Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Knowledge Distillation Inspired Variational Quantum Eigensolver with Virtual Annealing

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read KD-VQE runs several trial wavefunctions at once and uses a virtual annealing temperature to shift measurement shots toward the lowest-energy candidate, improving VQE's convergence reliability on the two-site Fermi-Hubbard model.

desk verdict Coherent but unproven: the single four-qubit demo cannot carry the claimed reliability, and the annealing dynamics need analysis rather than assertion. read the letter →

arxiv 2505.03998 v1 pith:NJKWPISS submitted 2025-05-06 quant-ph

Junxu Li This is my paper · ORCID
classification quant-ph MSC 81P68 PACS 03.67.Ac03.67.Lx
keywords variationalquantumeigensolverknowledgedistillationvirtualannealingBoltzmanndistributionmeasurementshotallocationFermi-Hubbardmodelhybridquantum-classicalalgorithmoptimization
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 proposes an extension of the variational quantum eigensolver (VQE) called KD-VQE, which runs several trial wavefunctions at once and distributes measurement shots among them according to a Boltzmann distribution at a virtual temperature. As the temperature is lowered step by step, shots are pulled away from high-energy candidates and concentrated on the lowest-energy one, a process the paper calls virtual annealing. The claim is that this filters out suboptimal states, lets the optimization explore a broader region of the solution space, and makes convergence to the ground state more reliable and less sensitive to initialization. The method is demonstrated on the two-site Fermi-Hubbard model, where it converges to the exact ground-state energy after pruning five of six initial trial states.

What carries the argument

The load-bearing object is the Boltzmann-weighted mixed state $\rho(T) = \frac{1}{Z}\sum_k \exp(-\varepsilon_k/T)|\psi_k(\theta_k)\rangle\langle\psi_k(\theta_k)|$ with partition function $Z$, together with a virtual annealing schedule that lowers $T$ (from $kT=25$ by 5% per step). This construction turns measurement-shot allocation into a resource-allocation policy: candidate $k$ receives $N_s \exp(-\varepsilon_k/T)/Z$ of the total $N_s$ shots, so the optimization automatically spends more measurements on promising states as the temperature drops. The variational principle $\mathrm{tr}(\rho H) \ge E_{gs}$ guarantees the mixed-state energy is an upper bound on the ground-state energy, so the annealing process can only improve the bound. The demonstration also uses a reduced subspace fixed by particle-number conservation (half filling), a penalty term to enforce the particle number, and a hardware-efficient ansatz whose initial states are eigenstates of the hopping term obtained by Fourier transformation.

What would settle it

Run KD-VQE on a Hamiltonian whose known ground state is reached only by a trial wavefunction that starts with the highest initial energy of the set; under the paper's schedule (kT=25, 5% decrease per step), if that candidate is pruned before its energy estimate drops below the others, the central claim of reliable convergence to the global minimum fails.

Watch

Extended reading notes

Core claim

KD-VQE's central claim is that a VQE optimizer need not commit to a single trial wavefunction. Instead, a mixed state of several candidates, weighted by Boltzmann factors $\exp(-\varepsilon_k/T)/Z$ at a virtual temperature $T$, serves as the ansatz; each candidate is optimized with standard gradient descent while the shot budget is reallocated in proportion to those weights. As $T$ is annealed from a high value toward zero, the algorithm progressively removes candidates whose estimated energies are high, pruning them once their shot allocation falls below a threshold, and finally condenses onto the surviving state. On the half-filled two-site Fermi-Hubbard model with six initial states, the method prunes the five states that start at or relax to higher energies and converges to the exact ground-state energy of $-1.56$ with all shots assigned to the best candidate $\psi_{VI}$.

Load-bearing premise

The load-bearing assumption is that the virtual annealing schedule and the hand-selected initial wavefunctions ensure the best candidate is not pruned before it reaches the global minimum, and no procedure is given for choosing either on a new problem.

