Pith. sign in

REVIEW 3 major objections 5 minor 58 references

Generative Learning for Quantum Measurement Design

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read This paper introduces FlowMeas, a generative model that learns shallow Clifford measurement ensembles which match or beat leading product-measurement baselines and reduce energy-estimation error by up to 27%.

desk verdict FlowMeas is a genuinely new angle on measurement scheduling, but the 27% claims rest on single training runs; send it to review with a request for code and error bars. read the letter →

arxiv 2608.11396 v1 pith:HHG4EZGW submitted 2026-08-11 quant-ph cs.LG

classification quant-phcs.LG
keywords quantummeasurementdesigngenerativeflownetworksCliffordcircuitsPauliobservablesenergyestimationshallowshadowsschedulingvariationalalgorithms
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 sets out to show that resource-constrained quantum measurement design can be framed as a generative learning problem, not as hand-built grouping or sequential derandomization. Its model, FlowMeas, trains a generative flow network to sample finite ensembles of shallow Clifford measurement circuits under a fixed shot budget and hardware constraints; each circuit in the ensemble is one measurement shot, so multiplicities are learned rather than assigned separately. The paper's central numerical claim is that at zero entangling depth FlowMeas learns qubit-wise commuting schedules that match or improve leading product-measurement methods on nearly all molecular benchmarks, and that allowing one or two CNOT layers reduces ground-state energy estimation error by up to 27% relative to the strongest state-independent product baseline. If the claim holds, the measurement layer of a quantum algorithm becomes a learned, hardware-aware object that interpolates between product measurements and deep commuting measurements without per-problem hand design.

What carries the argument

The mechanism that carries the argument is a generative flow network policy over Clifford tableaux. Starting from empty circuits, the policy samples one gate at a time from the local Clifford gates $H$, $S$, $HS$, $SH$, $HSH$, nearest-neighbor CNOTs, and a stop symbol, building $N$ circuits in parallel under masks that enforce the gate set, connectivity, and CNOT-depth limit. Coverage is evaluated exactly by stabilizer-tableau propagation, producing hit counts $h_k(U)$ for each Pauli term, and the ensemble reward is a monotone decreasing function of a state-independent proxy cost: variance-plus-bias, derandomized-shallow-shadow confidence, or overlapped-grouping diagonal variance. The ensemble trajectory-balance objective trains the shared policy toward low-proxy-cost ensembles, and because reward is assigned only after the complete ensemble is generated, the policy learns complementary rather than redundant coverage.

What would settle it

Take any FlowMeas-selected schedule, compute the full conditional RMSE including Pauli covariances from the reference state, and compare that RMSE with the proxy ranking across many independent training runs; if the lowest-proxy schedule is not among the lowest-RMSE schedules, or if the best-seed 27% gap over overlapped grouping becomes negligible when averaged over seeds, the central claim would be refuted.

Watch

Extended reading notes

Core claim

FlowMeas treats a measurement schedule as an ordered list $U=(U_1,\ldots,U_N)$ of deterministic Clifford circuits, one per shot, and learns a distribution over such ensembles with a generative flow network. A shared policy constructs the circuits gate by gate from local Clifford rotations, nearest-neighbor CNOTs, and a stop action, with action masks enforcing connectivity and a maximum CNOT depth; coverage of each Pauli term is computed exactly by stabilizer-tableau propagation. The terminal reward is a decreasing function of a state-independent proxy cost built from Hamiltonian coefficients and the ensemble hit counts, so training never needs the target quantum state. The paper reports that the learned product-measurement schedules match or improve published overlapped-grouping, locally-biased-shadow, derandomized-shadow, and largest-degree-first results on nearly all molecular benchmarks, and that one or two CNOT layers give further RMSE reductions of up to 27% relative to overlapped grouping; in a direct comparison using the derandomized-shallow-shadow confidence objective, FlowMeas matches or improves derandomized shallow shadows on four of five shared molecules. It further shows that a policy trained at one water geometry accelerates retraining across a potential-energy surface by factors of three to more than ten, and that the pipeline runs on 20-qubit molecules and a compactly encoded 54-qubit Hubbard model.

Load-bearing premise

The load-bearing assumption is that the training score, computed only from the Hamiltonian coefficients and how many circuits can read out each term, ranks measurement schedules in the same order as the true estimation error; the paper tests three such scores empirically but proves no bound that the score ordering equals the RMSE ordering, so if it misorders schedules the reported improvements would not transfer.

