Exploring fixed points and eigenstates of quantum systems with reinforcement learning

Denis Lacroix; Jes\'us Casado-Pascual; Mar\'ia Laura Olivera-Atencio

REVIEW 3 major objections 4 minor 80 references

Exploring fixed points and eigenstates of quantum systems with reinforcement learning

T0 review · 3 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read This paper presents a reinforcement-learning algorithm that learns the unitary mapping a computational basis onto the fixed-point basis of a quantum operation, which for Hamiltonian evolution is the eigenbasis.

desk verdict A clear, well-explained numerical methods paper with a genuinely new simultaneous-eigenbasis RL construction, but the general convergence claim rests on an unproven tau-sampling heuristic that can fail on near-degenerate gaps. read the letter →

arxiv 2511.17491 v2 pith:SJFC7JQE submitted 2025-11-21 quant-ph cond-mat.str-elnucl-th

María Laura Olivera-Atencio , Jesús Casado-Pascual , Denis Lacroix This is my paper

classification quant-phcond-mat.str-elnucl-th PACS 03.67.-a03.67.Lx

keywords quantumreinforcementlearningfixedpointsoperationHamiltonianeigenstatesunitarytransverse-fieldspinmodelpairingpost-selection

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper introduces a reinforcement-learning algorithm that, given a quantum operation, learns the unitary transformation sending the computational basis onto the operation's fixed-point basis. When the operation is unitary evolution generated by a Hamiltonian, the fixed points are the Hamiltonian eigenstates, so the algorithm produces the full eigenbasis in one pass rather than state by state. Learning is driven purely by projective measurements and a reward-punishment rule that shrinks or grows random two-level rotations. On random two- and three-qubit Hamiltonians, a transverse-field spin chain, and an all-to-all pairing model, the algorithm attains mean fidelities around 0.96–0.99, and a variance-based post-selection keeps accurate eigenstates even before full convergence. A reader should care because this is a non-variational, measurement-only route to eigenstates that also exposes hidden symmetries.

What carries the argument

The learning engine is the recurrence D_{k+1}=D_k ∏_{j<l}D^{(j,l)}_k, where each D^{(j,l)}_k is a rotation in the subspace of computational basis states |j⟩ and |l⟩. After applying D_k, evolving with the quantum operation, and undoing D_k, the algorithm measures in the computational basis: an outcome that returns the original label is rewarded by shrinking that pair's exploration width by r², while a mismatch expands it by p²; the rotation angles are drawn from intervals set by that width. Randomly sampling the evolution time τ at each iteration protects against converging to resonant superpositions that are invariant under a single fixed τ but are not eigenstates, and a fine-tuning stage th

What would settle it

Run the unrestricted algorithm on a Hamiltonian with a known exact degeneracy, as in the four-qubit pairing case, and test every converged output state with the paper's own zero-variance condition, σ_j = 0, while also checking that the learned states span the Hilbert space; a converged run whose states have σ_j far above zero would show the reward rule rewarded a non-fixed point, contradicting the central claim.

Watch

Extended reading notes

Core claim

The central claim is that the fixed-point basis of a quantum operation can be learned concurrently by iterating a unitary D_k and using computational-basis measurement outcomes to reward or penalize rotations on each two-dimensional subspace spanned by basis states, so that in the limit D_k|j⟩ approximates |Φ_α⟩. For Hamiltonian evolution U(τ)=e^{-iτH/ℏ}, the fixed points are the eigenstates, and the algorithm returns the whole eigenbasis at once. The paper demonstrates this numerically for random Hamiltonians on two and three qubits, for the transverse-field spin model up to four qubits, and for the all-to-all pairing Hamiltonian up to six qubits (with symmetry-restricted learning), reporti

Load-bearing premise

The central claim rests on the unproved heuristic that randomly sampling the evolution time from a hand-chosen interval prevents convergence to spurious invariant superpositions, plus the unproved convergence of the stochastic rotation-update rule; both are supported only by numerical examples.

Editorial extensions

If this is right

The full set of fixed points—and hence the entire eigenbasis—can be learned in parallel using only the ability to apply the operation and measure in a fixed basis, so no variational ansatz or gradient is required.
Known symmetries can be used to restrict learning to a single sector, cutting the search dimension and extending the method to larger systems (demonstrated on six qubits for the pairing Hamiltonian).
A post-selection threshold on energy variance identifies high-accuracy eigenstates even when the algorithm has not fully converged for all states.
The algorithm reveals hidden symmetries without being told about them, visible as separate convergence timescales tied to the dimensions of the invariant subspaces.
The construction is formulated for general quantum operations, not only unitary ones, so the same reward-penalty loop is a candidate for finding fixed points of dissipative dynamics.