Editorial extensions

If this is right

  • A user of KD-VQE only needs to allocate shots by Boltzmann weights and anneal the temperature; no new ansatz or circuit structure is required.
  • The variational principle for mixed states means the energy estimate produced by KD-VQE remains a valid upper bound on the ground-state energy throughout the annealing process.
  • KD-VQE remains compatible with extensions developed for standard VQE, so improvements such as better optimizers or error mitigation can be applied after the annealing phase.
  • In the demonstrated two-site Hubbard case, KD-VQE filters out all five suboptimal candidates and concentrates all shots on the state that reaches the exact ground-state energy.

Reading between the lines

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

  • A natural extension the paper leaves implicit is an adaptive schedule: choosing the temperature drop from the spread of estimated energies would remove the hand-tuned $kT=25$, 5%-per-step choice and could make the method portable to new problems.
  • Because the method is essentially simulated annealing over a discrete set of ansatze, it could be connected to population-based VQE strategies; a formal comparison of shot overhead versus single-ansatz VQE would clarify when the broader search is worth the cost.
  • On noisy hardware, energy estimates fluctuate more, which could cause premature pruning of the true best candidate; a sensitivity analysis of the pruning threshold would be a concrete test of the method's robustness.
  • The two-site Hubbard demonstration uses only six initial states; testing on a system where the ground state is not representable by any single initial ansatz would probe whether the annealing procedure can still steer the ensemble correctly.
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 / 5 minor

Summary. The paper proposes KD-VQE, a variational quantum eigensolver variant that maintains several trial wavefunctions concurrently, assigns measurement shots according to a Boltzmann distribution with a virtual temperature, and anneals the temperature so that resources concentrate on the lowest-energy candidate. The method is illustrated on the half-filled two-site Fermi-Hubbard model with six initial trial wavefunctions and a total of 10^4 shots, reporting convergence of one candidate to the exact ground-state energy -1.56. The paper claims that, compared with standard VQE, KD-VQE explores a broader region of solution space and offers improved convergence and reliability.

Significance. If the claims were supported, KD-VQE would be a simple, drop-in heuristic for reducing initialization sensitivity in VQE, and the connection to knowledge distillation is a nice conceptual analogy. The paper's Eq. (3) is a correct variational bound for the mixed-state expectation value, and the idea of dynamic shot reallocation is intuitive. However, the current evidence is limited to a single noiseless simulation on a four-qubit model with hand-picked parameters, and no comparison to standard VQE is provided. The significance of the paper will depend on whether the authors can supply a fair comparative study and a robustness analysis of the annealing procedure under finite-shot noise.

major comments (4)
  1. [Abstract; Section III, Fig. 2] The central claim of improved convergence and increased reliability relative to standard VQE is not supported by any direct comparison. The numerical section reports only the trajectories of the six trial states and the shot allocation for a single run. A fair baseline would use the same hardware-efficient ansatz, the same total shot budget, and the same number of runs for standard VQE with (i) the best initial state, (ii) random initializations, and (iii) all six initial states, reporting the distribution of final energies. Without such a baseline, the comparative statements in the abstract and conclusions are assertions, not demonstrated results.
  2. [Section II, after Eq. (3)] The sentence 'Eq.(3) ensures that KD-VQE can converge to the ground state through the optimization process' overstates what the inequality proves. Eq. (3) is a valid variational upper bound for the exact expectation value of the instantaneous mixed state, but it does not constrain the shot-reallocation dynamics, the annealing schedule, or the influence of finite measurement noise on the Boltzmann weights. As T decreases, the ensemble condenses onto the candidate with the currently lowest estimated energy, which need not be the global minimum. The role of Eq. (3) should be restated as a per-iteration bound rather than a convergence guarantee.
  3. [Section III, 'The virtual annealing schedule...' paragraph] The annealing schedule and pruning threshold are chosen by hand (kT=25, 5% reduction per step, pruning below 100 shots), and no criterion or sensitivity study is provided. The paper explicitly acknowledges 'premature condensation' as a failure mode, but does not quantify its probability. Because the Boltzmann weights are computed from the same noisy energy estimates that the algorithm is meant to improve, a candidate with a favorable early noise fluctuation can capture most of the 10^4 shots and survive pruning while the true ground-state candidate is starved. To support the reliability claim, the authors should report the success probability over many independent noise realizations and show how it depends on the initial temperature, the cooling rate, and the pruning threshold.
  4. [Section III, Fig. 2] The numerical demonstration consists of a single run on one Hamiltonian instance (t=1, U=1) at half filling. This is insufficient to establish the claimed reliability and generalizability. At minimum, the authors should provide statistics over repeated runs with different shot-noise realizations and over a small range of U/t values, and ideally compare against standard VQE under the same conditions.
minor comments (5)
  1. [Footnote 1] The word 'Emial' should be 'Email'.
  2. [Section III, paragraph beginning 'As depicted in'] The phrase 'As depicted in n Fig.(2d)' contains a typographical error ('n'); it should read 'As depicted in Fig.(2d)'.
  3. [Figure 2 caption] The caption does not state which of panels (d)-(g) corresponds to which trial state; please label the panels or state the correspondence explicitly in the caption.
  4. [Equation (5) and surrounding text] The penalty coefficient λ and the learning rate η are introduced but their numerical values are never specified; please provide them.
  5. [Section III, paragraph after Eq. (5)] The sentence 'εIV are εV are trapped around 0' is ungrammatical, and 'ψII and ψIIIexhibit' is missing a space between 'ψIII' and 'exhibit'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the variational bound is standard and the benchmark energy is external; the hand-picked run and overclaimed guarantees are correctness concerns, not circular reasoning.