Editorial extensions

If this is right

  • Learned qubit-wise commuting schedules alone are competitive with or better than published product-measurement methods, so the benefit of the generative formulation does not depend on entangling gates.
  • Allowing one or two CNOT layers yields RMSE reductions of up to 27% over the strongest state-independent product baseline, showing the depth–sampling trade-off can be exploited automatically.
  • A policy trained at one molecular geometry can be reused across a potential-energy surface, cutting the iterations needed to converge by factors of about 3 to more than 10 without a systematic loss in final accuracy.
  • The same pipeline scales to 20-qubit molecular Hamiltonians and a compactly encoded 54-qubit Hubbard model, extending optimized, state-independent measurement design beyond previous molecular benchmarks.

Reading between the lines

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

  • The 27% improvement is a single-policy statement: the reported RMSE averages 500 shot-noise trials for a fixed schedule, not independent training runs, so the exact magnitude could shift under a multi-seed evaluation.
  • Because the proxy enters only through the terminal reward, the same policy could be fine-tuned with measured Pauli covariances once a few device runs are available, making the schedule partially state-dependent without changing the generator.
  • If the proxy-ranking assumption transfers, the approach should also apply to other weighted Pauli-estimation tasks—Hamiltonian learning, correlation functions, and logical-level Clifford scheduling under fault-tolerant routing constraints—since these share the same discrete, compositional, resource-constrained structure.
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

3 major / 5 minor

Summary. FlowMeas recasts resource-constrained quantum measurement design as a generative learning problem: a GFlowNet policy samples finite ensembles of shallow Clifford measurement circuits under a shot budget and hardware constraints, with rewards derived from state-independent proxy costs built on exact Pauli hit counts. The paper reports ground-state energy RMSE for eight Jordan-Wigner molecular Hamiltonians (4-20 qubits) at zero CNOT depth (QWC schedules) and at one or two CNOT layers, claiming that FlowMeas matches or improves leading product-measurement baselines (LDF, LBCS, Derand, OGM) on nearly all benchmarks and outperforms DSS on four of five shared systems. It also reports a potential-energy-surface transfer experiment for H2O with 3-10x faster retraining, and demonstrations on compactly encoded spinless Hubbard models with 24 and 54 qubits. The methods section derives three proxy objectives, gives an exact stabilizer-tableau coverage computation, and specifies the ensemble trajectory-balance training objective.

Significance. If the empirical claims hold, the paper makes a useful contribution: FlowMeas directly optimizes executable measurement ensembles with exact coverage statistics, requires no target-state information during training, interpolates between QWC and shallow-entangling measurement families, and extends the demonstrated scale beyond previous molecular benchmarks. Strengths include the clean derivations of the proxy objectives in Appendix A, the use of external published baselines rather than re-fitted parameters, and the explicitly state-agnostic training protocol. The central quantitative claims are not yet fully supported, however: the reported improvements rest on single training runs, and the link between the state-independent proxy used for schedule selection and the true finite-shot RMSE is validated only indirectly. These issues are load-bearing for the headline 27% reduction and the 'nearly all benchmarks' claim, but they are addressable with additional experiments rather than being fundamental flaws in the framework.