Reading between the lines

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

The random-τ heuristic is effectively an energy filter: it prevents spurious superpositions by varying the phase accumulated by each eigenstate. A deterministic schedule that adapts the τ distribution to measured energy variances could replace it and make the convergence argument provable.
Because the algorithm never needs the Hamiltonian matrix, only the ability to evolve and measure, it is a plausible primitive for black-box quantum systems—such as unknown or noisy devices—where the fixed-point basis is the information-preserving basis.
The per-pair rotation structure hints at an O(d²) per-iteration circuit cost; restricting to symmetry sectors or to sparse graphs among basis states is the natural next scaling test.
The energy-variance post-selection could be promoted from a final filter to an online stopping rule during learning, since it needs no knowledge of the exact eigenstates.

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.

Desk Editor's Note

read the letter

The paper is worth your time if you work on quantum eigensolvers or RL for quantum tasks. What's actually new: instead of learning eigenstates one at a time as in Albarrán-Arriagada et al., the algorithm builds a single global unitary mapping the whole computational basis onto the full set of fixed points of a quantum operation. That's a real structural difference, and the applications to the TFIM and the pairing Hamiltonian, the symmetry-restricted version, and the energy-fluctuation post-selection are all reasonable additions. The algorithm description is detailed enough to re-implement, and the reported fidelities support the claims for the tested systems: random 2–3 qubit Hamiltonians, TFIM up to 4 qubits, pairing up to 6 qubits. The authors also honestly flag the degeneracy issue in the pairing model. Credit where due: this is a competent numerical study, not an overclaiming one.

The soft spots are real but not fatal. The biggest one is the tau-sampling heuristic in Sec. II D. The paper correctly identifies that fixed tau can make superpositions of eigenstates invariant, and randomizing tau over an interval mitigates this. But the stress-test concern lands: if your interval's upper bound is far below 2π/ω_min, then for every sampled tau the survival probability in Eq. (15) is within O((ω τ_max)^2) of 1, so the reward signal is too weak to resolve near-degenerate components and the exploration weights collapse. The paper admits tau_max can only be 'roughly estimated' when the spectrum is unknown, so this is a genuine correctness risk for the general case, not just a statistical fluctuation. The demonstrations choose [0,600] for rescaled spectra and work, but that doesn't prove the heuristic.

Other issues: no code or data, several hand-picked hyperparameters (r, p, w_th, w_r, the tau interval, the reset schedule), no convergence proof, and the cleanest energy plots in Fig. 10 rely on post-selection with a fluctuation threshold. The mean fidelity curves also lack error bars. The typo in Eq. (15) — β runs from j+1 instead of α+1 — is minor but should be fixed.

Who is this for? Researchers working on hybrid quantum-classical algorithms, especially those interested in non-variational alternatives to VQE or in RL-based state preparation. It's not a breakthrough that changes the landscape — the system sizes are classically trivial — but it's a legitimate incremental contribution with a clearly described algorithm and honest benchmarking.

I'd send it to peer review. The referee should ask for code/data and a more careful treatment of the tau-sampling failure mode, perhaps with a diagnostic or adaptive scheme. But the paper is coherent, the central idea is new enough, and the numerical evidence is consistent with the claims as stated.

Referee Report

3 major / 4 minor

Summary. The paper introduces a reinforcement-learning algorithm aimed at finding the unitary transformation D that maps the computational basis onto the basis of pure fixed points of a quantum operation E. For Hamiltonian time evolutions E(ρ)=U(τ)ρU†(τ), this task is equivalent to finding the eigenbasis. The algorithm updates D through pairwise rotations in the computational basis, using measurement outcomes on d simultaneously processed qudits to reward or punish exploration steps; the exploration widths w^{(j,l)} shrink on rewards and expand on penalties. The authors benchmark the method on random two- and three-qubit Hamiltonians, on the transverse-field Ising model up to four qubits, and on the Richardson pairing Hamiltonian up to five qubits in the full space and six qubits with symmetry-restricted learning. They report mean fidelities around 0.96–0.99, and they propose an energy-variance-based post-selection to improve the final energy estimates.

Significance. If the results hold, the algorithm offers a distinct, non-variational route to computing eigenstates and fixed points of quantum operations, with the appealing feature of obtaining all vectors simultaneously rather than sequentially. The symmetry-sector adaptation and the post-selection idea are useful contributions. The numerical demonstrations are suggestive, but the central claim in Eq. (2) is supported only by heuristics, especially the random-time sampling in Sec. II.D and the stochastic update rule. The paper is honest about these limitations, yet the load-bearing assumptions need quantitative validation before the general claim can be accepted.

major comments (3)