full rationale

The paper's central derivation is not circular. Equation (3) restates the standard variational principle for a mixed state, E_gs <= tr(rho H); it is a valid inequality for any density matrix and does not presuppose the conclusion that the algorithm reaches the global minimum. The Boltzmann reallocation of shots among trial wavefunctions is an algorithmic update rule, not a prediction derived from itself, and the exact ground-state energy (-1.56) of the two-site Fermi-Hubbard model is an external benchmark. The six trial states and the annealing schedule (kT = 25, 5% reduction) are hand-selected inputs, but they are not fitted to the target energy, so the demonstration is not a fitted-input-called-prediction. The paper's cited self-references ([38] and [41]) are peripheral technical citations for qubit mapping and Fourier diagonalization, not load-bearing justifications of the main claim. The statement that Eq. (3) 'ensures that KD-VQE can converge to the ground state' is an overstatement, and the absence of a rigorous guarantee against premature condensation or a comparison with standard VQE is a support/correctness weakness, but these are not circularity. No step in the paper reduces by construction to its own inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central demonstration is a self-contained numerical experiment, not a derivation. It rests on standard quantum mechanics (the variational principle for mixed states), on the expressibility of the hardware-efficient ansatz, on the accuracy of finite-shot energy estimates used to rank and prune trial states, and on a hand-chosen annealing schedule whose failure modes the paper itself describes. No new physical entities are introduced; the virtual temperature is an algorithmic control parameter.