major comments (3)
  1. [Sec. 3.1, Table 1, Fig. 3, Appendix B.7] The reported results are based on a single GFlowNet training run per system and depth setting, with the final schedule selected by the lowest proxy cost encountered during that run. Appendix B.7 explicitly states that the 500 trials 'quantify shot noise for a fixed schedule; they do not quantify variability across independent training runs.' Consequently, the claimed improvements (27% for H2(8), 17% for BeH2(14), 19% for H2O(14), 9% for NH3(16), and the DSS MAE reductions in Fig. 3) cannot yet be distinguished from training-run variability. This is particularly concerning for the near-tie entries such as LiH(12) at 0.036 vs. 0.036 and H2(4) at 0.013 vs. OGM 0.011, where a single run is insufficient to establish a match or a small difference. Please provide multiple independent training runs (or seeds) with means and standard deviations, and state the number of runs used for every numerical claim.
  2. [Sec. 5.2, Eq. (11); Sec. 3.2; Eq. (9)] The proxy-to-risk link is the load-bearing assumption and it is not yet established. The terminal reward is built from CDSS, which depends only on |c_k| and hit counts, while the true conditional risk R(U;ρ) in Eq. (9) includes state-dependent Pauli covariances. CDSS is not derived from any bound on R(U;ρ), so minimizing it need not minimize the finite-shot energy error. The only validation in Sec. 3.2 and Fig. 4 compares three proxies using one training run per proxy and one selected schedule per system, and no proxy-versus-RMSE validation is reported at all for the 54-qubit Hubbard experiment. I would like to see either (i) a multi-seed correlation analysis between proxy values and RMSE over held-out schedules, or (ii) a formal statement of the conditions under which CDSS ranking matches RMSE ranking, or (iii) at minimum an explicit sensitivity analysis showing that the headline margins are stable when the schedule is selected by a different but equally plausible proxy.
  3. [Sec. 3.4, Fig. 6, Appendix B.6.4] The Hubbard scaling demonstration is currently qualitative. FlowMeas is compared with an oracle Neyman allocation that uses target-state variances, which is appropriate as a demanding baseline, but only one FlowMeas schedule per lattice size is shown and the RMSE values are reported in arbitrary units without normalization details. The statement that FlowMeas 'remains within the same error scale' is not quantified as a ratio or interval. Please report the FlowMeas-to-oracle RMSE ratio for both lattice sizes, include repeated training runs or a sensitivity range, and specify the normalization of the vertical axis in Fig. 6.
minor comments (5)
  1. [Sec. 5.3] There is a typo in the sentence defining f_theta: 'the neural policy network where , which maps' should read 'the neural policy network, which maps'.
  2. [Sec. B.6.1] The sentence 'This gives the number of Pauli terms 88, and 216 for 4×4 and 6×6 respectively' is grammatically awkward; please rephrase, for example as 'This gives 88 and 216 Pauli terms for the 4×4 and 6×6 lattices, respectively.'
  3. [Fig. 4] The percentage annotations in panel (b) are difficult to parse because they do not clearly indicate which proxy is the reference and which proxy each annotation refers to. Please add explicit labels, such as 'relative to VB' on the axis and markers for each proxy in the legend.
  4. [Fig. 6] The vertical axis is labeled 'RMSE (arbitrary units)'; please state the normalization convention used, since arbitrary units make the quantitative comparison between the two lattice sizes hard to interpret.
  5. [Sec. 5.2 and Appendix A.2] The default value epsilon = 0.9 for the DSS proxy is stated only in Appendix A.2; please state it in the main text near Eq. (11) so that the main results are self-contained.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found; the final schedules are selected by state-independent proxies and the RMSE is evaluated retrospectively against external benchmarks.

full rationale

FlowMeas's derivation chain is self-contained against external benchmarks. The terminal reward is a state-independent proxy cost (Eqs. 10-12) computed only from Hamiltonian coefficients and exact Pauli hit counts; the final schedule is retained under the same proxy (Sec. 5.5), and the target-state RMSE is used only for retrospective evaluation, so no fitted quantity is renamed as a prediction. The DSS cost in Eq. (11) is imported from Ref. [32] rather than derived from FlowMeas's own outputs, and the DSS baseline values in Fig. 3 are published values that are not recomputed by this paper. The only self-citation, Ref. [44] (FlowQ-Net), is cited as related work on GFlowNet circuit synthesis and is not used to justify measurement-design performance. The choice of the DSS proxy after the three-proxy comparison in Fig. 4 is a model-selection limitation on a fixed benchmark set, and Appendix B.7 itself notes that the 500 trials quantify shot noise rather than variability across independent training runs; these are statistical robustness concerns, not circular reductions. No equation or construction makes the claimed RMSE improvements equivalent to the paper's training inputs.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. The free parameters are hand-chosen hyperparameters of the proxy and reward (epsilon, alpha, beta) plus a stabilizer penalty for the Hubbard reference calculation; none are fitted to the target-state data used in the reported RMSE. The axioms are the state-independent proxy assumption, the approximate product-form GFlowNet objective, the completeness of the action set, and the accuracy of the DMRG reference states.