[Sec. II.D, Eq. (15)] The convergence to the true fixed-point basis depends critically on the claim that randomly sampling τ prevents convergence to spurious invariant superpositions. No proof or quantitative criterion is given. For a superposition of two eigenstates with energy gap ℏω, Eq. (15) gives survival probability P = 1 − 4|c_α|^2|c_β|^2 sin²(ωτ/2). If τ is sampled uniformly from [0,T] with T ≪ 2π/ω_min, then P ≥ 1 − (ω_min T/2)² for all sampled τ, so rewards occur with probability nearly one and w^{(j,l)} shrinks before the near-degenerate components are resolved. The manuscript itself states that τ_max can only be “roughly estimated” when the spectrum is unknown (Sec. II.D). Thus an unlucky interval can lock the algorithm onto a non-eigenstate superposition. Please add either a provable condition, a concrete failure test, or a numerical stress test with a small-gap Hamiltonian (e.g., with ω_min T ≪
[Sec. III.D, Fig. 10] The post-selection result is presented as a way to extract high-fidelity states, but the filter is based on the energy variance σ defined in Eq. (21), which is zero by definition only for exact eigenstates. The improvement of the energy estimates in Fig. 10 is therefore partly tautological: states with small σ are, by construction, close to eigenstates in the variance sense. To support the abstract’s claim of post-selecting high-fidelity states, the authors should compare the selected states directly with exact eigenstates (e.g., by overlap) or explicitly state that the post-selection criterion is for energy estimation only. The issue is compounded in degenerate subspaces, where a low-variance state may have no meaningful overlap with the particular eigenbasis used in Eq. (16).
[Sec. II.E, Sec. II.G] The convergence of the stochastic reward-penalty update and the benefit of the reset/fine-tuning schedule are supported only by numerical examples. The manuscript reports mean fidelities but no spread across the N_r realizations, so it is difficult to assess whether the quoted F_min and F_max are typical or arise from a few favorable runs. Please provide either standard deviations/quantiles for the fidelity curves or an explicit robustness scan over the hand-picked parameters r, p, w_th, w_r, k_0, and k_M. This is important because the central claim of reliable convergence (Eq. (2)) is based on these averages.

minor comments (4)

[Eq. (15)] The second summation is written with lower limit β=j+1, but j is not defined in that expression; it should presumably be β=α+1.
[Figs. 2–6] The text repeatedly states that the color-coded fidelity curves are difficult to distinguish. Using distinct line styles or labeled endpoints would improve readability.
[Sec. II.D and Sec. III.A] The choice of dimensionless interval [0,100] for random Hamiltonians and [0,600] for the physical models is stated but not derived. A brief explanation of how these ranges are selected from the spectral scales would help the reader reproduce the results.
[Data Availability] The statement that data are available ‘upon reasonable request’ is vague; depositing code and data in a public repository would significantly strengthen reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the RL target is defined by fixed-point invariance, while correctness is benchmarked against independently diagonalized eigenstates.

full rationale

The paper's central claim is that the iterative reward–penalty scheme constructs a unitary D_k such that D_k|j> approximates the fixed-point basis of a quantum operation E, and in the Hamiltonian case the eigenbasis of H. This is not circular: the learning signal is the measurement outcome in step 4, which tests whether D_k|j> is invariant under E, i.e. whether the inverse-evolved state returns to the computational basis. Invariance under E(ρ)=U(τ)ρU†(τ) is not definitionally the same as being an eigenstate of H; the paper explicitly identifies the gap in Sec. II D and Eq. (15), showing that superpositions of nondegenerate eigenstates can also be invariant for special or near-resonant τ. The proposed remedy (randomly sampling τ, with τ_max only 'roughly estimated' when the spectrum is unknown) is an acknowledged heuristic and a real correctness/convergence risk, but it is not a circular step: the paper does not define the target basis in terms of the algorithm's output, nor does it fit any parameter to the benchmark fidelities. The accuracy metric in Eq. (16) compares learned states with eigenstates obtained by independent numerical diagonalization, so the reported fidelities are external tests rather than constructed outputs. The post-selection criterion based on energy fluctuations in Eq. (21) uses the necessary eigenstate condition σ_j=0, but it is applied as a filter after convergence and is validated empirically in Fig. 9 by showing that stricter thresholds reduce the distance to independently computed eigenenergies; it is not used inside the learning update or as a fitted target. Hyperparameters such as r, p, w_th, w_r and the sampling interval [0,600] are hand-tuned and problem-specific, but they are not fitted to the exact eigenstates and do not make the prediction equivalent to the input. Self-citations [31–33] are mentioned only as preliminary directions for future non-unitary extensions and are not load-bearing for the central unitary-eigenstate claim; no uniqueness theorem or prior result is invoked to forbid alternatives. Overall, the derivation chain is self-contained against external benchmarks, and no specific reduction of a prediction to its inputs can be exhibited.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The algorithm relies on several hand-set hyperparameters (r, p, w_th, w_r, tau interval, sigma_th, reset schedule) and heuristic assumptions about random tau sampling and convergence of the stochastic rotation updates. No new physical entities are introduced. The target eigenstates are not used during training, so the benchmarks are external to the learning loop.