free parameters (6)
  • Initial virtual temperature kT = 25
    Hand-selected starting temperature for the Boltzmann distribution over trial wavefunctions; it controls the initial degree of mixing and is not derived from any criterion.
  • Temperature reduction per step = 5%
    The virtual temperature is reduced by 5% after each optimization step; the paper itself notes that the schedule is critical and that a wrong choice causes premature condensation or wasted resources.
  • Total measurement shots = 10^4
    The total shot budget across all trial wavefunctions; chosen by hand and affects the accuracy of every energy estimate used for reallocation.
  • Pruning threshold = 100 shots
    Trial wavefunctions receiving fewer than 100 shots are removed from the pool; this threshold is introduced without justification and affects which candidates survive.
  • Penalty coefficient lambda
    Appears in Eq. (5) to enforce half filling; its numerical value is not reported and it controls how strongly the optimization is confined to the correct particle-number sector.
  • Learning rate eta
    Used in the gradient descent update theta -> theta - eta * grad; the value is not reported, and the paper only says KD-VQE benefits from a relatively large learning rate.
assumptions (5)
  • standard math Variational principle for mixed states: for any density matrix rho, E_gs <= tr(rho H).
    Invoked in Eq. (3) to justify that the mixed-state energy upper-bounds the ground-state energy; this is a standard quantum mechanics result.
  • domain assumption The hardware-efficient ansatz can express the ground state within the reduced subspace.
    The demonstration requires that at least one trial wavefunction, psi_VI, can be evolved by the ansatz to the exact ground state; no expressibility analysis is provided.
  • domain assumption Finite-shot energy estimates are accurate enough to rank trial states for Boltzmann reallocation and pruning.
    The reallocation formula uses estimated energies epsilon_k, and the pruning rule removes states with fewer than 100 shots; shot noise can misrank states, and no statistical analysis is given.
  • ad hoc to paper The hand-chosen annealing schedule (kT=25, 5% reduction per step) avoids premature condensation and local-minimum trapping.
    The paper itself says the schedule is critical and describes failure modes for both too-fast and too-slow cooling, yet no procedure is given for choosing the schedule.
  • domain assumption The penalty method with coefficient lambda keeps the optimization in the half-filling subspace.
    Eq. (5) adds a penalty term lambda * |<psi|n|psi> - 2|, which is assumed to be strong enough to enforce particle-number conservation; the value of lambda is not reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Knowledge Distillation Inspired Variational Quantum Eigensolver with Virtual Annealing." pith.science (2026). https://pith.science/paper/NJKWPISS

@misc{pith2026250503998,
  author       = {Pith},
  title        = {Pith review of: Knowledge Distillation Inspired Variational Quantum Eigensolver with Virtual Annealing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NJKWPISS}},
  note         = {Machine review of arXiv:2505.03998}
}
read the original abstract

In this paper, we propose a Knowledge Distillation Inspired Variational Quantum Eigensolver (KD-VQE). Inspired by the virtual distillation process in knowledge distillation (KD), KD-VQE introduces a virtual annealing mechanism to the variational quantum eigensolver (VQE) framework. In KD-VQE, measurement resources (shots) are dynamically allocated among multiple trial wavefunctions, each weighted according to a Boltzmann distribution with a virtual temperature. As the temperature decreases gradually, the algorithm progressively reallocates resources toward lower-energy candidates, effectively filtering out suboptimal states and steering the system toward the global minimum. Moreover, we demonstrate the effectiveness of KD-VQE by applying it to the two-site Fermi-Hubbard model. Compared to standard VQE framework, KD-VQE explores a broader region of the solution space, and offers improved convergence behavior and increased reliability.

Figures

Figures reproduced from arXiv: 2505.03998 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Quantum Algorithm Software for Condensed Matter Physics

    cond-mat.str-el 2025-06 reject novelty 2.0 of 10

    A review of quantum algorithm software that advertises a benchmark suite, yet the body contains no benchmarks, data, or code.

Reference graph

Works this paper leans on

48 extracted references · 41 canonical work pages · cited by 1 Pith paper

  1. [1]

    The first step of KD-VQE is to examine the conservation laws and symmetry properties of the given Hamil- tonian H

    Identification of the Reduced Subspace. The first step of KD-VQE is to examine the conservation laws and symmetry properties of the given Hamil- tonian H. In KD techniques, pre-trained ‘teacher’ models possess a greater knowledge capacity than the smaller ‘student’ models. However, in most practical applications, it is unnecessary to fully ex- ploit this ...

    work page

  2. [2]

    Once the reduced subspace is determined, the next step is to construct the trial wavefunctions and prepare the mixed state according to the Boltzmann distri- bution

    Preparation of the Mixed State Ansatz. Once the reduced subspace is determined, the next step is to construct the trial wavefunctions and prepare the mixed state according to the Boltzmann distri- bution. Initially, several candidate solutions within the reduced subspace are selected, serving as ini- tial guesses for KD-VQE. A practical approach is to ide...

    work page

  3. [3]