free parameters (4)
  • DSS proxy epsilon = 0.9
    Chosen by hand in Eq. (11) and Appendix A.2; controls how strongly small hit counts are penalized. Results may depend on it, and it is not fitted to target-state data.
  • Reward scaling alpha = 10
    Appendix A.4 defines log R = alpha - beta * Ctilde; chosen for numerical stability. Affects GFlowNet training but not the proxy ordering.
  • Reward scaling beta = 100
    Same as alpha; chosen by hand to shape the terminal reward in Eq. (42).
  • Hubbard stabilizer penalty lambda = 6
    Appendix B.6.1 Eq. (54); picks the desired stabilizer sector for the compact encoding. Not part of the measurement proxy but part of the reference-state construction.
assumptions (5)
  • domain assumption The variance-plus-bias proxy uses Haar-averaged Pauli covariances and assumes distinct Pauli estimator covariances vanish on average.
    Appendix A.1 Eq. (29); this makes the proxy state-independent, but the actual target state may have nonzero covariances that shift the ranking.
  • domain assumption State-independent proxies (VB, DSS, OGM) predict finite-shot energy RMSE.
    Sec. 3.2 tests three proxies empirically, but schedule selection in all benchmarks relies on this alignment.
  • domain assumption The product-form GFlowNet policy, trained with trajectory balance, finds low-cost ensembles even though no product distribution can match the ensemble reward exactly.
    Sec. 5.4 states explicitly that R(U) does not factorize; the method relies on the approximate objective being good enough.
  • domain assumption The action set and masks generate all relevant shallow Clifford circuits up to depth dmax.
    Sec. 5.3 gives the normal form Eq. (14) and bounds the state space, but does not prove the generated family contains the optimal shallow circuit for every Hamiltonian.
  • domain assumption DMRG reference states with bond dimensions 256 (4x4) and 756 (6x6) give accurate moments for the Hubbard evaluation.
    Appendix B.6.2; exact ground states for 36 JW qubits are unavailable, so the reported RMSE inherits DMRG approximation error.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generative Learning for Quantum Measurement Design." pith.science (2026). https://pith.science/paper/HHG4EZGW

@misc{pith2026260811396,
  author       = {Pith},
  title        = {Pith review of: Generative Learning for Quantum Measurement Design},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HHG4EZGW}},
  note         = {Machine review of arXiv:2608.11396}
}
abstract

Extracting quantum information from a quantum state is a fundamental task of quantum computation, often requiring the estimation of many non-commuting observables under a finite measurement budget. For both near-term and early fault-tolerant settings, the measurement protocol must balance statistical efficiency against implementation resources such as circuit depth, connectivity, and entangling-gate count. Many existing strategies focus on two extremes: hardware-friendly product measurements with high sampling cost, and fully commuting measurements with deep circuits. Here we recast resource-constrained measurement design as a generative learning problem. We introduce FlowMeas, which uses a generative flow network to directly sample finite ensembles of shallow Clifford measurement circuits subject to a prescribed shot budget and hardware constraints. At zero entangling depth, FlowMeas learns qubit-wise commuting measurement schedules and already matches or improves leading product-measurement methods on nearly all molecular benchmarks. Allowing one or two entangling gate layers yields further reductions in energy estimation error of up to $27\%$ relative to the strongest state-independent product-measurement baseline. The learned policy can also be reused across related Hamiltonians, substantially accelerating retraining along a molecular potential-energy surface. We further obtain results for molecular Hamiltonians with up to 20 qubits and apply the framework to a compactly encoded 54-qubit interacting fermionic model, extending the demonstrated scale beyond prior molecular benchmarks. These results establish generative learning as a flexible and unified framework for quantum measurement design under practical resource constraints.

Figures

Figures reproduced from arXiv: 2608.11396 by the authors.

Figure 1
Figure 1. Resource-constrained quantum measurement design. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Generative learning of shallow measurement ensembles. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Direct comparison with derandomized shallow shadows. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Effect of the state-independent training proxy. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Reuse of the learned policy across an H2O potential energy surface. Number of gradient updates required to satisfy the convergence criterion (a) and final ground-state energy RMSE (b) as functions of the symmetric O–H bond length. At each bond length, the markers show …
Figure 6
Figure 6. Figure 6: Scaling to compactly encoded spinless Hubbard models. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Single-qubit Clifford representatives. Conjugation by H, S, HSH, SH, and HS generates the five nonidentity permutations of the Pauli axes, up to signs. Together with the identity, these actions represent the six elements of the single-qubit Clifford group modulo Pauli …
Figure 8
Figure 8. Figure 8: Training throughput for the molecular benchmarks. [PITH_FULL_IMAGE:figures/full_fig_p034_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 22 canonical work pages