free parameters (7)

reward rate r = 0.9 (0.93 in Fig. 5)
Controls exploration contraction; chosen by hand, not derived from first principles.
punishment rate p = 2/r
Chosen by hand together with r; affects how quickly exploration expands after penalties.
convergence threshold w_th = 0.005
Hand-picked threshold for declaring convergence of exploration parameters.
reset exploration value w_r = 0.01 or 0.05
Problem-dependent reset value; authors state the optimal value must be adjusted per problem.
evolution time sampling interval = [0,100] for random Hamiltonians, [0,600] for TFIM and pairing
Hand-chosen to avoid spurious fixed points; Sec. II D says tau_max can only be roughly estimated for unknown spectra.
post-selection threshold sigma_th = 0.02
Chosen by inspection in Fig. 8 to filter final states for the energy plots in Fig. 10.
reset schedule parameters k0, kM = not reported
Sec. II E defines the schedule w_r(kM-k)/(kM-k0) but does not give the numerical values used in the simulations.

assumptions (5)

domain assumption The target fixed points form an orthonormal basis of pure states
Sec. II A assumes B={|Phi_alpha>} is an orthonormal basis; the authors note that general quantum operations may have different fixed-point structures and defer that case.
domain assumption The quantum operation E and projective computational-basis measurements can be applied on every qudit
Algorithm steps 2 and 4 require repeated access to E and measurement outcomes; the paper assumes this access throughout.
ad hoc to paper Random sampling of tau prevents convergence to spurious invariant superpositions
Sec. II D argues this heuristic mitigates the problem of near-resonant evolution times, but provides no proof and relies on rough estimates of tau_max.
ad hoc to paper The multiplicative reward-penalty update with random rotations converges to the target unitary
No convergence theorem is given; success is established only through numerical examples.
ad hoc to paper The reset/fine-tuning schedule preserves learned knowledge and improves fidelity
Sec. II E states that w_r should be carefully selected and that the schedule enables refinement; this is an empirical design choice.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Exploring fixed points and eigenstates of quantum systems with reinforcement learning." pith.science (2026). https://pith.science/paper/SJFC7JQE

@misc{pith2026251117491,
  author       = {Pith},
  title        = {Pith review of: Exploring fixed points and eigenstates of quantum systems with reinforcement learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SJFC7JQE}},
  note         = {Machine review of arXiv:2511.17491}
}

read the original abstract

We introduce a reinforcement learning algorithm designed to identify the fixed points of a given quantum operation. The method iteratively constructs the unitary transformation that maps the computational basis onto the basis of fixed points through a reward-penalty scheme based on quantum measurements. In cases where the operation corresponds to a Hamiltonian evolution, this task reduces to determining the Hamiltonian eigenstates. The algorithm is first benchmarked on random Hamiltonians acting on two and three qubits and then applied to many-body systems of up to six qubits, including the transverse-field Ising model and the all-to-all pairing Hamiltonian. In both cases, the algorithm is demonstrated to perform successfully; in the pairing model, it can also reveal hidden symmetries, which can be exploited to restrict learning to specific symmetry sectors. Finally, we discuss the possibility of post-selecting high-fidelity states even when full convergence has not been reached.

Figures

Figures reproduced from arXiv: 2511.17491 by the authors.

**Figure 2.** FIG. 2. Algorithm results for random two-qubit Hamiltonians [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** FIG. 3. Same as Fig. 2, but for the case of three qubits [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: FIG. 4. Results of the algorithm applied to the TFIM Hamil [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: FIG. 5. Results analogous to those shown in Fig. 4, obtained [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

**Figure 6.** Figure 6: To account for particle-number conservation, in [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

**Figure 8.** Figure 8: FIG. 8. Illustration of the evaluation of the energy fluctua [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]

**Figure 9.** Figure 9: FIG. 9. Average distance between the expectation value [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]

**Figure 10.** Figure 10: FIG. 10. Effect of introducing a fluctuation threshold on the [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

80 extracted references · 1 canonical work pages

[1]

M. A. Nielsen and I. L. Chuang,Quantum Computing and Quantum Information(Cambridge University Press, Cambridge, 2000)

2000
[2]

Schuld, I

M. Schuld, I. Sinayskiy, and F. Petruccione, An introduc- tion to quantum machine learning, Contemp. Phys.56, 172 (2015)

2015
[3]

Torlai, G

G. Torlai, G. Mazzola, J. Carrasquilla, M. Troyer, R. Melko, and G. Carleo, Neural-network quantum state tomography, Nat. Phys.14, 447 (2018)

2018
[4]

Biamonte, P

J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, Quantum machine learning, Na- ture549, 195 (2017)

2017
[5]

G. Liu, M. Chen, Y.-X. Liu, D. Layden, and P. Cappel- laro, Repetitive readout enhanced by machine learning, Mach. Learn.: Sci. Technol.1, 015003 (2020)

2020
[6]

Heese, T

R. Heese, T. Gerlach, S. M¨ ucke, S. M¨ uller, M. Jakobs, and N. Piatkowski, Explaining quantum circuits with shapley values: towards explainable quantum machine learning, Quantum Mach. Intell.7, 10.1007/s42484-025- 00254-8 (2025)

work page doi:10.1007/s42484-025- 2025
[7]