    The optimization process in KD-VQE follows the standard VQE method

    Variational Optimization along with Virtual Annealing. The optimization process in KD-VQE follows the standard VQE method. For each trial wavefunction|ψk(θk)⟩, the variational parameters θk are independently updated by minimizing the ex- pectation value⟨ψk(θk)|H|ψk(θk)⟩. The standard VQE framework employs the variational principle to approximate the groun...

    work page

  4. [4]

    Peruzzo, J

    A. Peruzzo, J. McClean, P. Shadbolt, M.-H. Yung, X.-Q. Zhou, P. J. Love, A. Aspuru-Guzik, and J. L. O’brien, A variational eigenvalue solver on a pho- tonic quantum processor, Nature communications 5, 4213 (2014)

    work page 2014

  5. [5]

    J. R. McClean, J. Romero, R. Babbush, and A. Aspuru-Guzik, The theory of variational hy- brid quantum-classical algorithms, New Journal of Physics 18, 023023 (2016)

    work page 2016

  6. [6]

    Preskill, Quantum computing in the nisq era and beyond, Quantum 2, 79 (2018)

    J. Preskill, Quantum computing in the nisq era and beyond, Quantum 2, 79 (2018)

    2018

  7. [7]

    Bittel and M

    L. Bittel and M. Kliesch, Training variational quan- tum algorithms is np-hard, Physical review letters 127, 120502 (2021)

    2021

  8. [8]

    P. J. O’Malley, R. Babbush, I. D. Kivlichan, J. Romero, J. R. McClean, R. Barends, J. Kelly, P. Roushan, A. Tranter, N. Ding,et al. , Scalable quantum simulation of molecular energies, Physical Review X 6, 031007 (2016)

    work page 2016

Show all 48 references
  1. [9]

    Hempel, C

    C. Hempel, C. Maier, J. Romero, J. McClean, T. Monz, H. Shen, P. Jurcevic, B. P. Lanyon, P. Love, R. Babbush, et al. , Quantum chemistry calculations on a trapped-ion quantum simulator, Physical Review X8, 031022 (2018)

    2018

  2. [10]

    Y. Cao, J. Romero, J. P. Olson, M. Degroote, P. D. Johnson, M. Kieferová, I. D. Kivlichan, T. Menke, B. Peropadre, N. P. Sawaya,et al., Quantum chem- istry in the age of quantum computing, Chemical reviews 119, 10856 (2019)

    2019

  3. [11]

    R. M. Parrish, E. G. Hohenstein, P. L. McMa- hon, and T. J. Martínez, Quantum computation of electronic transitions using a variational quan- tumeigensolver,Physicalreviewletters 122,230401 (2019)

    2019

  4. [12]

    G. A. Quantum, Collaborators* †, F. Arute, K. Arya, R. Babbush, D. Bacon, J. C. Bardin, R. Barends, S. Boixo, M. Broughton, B. B. Buck- ley, et al., Hartree-fock on a superconducting qubit quantum computer, Science369, 1084 (2020)

    2020

  5. [13]

    Cerezo, A

    M. Cerezo, A. Arrasmith, R. Babbush, S. C. Ben- jamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio,et al., Variational quantum al- gorithms, Nature Reviews Physics3, 625 (2021)

    2021

  6. [14]

    Tilly, H

    J. Tilly, H. Chen, S. Cao, D. Picozzi, K. Setia, Y. Li, E. Grant, L. Wossnig, I. Rungger, G. H. Booth, et al. , The variational quantum eigensolver: a re- viewofmethodsandbestpractices,PhysicsReports 986, 1 (2022)

    2022

  7. [15]

    Sajjan, J

    M. Sajjan, J. Li, R. Selvarajan, S. H. Sureshbabu, S. S. Kale, R. Gupta, V. Singh, and S. Kais, Quan- tum machine learning for chemistry and physics, Chemical Society Reviews51, 6475 (2022)

    2022

  8. [16]

    H. Ma, J. Liu, H. Shang, Y. Fan, Z. Li, and J. Yang, Multiscale quantum algorithms for quantum chem- istry, Chemical Science14, 3190 (2023)

    2023

  9. [17]