  1. [1]

    A variational eigenvalue solver on a photonic quantum processor

    Alberto Peruzzo et al. “A variational eigenvalue solver on a photonic quantum processor”. In: Nature Communications5.1 (2014), p. 4213.doi:10.1038/ncomms5213

  2. [2]

    Quantum computing in the NISQ era and beyond

    John Preskill. “Quantum computing in the NISQ era and beyond”. In:Quantum2 (2018), p. 79.doi:10.22331/q-2018-08-06-79

  3. [3]

    The variational quantum eigensolver: A review of methods and best practices

    Jules Tilly et al. “The variational quantum eigensolver: A review of methods and best practices”. In:Physics Reports986 (2022), pp. 1–128.doi:10.1016/j.physrep.2022.08.003

  4. [4]

    Quantum Measurement for Quantum Chemistry on a Quantum Computer

    Smik Patel et al. “Quantum Measurement for Quantum Chemistry on a Quantum Computer”. In:Chemical Reviews125.16 (2025), pp. 7490–7524.doi:10.1021/acs.chemrev.5c00055

  5. [5]

    Learning many-body Hamiltonians with Heisenberg-limited scaling

    Hsin-Yuan Huang et al. “Learning many-body Hamiltonians with Heisenberg-limited scaling”. In:Physical Review Letters130.20 (2023), p. 200403.doi: 10.1103/PhysRevLett.130.200403

  6. [6]

    Predicting many properties of a quantum system from very few measurements

    Hsin-Yuan Huang, Richard Kueng, and John Preskill. “Predicting many properties of a quantum system from very few measurements”. In:Nature Physics16.10 (2020), pp. 1050– 1057.doi:10.1038/s41567-020-0932-7

  7. [7]

    Theoretical and experimental perspectives of quantum verification

    Jose Carrasco et al. “Theoretical and experimental perspectives of quantum verification”. In: PRX Quantum2.1 (2021), p. 010102.doi:10.1103/PRXQuantum.2.010102

  8. [8]

    Progress towards practical quantum variational algorithms

    Dave Wecker, Matthew B. Hastings, and Matthias Troyer. “Progress towards practical quantum variational algorithms”. In:Physical Review A92.4 (2015), p. 042303.doi: 10.1103/PhysRevA. 92.042303

Show all 58 references
  1. [9]

    Measurements as a roadblock to near-term practical quantum advantage in chemistry: Resource analysis

    J´ erˆ ome F. Gonthier et al. “Measurements as a roadblock to near-term practical quantum advantage in chemistry: Resource analysis”. In:Physical Review Research4.3 (2022), p. 033154. doi:10.1103/PhysRevResearch.4.033154

  2. [10]

    Nearly optimal measurement scheduling for partial tomography of quantum states

    Xavier Bonet-Monroig, Ryan Babbush, and Thomas E. O’Brien. “Nearly optimal measurement scheduling for partial tomography of quantum states”. In:Physical Review X10.3 (2020), p. 031064.doi:10.1103/PhysRevX.10.031064

  3. [11]

    Measurements of quantum Hamiltonians with locally-biased classical shadows

    Charles Hadfield et al. “Measurements of quantum Hamiltonians with locally-biased classical shadows”. In:Communications in Mathematical Physics391.3 (2022), pp. 951–967.doi: 10.1007/s00220-022-04343-8

  4. [12]

    Efficient estimation of Pauli observables by derandomization

    Hsin-Yuan Huang, Richard Kueng, and John Preskill. “Efficient estimation of Pauli observables by derandomization”. In:Physical Review Letters127.3 (2021), p. 030503.doi: 10.1103/ PhysRevLett.127.030503

  5. [13]

    Overlapped grouping measurement: A unified framework for measuring quantum states

    Bujiao Wu et al. “Overlapped grouping measurement: A unified framework for measuring quantum states”. In:Quantum7 (2023), p. 896.doi:10.22331/q-2023-01-13-896

  6. [14]

    Measurement optimization in the variational quantum eigensolver using a minimum clique cover

    Vladyslav Verteletskyi, Tzu-Ching Yen, and Artur F. Izmaylov. “Measurement optimization in the variational quantum eigensolver using a minimum clique cover”. In:The Journal of Chemical Physics152.12 (2020), p. 124114.doi:10.1063/1.5141458