T. Haug, C. N. Self, and M. S. Kim, Quantum ma- chine learning of large datasets using randomized mea- surements, Mach. Learn.: Sci. Technol.4, 015005 (2023)

2023
[8]

Perrier, D

E. Perrier, D. Tao, and C. Ferrie, Quantum geometric machine learning for quantum circuits and control, New J. Phys.22, 103056 (2020)

2020

Show all 80 references

[9]

F. A. C´ ardenas-L´ opez, M. Sanz, J. C. Retamal, and E. Solano, Enhanced quantum synchronization via quan- tum machine learning, Adv. Quantum Technol.2, 1800076 (2019)

2019
[10]

Saini, P

S. Saini, P. Khosla, M. Kaur, and G. Singh, Quantum driven machine learning, Int. J. Theor. Phys.59, 4013 (2020)

2020
[11]

Innan, O

N. Innan, O. I. Siddiqui, S. Arora, T. Ghosh, Y. P. Ko¸ cak, D. Paragas, A. A. O. Galib, M. A.-Z. Khan, and M. Ben- nai, Quantum state tomography using quantum machine learning, Quantum Mach. Intell.6(2024)

2024
[12]

Sutton and A

R. Sutton and A. Barto,Reinforcement Learning: An In- troduction., 2nd ed., Adaptative Computation and Ma- chine Learning (MIT Press, Cambridge, 2018)

2018
[13]

Shrapnel, F

S. Shrapnel, F. Costa, and G. Milburn, Quantum Marko- vianity as a supervised learning task, Int. J. Quantum Info.16, 1840010 (2018)

2018
[14]

Agresti, N

I. Agresti, N. Viggianiello, F. Flamini, N. Spagnolo, A. Crespi, R. Osellame, N. Wiebe, and F. Sciarrino, Pat- tern recognition techniques for boson sampling valida- tion, Phys. Rev. X9, 011013 (2019)

2019
[15]

Youssry, R

A. Youssry, R. J. Chapman, A. Peruzzo, C. Ferrie, and M. Tomamichel, Modeling and control of a reconfigurable photonic circuit using deep learning, Quantum Sci. Tech- nol.5, 025001 (2020)

2020
[16]

I. A. Luchnikov, S. V. Vintskevich, D. A. Grigoriev, and S. N. Filippov, Machine Learning Non-Markovian Quan- tum Dynamics, Phys. Rev. Lett.124, 140502 (2020)

2020
[17]

A. A. Melnikov, L. E. Fedichkin, R.-K. Lee, and A. Alod- jants, Machine learning transfer efficiencies for noisy quantum walks, Adv. Quantum Technol.3, 1900115 (2020)

2020
[18]

Lucian, B

B. Lucian, B. Robert, S. B. De, and E. Damien,Re- inforcement Learning and Dynamic Programming Using Function Approximators, 1st ed. (CRC Press Inc., Boca Raton, FL, USA, 2010)

2010
[19]

Agostinelli, G

F. Agostinelli, G. Hocquet, S. Singh, and P. Baldi, From reinforcement learning to deep reinforcement learning: An overview, inBraverman Readings in Machine Learn- ing. Key Ideas from Inception to Current State: Interna- tional Conference Commemorating the 40th Anniversary of ...

2017
[20]

D. Dong, C. Chen, H. Li, and T.-J. Tarn, Quantum rein- forcement learning, IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics)38, 1207 (2008)

2008
[21]

G. D. Paparo, V. Dunjko, A. Makmal, M. A. Martin- Delgado, and H. J. Briegel, Quantum speedup for active learning agents, Phys. Rev. X4, 031002 (2014)

2014
[22]

Dunjko, J

V. Dunjko, J. M. Taylor, and H. J. Briegel, Quantum- enhanced machine learning, Phys. Rev. Lett.117, 130501 (2016). 14

2016
[23]

Bukov, A

M. Bukov, A. G. R. Day, D. Sels, P. Weinberg, A. Polkovnikov, and P. Mehta, Reinforcement learning in different phases of quantum control, Phys. Rev. X8, 031086 (2018)

2018
[24]

Bukov, Reinforcement learning for autonomous preparation of floquet-engineered states: Inverting the quantum kapitza oscillator, Phys

M. Bukov, Reinforcement learning for autonomous preparation of floquet-engineered states: Inverting the quantum kapitza oscillator, Phys. Rev. B98, 224305 (2018)

2018
[25]

F¨ osel, P

T. F¨ osel, P. Tighineanu, T. Weiss, and F. Marquardt, Re- inforcement learning with neural networks for quantum feedback, Phys. Rev. X8, 031084 (2018)

2018
[26]

Kaldari, S

J. Kaldari, S. Tariq, S. Al-Kuwari, S. Y.-C. Chen, S. Chatzinotas, and H. Shin, Quantum reinforcement learning: Recent advances and future directions (2025), arXiv:2510.14595 [quant-ph]