    S. Guo, J. Sun, H. Qian, M. Gong, Y. Zhang, F. Chen, Y. Ye, Y. Wu, S. Cao, K. Liu,et al., Ex- perimental quantum computational chemistry with optimized unitary coupled cluster ansatz, Nature Physics 20, 1240 (2024)

    2024

  10. [18]

    Kim, K.-M

    B. Kim, K.-M. Hu, M.-H. Sohn, Y. Kim, Y.-S. Kim, S.-W. Lee, and H.-T. Lim, Qudit-based variational quantum eigensolver using photonic orbital angular momentum states, Science Advances10, eado3472 (2024)

    2024

  11. [19]

    L. H. Delgado-Granados, T. J. Krogmeier, L. M. Sager-Smith, I. Avdic, Z. Hu, M. Sajjan, M. Ab- basi, S. E. Smart, P. Narang, S. Kais,et al., Quan- tum algorithms and applications for open quantum systems, Chemical Reviews125, 1823 (2025). 6

    2025

  12. [20]

    Kandala, A

    A. Kandala, A. Mezzacapo, K. Temme, M. Takita, M. Brink, J. M. Chow, and J. M. Gambetta, Hardware-efficient variational quantum eigensolver for small molecules and quantum magnets, nature 549, 242 (2017)

    2017

  13. [21]

    Kokail, C

    C. Kokail, C. Maier, R. van Bijnen, T. Brydges, M. K. Joshi, P. Jurcevic, C. A. Muschik, P. Silvi, R.Blatt, C.F.Roos, et al.,Self-verifyingvariational quantum simulation of lattice models, Nature569, 355 (2019)

    2019

  14. [22]

    Uvarov, J

    A. Uvarov, J. D. Biamonte, and D. Yudin, Varia- tional quantum eigensolver for frustrated quantum systems, Physical Review B102, 075104 (2020)

    2020

  15. [23]

    Mizuta, M

    K. Mizuta, M. Fujii, S. Fujii, K. Ichikawa, Y. Ima- mura, Y. Okuno, and Y. O. Nakagawa, Deep varia- tional quantum eigensolver for excited states and its application to quantum chemistry calculation of periodic materials, Physical Review Research3, 043121 (2021)

    2021

  16. [24]

    Gyawali and M

    G. Gyawali and M. J. Lawler, Adaptive variational preparation of the fermi-hubbard eigenstates, Phys- ical Review A105, 012413 (2022)

    2022

  17. [25]

    Anselme Martin, P

    B. Anselme Martin, P. Simon, and M. J. Rančić, Simulating strongly interacting hubbard chains with the variational hamiltonian ansatz on a quan- tum computer, Physical Review Research4, 023190 (2022)

    2022

  18. [26]

    Kattemölle and J

    J. Kattemölle and J. Van Wezel, Variational quan- tum eigensolver for the heisenberg antiferromag- net on the kagome lattice, Physical Review B106, 214429 (2022)

    2022

  19. [27]

    Wecker, M

    D. Wecker, M. B. Hastings, and M. Troyer, Progress towards practical quantum variational algorithms, Physical Review A92, 042303 (2015)

    2015

  20. [28]

    J. R. McClean, S. Boixo, V. N. Smelyanskiy, R. Babbush, and H. Neven, Barren plateaus in quantum neural network training landscapes, Na- ture communications 9, 4812 (2018)

    2018

  21. [29]

    D.Wierichs, C.Gogolin,andM.Kastoryano,Avoid- ing local minima in variational quantum eigen- solvers with the natural gradient optimizer, Physi- cal Review Research2, 043246 (2020)

    2020

  22. [30]

    Hinton, O

    G. Hinton, O. Vinyals, and J. Dean, Distilling the knowledge in a neural network, arXiv preprint arXiv:1503.02531 (2015)

    2015 arXiv

  23. [31]

    G. Chen, W. Choi, X. Yu, T. Han, and M. Chan- draker, Learning efficient object detection models with knowledge distillation, Advances in neural in- formation processing systems30 (2017)

    2017

  24. [32]

    J. H. Cho and B. Hariharan, On the efficacy of knowledge distillation, in Proceedings of the IEEE/CVF international conference on computer vision (2019) pp. 4794–4802