  7. [15]

    Unitary partitioning approach to the measurement problem in the variational quantum eigensolver method

    Artur F. Izmaylov et al. “Unitary partitioning approach to the measurement problem in the variational quantum eigensolver method”. In:Journal of Chemical Theory and Computation 16.1 (2020), pp. 190–195.doi:10.1021/acs.jctc.9b00791

  8. [16]

    Measuring all compatible operators in one series of single-qubit measurements using unitary transformations

    Tzu-Ching Yen, Vladyslav Verteletskyi, and Artur F. Izmaylov. “Measuring all compatible operators in one series of single-qubit measurements using unitary transformations”. In: Journal of Chemical Theory and Computation16.4 (2020), pp. 2400–2409.doi: 10.1021/acs. jctc.0c00008

  9. [17]

    Measurement reduction in variational quantum algorithms

    Andrew Zhao et al. “Measurement reduction in variational quantum algorithms”. In:Physical Review A101.6 (2020), p. 062322.doi:10.1103/PhysRevA.101.062322. 20

  10. [18]

    Efficient and noise resilient measurements for quantum chemistry on near-term quantum computers

    William J. Huggins et al. “Efficient and noise resilient measurements for quantum chemistry on near-term quantum computers”. In:npj Quantum Information7.1 (2021), p. 23.doi: 10.1038/s41534-020-00341-7

  11. [19]

    Cartan subalgebra approach to efficient measurements of quantum observables

    Tzu-Ching Yen and Artur F. Izmaylov. “Cartan subalgebra approach to efficient measurements of quantum observables”. In:PRX Quantum2.4 (2021), p. 040320.doi: 10.1103/PRXQuantum. 2.040320

  12. [20]

    Efficient quantum measurement of Pauli operators in the presence of finite sampling error

    Ophelia Crawford et al. “Efficient quantum measurement of Pauli operators in the presence of finite sampling error”. In:Quantum5 (2021), p. 385.doi:10.22331/q-2021-01-20-385

  13. [21]

    1998.doi: 10

    Daniel Gottesman.The Heisenberg representation of quantum computers. 1998.doi: 10 . 48550/arXiv.quant-ph/9807006. arXiv:quant-ph/9807006 [quant-ph]

  14. [22]

    Improved simulation of stabilizer circuits

    Scott Aaronson and Daniel Gottesman. “Improved simulation of stabilizer circuits”. In: Physical Review A70.5 (2004), p. 052328.doi:10.1103/PhysRevA.70.052328

  15. [23]

    Hadamard-free circuits expose the structure of the Clifford group

    Sergey Bravyi and Dmitri Maslov. “Hadamard-free circuits expose the structure of the Clifford group”. In:IEEE Transactions on Information Theory67.7 (2021), pp. 4546–4563.doi: 10.1109/TIT.2021.3081415

  16. [24]

    The randomized measurement toolbox

    Andreas Elben et al. “The randomized measurement toolbox”. In:Nature Reviews Physics5.1 (2023), pp. 9–24.doi:10.1038/s42254-022-00535-2

  17. [25]

    Learning to measure: Adaptive informationally complete generalized measurements for quantum algorithms

    Guillermo Garc´ ıa-P´ erez et al. “Learning to measure: Adaptive informationally complete generalized measurements for quantum algorithms”. In:PRX Quantum2.4 (2021), p. 040342. doi:10.1103/PRXQuantum.2.040342

  18. [26]

    Fermionic partial tomography via classical shadows

    Andrew Zhao, Nicholas C. Rubin, and Akimasa Miyake. “Fermionic partial tomography via classical shadows”. In:Physical Review Letters127.11 (2021), p. 110504.doi: 10 . 1103 / PhysRevLett.127.110504

  19. [27]

    Shallow shadows: Expectation estimation using low-depth random Clifford circuits

    Christian Bertoni et al. “Shallow shadows: Expectation estimation using low-depth random Clifford circuits”. In:Physical Review Letters133.2 (2024). arXiv:2209.12924 [quant-ph], p. 020602.doi:10.1103/PhysRevLett.133.020602

  20. [28]

    Operator relaxation and the optimal depth of classical shadows

    Matteo Ippoliti et al. “Operator relaxation and the optimal depth of classical shadows”. In: Physical Review Letters130.23 (2023), p. 230403.doi:10.1103/PhysRevLett.130.230403