2025
[27]

Jiang, Y

C. Jiang, Y. Pan, Z.-G. Wu, Q. Gao, and D. Dong, Robust optimization for quantum reinforcement learn- ing control using partial observations, Phys. Rev. A105, 062443 (2022)

2022
[28]

Cherrat, I

E. Cherrat, I. Kerenidis, and A. Prakash, Quantum rein- forcement learning via policy iteration, Quantum Mach. Intell.5, DOI10.1007/s42484-023-00116-1 (2023)

2023
[29]

Y. Liu, C. Xu, and S. Jin, Reinforcement learning for con- tinuous control: A quantum normalized advantage func- tion approach, in2023 IEEE International Conference on Quantum Software (QSW)(2023) pp. 83–87

2023
[30]

Franceschetto and A

G. Franceschetto and A. Ricou, Demonstration of quan- tum projective simulation on a single-photon-based quan- tum computer, Phys. Rev. A110, 062613 (2024)

2024
[31]

M. L. Olivera-Atencio, L. Lamata, M. Morillo, and J. Casado-Pascual, Quantum reinforcement learning in the presence of thermal dissipation, Phys. Rev. E108, 014128 (2023)

2023
[32]

M. L. Olivera-Atencio, L. Lamata, and J. Casado- Pascual, Benefits of open quantum systems for quan- tum machine learning, Adv. Quantum Technol. , 2300247 (2023)

2023
[33]

M. L. Olivera-Atencio, L. Lamata, and J. Casado- Pascual, Impact of amplitude and phase damping noise on quantum reinforcement learning: challenges and op- portunities, Eur. Phys. J. Spec. Top. (2025)

2025
[34]

Albarr´ an-Arriagada, J

F. Albarr´ an-Arriagada, J. C. Retamal, E. Solano, and L. Lamata, Reinforcement learning for semi-autonomous approximate quantum eigensolver, Mach. Learn.: Sci. Technol.1, 015002 (2020)

2020
[35]

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

2016
[36]

McArdle, S

S. McArdle, S. Endo, A. Aspuru-Guzik, S. C. Benjamin, and X. Yuan, Quantum computational chemistry, Rev. Mod. Phys.92, 015003 (2020)

2020
[37]

Motta and J

M. Motta and J. E. Rice, Emerging quantum computing algorithms for quantum chemistry, WIREs Comput. Mol. Sci.12, e1580 (2022)

2022
[38]

Cerezo, A

M. Cerezo, A. Arrasmith, R. Babbush, S. C. Benjamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio, and P. J. Coles, Variational quantum algo- rithms, Nat. Rev. Phys.3, 625 (2021)

2021
[39]

Tilly, H

J. Tilly, H. Chen, S. Cao, D. Picozzi, K. Setia, Y. Li, E. Grant, L. Wossnig, I. Rungger, G. H. Booth, and J. Tennyson, The variational quantum eigensolver: A re- view of methods and best practices, Phys. Rep.986, 1 (2022)

2022
[40]

Larocca, S

M. Larocca, S. Thanasilp, S. Wang, K. Sharma, J. Bia- monte, P. J. Coles, L. Cincio, J. R. McClean, Z. Holmes, and M. Cerezo, Barren plateaus in variational quantum computing, Nat. Rev. Phys.7, 174 (2025)

2025
[41]

J. R. McClean, M. E. Kimchi-Schwartz, J. Carter, and W. A. de Jong, Hybrid quantum-classical hierarchy for mitigation of decoherence and determination of excited states, Phys. Rev. A95, 042308 (2017)

2017
[42]

H. R. Grimsley, S. E. Economou, E. Barnes, and N. J. Mayhall, An adaptive variational algorithm for exact molecular simulations on a quantum computer, Nat. Commun.10, 3007 (2019)

2019
[43]

K. M. Nakanishi, K. Mitarai, and K. Fujii, Subspace- search variational quantum eigensolver for excited states, Phys. Rev. Research1, 033062 (2019)

2019
[44]

R. M. Parrish, E. G. Hohenstein, P. L. McMahon, and T. J. Mart ´ ınez, Quantum computation of electronic tran- sitions using a variational quantum eigensolver, Phys. Rev. Lett.122, 230401 (2019)

2019
[45]

Higgott, D

O. Higgott, D. Wang, and S. Brierley, Variational quan- tum computation of excited states, Quantum3, 156 (2019)

2019
[46]

Jones, S

T. Jones, S. Endo, S. McArdle, X. Yuan, and S. C. Ben- jamin, Variational quantum algorithms for discovering Hamiltonian spectra, Phys. Rev. A99, 062304 (2019)

2019
[47]

Motta, C

M. Motta, C. Sun, A. T. K. Tan, M. J. O’Rourke, E. Ye, A. J. Minnich, F. G. S. L. Brand˜ ao, and G. K.-L. Chan, Determining eigenstates and thermal states on a quan- tum computer using quantum imaginary time evolution, Nat. Phys.16, 205 (2020)