    2019

  25. [33]

    Cheng, Z

    X. Cheng, Z. Rao, Y. Chen, and Q. Zhang, Ex- plaining knowledge distillation by quantifying the knowledge, in Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition (2020) pp. 12925–12935

    2020

  26. [34]

    Ji and Z

    G. Ji and Z. Zhu, Knowledge distillation in wide neural networks: Risk bound, data efficiency and imperfect teacher, Advances in Neural Information Processing Systems 33, 20823 (2020)

    2020

  27. [35]

    Y. Gu, L. Dong, F. Wei, and M. Huang, Minillm: Knowledge distillation of large language models, arXiv preprint arXiv:2306.08543 (2023)

    2023 arXiv

  28. [36]

    A. Liu, B. Feng, B. Xue, B. Wang, B. Wu, C. Lu, C. Zhao, C. Deng, C. Zhang, C. Ruan, et al., Deepseek-v3 technical report, arXiv preprint arXiv:2412.19437 (2024)

    2024 arXiv

  29. [37]

    J. Gou, B. Yu, S. J. Maybank, and D. Tao, Knowl- edge distillation: A survey, International Journal of Computer Vision 129, 1789 (2021)

    2021

  30. [38]

    M. S. Jattana, F. Jin, H. De Raedt, and K. Michielsen, Improved variational quantum eigen- solver via quasidynamical evolution, Physical Re- view Applied 19, 024047 (2023)

    2023

  31. [39]

    Skogh, O

    M. Skogh, O. Leinonen, P. Lolur, and M. Rahm, Accelerating variational quantum eigensolver con- vergenceusingparametertransfer,ElectronicStruc- ture 5, 035002 (2023)

    2023

  32. [40]

    Hubbard, Electron correlations in narrow energy bands, Proceedings of the Royal Society of London

    J. Hubbard, Electron correlations in narrow energy bands, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences276, 238 (1963)

    1963

  33. [41]

    Li, Iterative method to improve the precision of the quantum-phase-estimation algorithm, Physical Review A 109, 032606 (2024)

    J. Li, Iterative method to improve the precision of the quantum-phase-estimation algorithm, Physical Review A 109, 032606 (2024)

    2024

  34. [42]

    Verstraete, J

    F. Verstraete, J. I. Cirac, and J. I. Latorre, Quan- tum circuits for strongly correlated quantum sys- tems, Physical Review A—Atomic, Molecular, and Optical Physics 79, 032316 (2009)

    2009

  35. [43]

    Cervera-Lierta, Exact ising model simulation on a quantum computer, Quantum2, 114 (2018)

    A. Cervera-Lierta, Exact ising model simulation on a quantum computer, Quantum2, 114 (2018)

    2018

  36. [44]

    J. Li, B. A. Jones, and S. Kais, Toward perturbation theory methods on a quantum computer, Science Advances 9, eadg4576 (2023)

    2023

  37. [45]

    Mitarai, T

    K. Mitarai, T. Yan, and K. Fujii, Generalization of the output of a variational quantum eigensolver by parameter interpolation with a low-depth ansatz, Physical Review Applied11, 044087 (2019)

    2019

  38. [46]

    Tilly, G

    J. Tilly, G. Jones, H. Chen, L. Wossnig, and E. Grant, Computation of molecular excited states on ibm quantum computers using a discriminative variational quantum eigensolver, Physical Review A 102, 062425 (2020)

    2020

  39. [47]

    Tilly, P

    J. Tilly, P. Sriluckshmy, A. Patel, E. Fontana, I. Rungger, E. Grant, R. Anderson, J. Tennyson, and G. H. Booth, Reduced density matrix sampling: Self-consistent embedding and multiscale electronic structure on current generation quantum comput- ers, Physical Review Research3,...

    2021

  40. [48]

    Kuroiwa and Y

    K. Kuroiwa and Y. O. Nakagawa, Penalty meth- ods for a variational quantum eigensolver, Physical Review Research 3, 013197 (2021)

    2021

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.