  21. [29]

    Classical shadow tomography with locally scrambled quantum dynamics

    Hong-Ye Hu, Soonwon Choi, and Yi-Zhuang You. “Classical shadow tomography with locally scrambled quantum dynamics”. In:Physical Review Research5.2 (2023), p. 023027.doi: 10.1103/PhysRevResearch.5.023027

  22. [30]

    Scalable and flexible classical shadow tomography with tensor networks

    Ahmed A. Akhtar, Hong-Ye Hu, and Yi-Zhuang You. “Scalable and flexible classical shadow tomography with tensor networks”. In:Quantum7 (2023), p. 1026.doi: 10.22331/q-2023- 06-01-1026

  23. [31]

    Demonstration of robust and efficient quantum property learning with shallow shadows

    Hong-Ye Hu et al. “Demonstration of robust and efficient quantum property learning with shallow shadows”. In:Nature Communications16.1 (2025), p. 2943.doi: 10.1038/s41467- 025-57349-w

  24. [32]

    2024.doi:10.48550/arXiv.2412.18973

    Katherine Van Kirk et al.Derandomized shallow shadows: Efficient Pauli learning with bounded- depth circuits. 2024.doi:10.48550/arXiv.2412.18973. arXiv:2412.18973 [quant-ph]

  25. [33]

    Flow network based generative models for non-iterative diverse candidate generation

    Emmanuel Bengio et al. “Flow network based generative models for non-iterative diverse candidate generation”. In:Advances in Neural Information Processing Systems34 (2021), pp. 27381–27394

  26. [34]

    Trajectory balance: Improved credit assignment in GFlowNets

    Nikolay Malkin et al. “Trajectory balance: Improved credit assignment in GFlowNets”. In: Advances in Neural Information Processing Systems35 (2022), pp. 5955–5967

  27. [35]

    GFlowNet foundations

    Yoshua Bengio et al. “GFlowNet foundations”. In:Journal of Machine Learning Research 24.210 (2023), pp. 1–55

  28. [36]

    Sutton and Andrew G

    Richard S. Sutton and Andrew G. Barto.Reinforcement Learning: An Introduction. 2nd ed. Cambridge, MA: MIT Press, 2018. 21

  29. [37]

    Biological sequence design with GFlowNets

    Moksh Jain et al. “Biological sequence design with GFlowNets”. In:Proceedings of the 39th International Conference on Machine Learning. Vol. 162. Proceedings of Machine Learning Research. 2022, pp. 9786–9801

  30. [38]

    Multi-objective GFlowNets

    Moksh Jain et al. “Multi-objective GFlowNets”. In:Proceedings of the 40th International Conference on Machine Learning. Vol. 202. Proceedings of Machine Learning Research. 2023, pp. 14631–14653

  31. [39]

    Let the flows tell: Solving graph combinatorial problems with GFlowNets

    Dinghuai Zhang et al. “Let the flows tell: Solving graph combinatorial problems with GFlowNets”. In:Advances in Neural Information Processing Systems36 (2023), pp. 11952– 11969

  32. [40]

    Robust scheduling with GFlowNets

    David W. Zhang et al. “Robust scheduling with GFlowNets”. In:11th International Conference on Learning Representations (ICLR). 2023. arXiv:2302.05446 [cs.AI]

  33. [41]

    Bayesian structure learning with generative flow networks

    Tristan Deleu et al. “Bayesian structure learning with generative flow networks”. In:Proceedings of the 38th Conference on Uncertainty in Artificial Intelligence (UAI). Vol. 180. Proceedings of Machine Learning Research. 2022, pp. 518–528

  34. [42]

    Generative flow networks for discrete probabilistic modeling

    Dinghuai Zhang et al. “Generative flow networks for discrete probabilistic modeling”. In: Proceedings of the 39th International Conference on Machine Learning. Vol. 162. Proceedings of Machine Learning Research. 2022, pp. 26412–26428

  35. [43]

    Huidobro-Meezs et al.GFlowNets for Hamiltonian decomposition in groups of compatible operators

    Isaac L. Huidobro-Meezs et al.GFlowNets for Hamiltonian decomposition in groups of compatible operators. Machine Learning and the Physical Sciences Workshop, NeurIPS 2024. 2024.doi:10.48550/arXiv.2410.16041. arXiv:2410.16041 [quant-ph]