2020
[48]

W. J. Huggins, J. Lee, U. Baek, B. O’Gorman, and K. B. Whaley, A non-orthogonal variational quantum eigen- solver, New J. Phys.22, 073009 (2020)

2020
[49]

N. H. Stair, R. Huang, and F. A. Evangelista, A Mul- tireference Quantum Krylov Algorithm for Strongly Cor- related Electrons, J. Chem. Theory Comput.16, 2236 (2020)

2020
[50]

Seki and S

K. Seki and S. Yunoki, Quantum power method by a superposition of time-evolved states, PRX Quantum2, 010333 (2021)

2021
[51]

Y. S. Yordanov, V. Armaos, C. H. W. Barnes, and D. R. M. Arvidsson-Shukur, Qubit-excitation-based adaptive variational quantum eigensolver, Commun. Phys.4, 1 (2021)

2021
[52]

H. L. Tang, V. O. Shkolnikov, G. S. Barron, H. R. Grim- sley, N. J. Mayhall, E. Barnes, and S. E. Economou, Qubit-adapt-vqe: An adaptive algorithm for construct- ing hardware-efficient ans¨ atze on a quantum processor, PRX Quantum2, 020310 (2021)

2021
[53]

Zhang, N

F. Zhang, N. Gomes, Y. Yao, P. P. Orth, and T. Iadecola, Adaptive variational quantum eigensolvers for highly ex- cited states, Phys. Rev. B104, 075159 (2021)

2021
[54]

Y. S. Yordanov, C. H. W. Barnes, and D. R. M. Arvidsson-Shukur, Molecular-excited-state calculations with the qubit-excitation-based adaptive variational quantum eigensolver protocol, Phys. Rev. A106, 032434 (2022)

2022
[55]

C. L. Cortes and S. K. Gray, Quantum Krylov subspace algorithms for ground- and excited-state energy estima- tion, Phys. Rev. A105, 022417 (2022)

2022
[56]

E. N. Epperly, L. Lin, and Y. Nakatsukasa, A theory of quantum subspace diagonalization, SIAM J. Matrix Anal. Appl.43, 1263 (2022). 15

2022
[57]

E. A. R. Guzman and D. Lacroix, Accessing ground-state and excited-state energies in a many-body system after symmetry restoration using quantum computers, Phys. Rev. C105, 024324 (2022)

2022
[58]

M. Q. Hlatshwayo, Y. Zhang, H. Wibowo, R. LaRose, D. Lacroix, and E. Litvinova, Simulating excited states of the Lipkin model on a quantum computer, Phys. Rev. C106, 024319 (2022)

2022
[59]

Wang and D

Y. Wang and D. A. Mazziotti, Electronic excited states from a variance-based contracted quantum eigensolver, Phys. Rev. A108, 022814 (2023)

2023
[60]

Y. Shen, K. Klymko, J. Sud, D. B. Williams-Young, W. A. de Jong, and N. M. Tubman, Real-Time Krylov Theory for Quantum Computing Algorithms, Quantum 7, 1066 (2023)

2023
[61]

Zheng, B

M. Zheng, B. Peng, N. Wiebe, A. Li, X. Yang, and K. Kowalski, Quantum algorithms for generator coordi- nate methods, Phys. Rev. Res.5, 023200 (2023)

2023
[62]

S. E. Smart, D. M. Welakuh, and P. Narang, Many-body excited states with a contracted quantum eigensolver, J. Chem. Theory Comput.20, 3580 (2024)

2024
[63]

Motta, W

M. Motta, W. Kirby, I. Liepuoniute, K. J. Sung, J. Cohn, A. Mezzacapo, K. Klymko, N. Nguyen, N. Yoshioka, and J. E. Rice, Subspace methods for electronic structure simulations on quantum computers, Electron. Struct.6, 013001 (2024)

2024
[64]

M. R. Hirsbrunner, J. W. Mullinax, Y. Shen, D. B. Williams-Young, K. Klymko, R. V. Beeumen, and N. M. Tubman, Diagnosing local minima and accelerat- ing convergence of variational quantum eigensolvers with quantum subspace techniques (2024), arXiv:2404.06534 [quant-ph]

2024 arXiv
[65]

Zheng, B

M. Zheng, B. Peng, A. Li, X. Yang, and K. Kowalski, Un- leashed from constrained optimization: Quantum com- puting for quantum chemistry employing generator coor- dinate inspired method, NPJ Quantum Inf.10, 1 (2024)

2024
[66]

Y. O. Nakagawa, M. Kamoshita, W. Mizukami, S. Sudo, and Y. ya Ohnishi, Adapt-qsci: Adaptive construction of an input state for quantum-selected configuration inter- action, J. Chem. Theory Comput.20, 10817 (2024)

2024
[67]

