measurements Have Infinite Resource Costs, Quantum4, 222 (2020)

    2020

  44. [52]

    L Latune and C

    C. L Latune and C. Elouard, A thermodynamically con- sistent approach to the energy costs of quantum measure- ments, Quantum9, 1614 (2025)

    2025

  45. [53]

    Van Vu and K

    T. Van Vu and K. Saito, Time-cost-error trade-off re- lation in thermodynamics: The third law and beyond, Phys. Rev. X15, 041029 (2025)

    2025

  46. [54]

    Freitas, J.-C

    N. Freitas, J.-C. Delvenne, and M. Esposito, Stochastic thermodynamics of nonlinear electronic circuits: A real- istic framework for computing aroundkt, Phys. Rev. X 11, 031064 (2021)

    2021

  47. [55]

    Lipka-Bartosik, M

    P. Lipka-Bartosik, M. Perarnau-Llobet, and N. Brunner, Thermodynamic comput- ing via autonomous quantum thermal ma- chines, Science Advances10, eadm8792 (2024), https://www.science.org/doi/pdf/10.1126/sciadv.adm8792

    2024 doi

  48. [56]

    Energetic Cost of Temporal Information Processing in Quantum Reservoirs

    G. Manzano, G. Karde¸ s, E. Rold´ an, and D. H. Wolpert, Thermodynamics of computations with absolute irre- versibility, unidirectional transitions, and stochastic com- putation times, Phys. Rev. X14, 021026 (2024). END MA TTER Additional PC and NARMA benchmark.– To demon- str...

    2024

  49. [57]

    Therefore, around the crossover h∼a, weak disorder can either increase or decrease the local transverse response, and its effect is not captured by the large-hexpansion alone. In the strong-input regime,h≪a, one finds instead Eδ[Wk]≃ −N(ak+1 −a k) 1− h2 2a2 k − D2 6a2 k +O h4 ...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.