  36. [44]

    2025.doi: 10

    Jun Dai, Michael Rizvi-Martel, and Guillaume Rabusseau.FlowQ-Net: A generative framework for automated quantum circuit design. 2025.doi: 10 . 48550 / arXiv . 2510 . 26688. arXiv: 2510.26688 [quant-ph]

  37. [45]

    ¨Uber das Paulische ¨Aquivalenzverbot

    Pascual Jordan and Eugene Wigner. “ ¨Uber das Paulische ¨Aquivalenzverbot”. In:Zeitschrift f¨ ur Physik47.9-10 (1928), pp. 631–651.doi:10.1007/BF01331938

  38. [46]

    https://github

    Charles Hadfield.Variances: Measurement in noisy quantum computers. https://github. com/charleshadfield/variances. Accessed 2026-08-01

  39. [47]

    Compact fermion to qubit mappings

    Charles Derby et al. “Compact fermion to qubit mappings”. In:Physical Review B104.3 (2021), p. 035118.doi:10.1103/PhysRevB.104.035118

  40. [48]

    Scalable simulation of fermionic encoding performance on noisy quantum computers

    Emiliia Dyrenkova, Raymond Laflamme, and Michael Vasmer. “Scalable simulation of fermionic encoding performance on noisy quantum computers”. In:Physical Review A113.2 (2026). arXiv:2506.06425 [quant-ph], p. 022607.doi:10.1103/dg5d-m728

  41. [49]

    Density matrix formulation for quantum renormalization groups

    Steven R. White. “Density matrix formulation for quantum renormalization groups”. In: Physical Review Letters69.19 (1992), pp. 2863–2866.doi:10.1103/PhysRevLett.69.2863

  42. [50]

    The density-matrix renormalization group in the age of matrix product states

    Ulrich Schollw¨ ock. “The density-matrix renormalization group in the age of matrix product states”. In:Annals of Physics326.1 (2011), pp. 96–192.doi:10.1016/j.aop.2010.09.012

  43. [51]

    On the two different aspects of the representative method: The method of stratified sampling and the method of purposive selection

    Jerzy Neyman. “On the two different aspects of the representative method: The method of stratified sampling and the method of purposive selection”. In:Journal of the Royal Statistical Society97.4 (1934), pp. 558–625.doi:10.2307/2342192

  44. [52]

    Improving quantum measurements by introducing “ghost

    Seonghoon Choi, Tzu-Ching Yen, and Artur F. Izmaylov. “Improving quantum measurements by introducing “ghost” Pauli products”. In:Journal of Chemical Theory and Computation 18.12 (2022), pp. 7394–7402.doi:10.1021/acs.jctc.2c00837

  45. [53]

    Deterministic improvements of quantum measurements with grouping of compatible operators, non-local transformations, and covariance estimates

    Tzu-Ching Yen, Aadithya Ganeshram, and Artur F. Izmaylov. “Deterministic improvements of quantum measurements with grouping of compatible operators, non-local transformations, and covariance estimates”. In:npj Quantum Information9.1 (2023), p. 14.doi: 10.1038/s41534- 023-00683-y

  46. [54]

    Yadong Wu et al.Designing Shadow Tomography Protocols by Natural Language Processing

  47. [55]

    Learning GFlowNets from partial episodes for improved convergence and stability

    Kanika Madan et al. “Learning GFlowNets from partial episodes for improved convergence and stability”. In:Proceedings of the 40th International Conference on Machine Learning. Vol. 202. Proceedings of Machine Learning Research. 2023, pp. 23467–23483

  48. [56]

    PyTorch: An imperative style, high-performance deep learning library

    Adam Paszke et al. “PyTorch: An imperative style, high-performance deep learning library”. In:Advances in Neural Information Processing Systems32 (2019), pp. 8024–8035

  49. [57]

    CuPy: A NumPy-compatible library for NVIDIA GPU calculations

    Ryosuke Okuta et al. “CuPy: A NumPy-compatible library for NVIDIA GPU calculations”. In:Proceedings of the Workshop on Machine Learning Systems (LearningSys) at the 31st Con- ference on Neural Information Processing Systems (NIPS). 2017.url: http://learningsys. org/nips17/asse...

  50. [2025]

    arXiv:2509.12782 [quant-ph].url:https://arxiv.org/abs/2509.12782. 22

Pith tools

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