Ding, C.-L

L. Ding, C.-L. Hong, and C. Schilling, Ground and ex- cited states from ensemble variational principles, Quan- tum8, 1525 (2024)

2024
[68]

Beaujeault-Taudi` ere and D

Y. Beaujeault-Taudi` ere and D. Lacroix, Solving the Lip- kin model using quantum computers with two qubits only with a hybrid quantum-classical technique based on the generator coordinate method, Phys. Rev. C109, 024327 (2024)

2024
[69]

M. Q. Hlatshwayo, J. Novak, and E. Litvinova, Quan- tum benefit of the quantum equation of motion for the strongly coupled many-body problem, Phys. Rev. C109, 014306 (2024)

2024
[70]

Barison, J

S. Barison, J. R. Moreno, and M. Motta, Quantum- centric computation of molecular excited states with ex- tended sample-based quantum diagonalization, Quantum Sci. Technol.10, 025034 (2025)

2025
[71]

Marti-Dafcik, H

D. Marti-Dafcik, H. G. A. Burton, and D. P. Tew, Spin coupling is all you need: Encoding strong electron cor- relation in molecules on quantum computers, Phys. Rev. Res.7, 013191 (2025)

2025
[72]

H. R. Grimsley and F. A. Evangelista, Challenging ex- cited states from adaptive quantum eigensolvers: Sub- space expansions vs. state-averaged strategies, Quantum Sci. Technol.10, 025003 (2025)

2025
[73]

Zhang and D

J. Zhang and D. Lacroix, Excited States from ADAPT- VQE convergence path in Many-Body Problems: appli- cation to nuclear pairing problem andH 4 molecule dis- sociation (2025), arXiv:2506.22275 [quant-ph]

2025 arXiv
[74]

Fradking,Field Theories of Condensed Matter Physics (Cambridge University Press, 2013)

E. Fradking,Field Theories of Condensed Matter Physics (Cambridge University Press, 2013)

2013
[75]

Anselme Martin, T

B. Anselme Martin, T. Ayral, F. m. c. Jamet, M. J. Ranˇ ci´ c, and P. Simon, Combining matrix product states and noisy quantum computers for quantum simulation, Phys. Rev. A109, 062437 (2024)

2024
[76]

Dukelsky, S

J. Dukelsky, S. Pittel, and G. Sierra, Colloquium: Ex- actly solvable Richardson-Gaudin models for many-body quantum systems, Rev. Mod. Phys.76, 643 (2004)

2004
[77]

von Delft and D

J. von Delft and D. Ralph, Spectroscopy of discrete en- ergy levels in ultrasmall metallic grains, Phys. Rep.345, 61 (2001)

2001
[78]

Zelevinsky and A

V. Zelevinsky and A. Volya, Nuclear pairing: New per- spectives, Phys. of At. Nucl.66, 1781 (2003)

2003
[79]

D. M. Brink and R. A. Broglia,Nuclear Superfluidity Pairing in Finite Systems(Cambridge University Press, 2005)

2005
[80]

E. A. Ruiz Guzman and D. Lacroix, Accessing ground- state and excited-state energies in a many-body system after symmetry restoration using quantum computers, Phys. Rev. C105, 024324 (2022)

2022

Pith tools

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

[1] [1]

M. A. Nielsen and I. L. Chuang,Quantum Computing and Quantum Information(Cambridge University Press, Cambridge, 2000)

2000

[2] [2]

Schuld, I

M. Schuld, I. Sinayskiy, and F. Petruccione, An introduc- tion to quantum machine learning, Contemp. Phys.56, 172 (2015)

2015

[3] [3]

Torlai, G

G. Torlai, G. Mazzola, J. Carrasquilla, M. Troyer, R. Melko, and G. Carleo, Neural-network quantum state tomography, Nat. Phys.14, 447 (2018)

2018

[4] [4]

Biamonte, P

J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, Quantum machine learning, Na- ture549, 195 (2017)

2017

[5] [5]

G. Liu, M. Chen, Y.-X. Liu, D. Layden, and P. Cappel- laro, Repetitive readout enhanced by machine learning, Mach. Learn.: Sci. Technol.1, 015003 (2020)

2020

[6] [6]

Heese, T

R. Heese, T. Gerlach, S. M¨ ucke, S. M¨ uller, M. Jakobs, and N. Piatkowski, Explaining quantum circuits with shapley values: towards explainable quantum machine learning, Quantum Mach. Intell.7, 10.1007/s42484-025- 00254-8 (2025)

work page doi:10.1007/s42484-025- 2025

[7] [7]

T. Haug, C. N. Self, and M. S. Kim, Quantum ma- chine learning of large datasets using randomized mea- surements, Mach. Learn.: Sci. Technol.4, 015005 (2023)

2023

[8] [8]

Perrier, D

E. Perrier, D. Tao, and C. Ferrie, Quantum geometric machine learning for quantum circuits and control, New J. Phys.22, 103056 (2020)

2020