Pith. sign in

REVIEW 1 major objections 1 minor 3 cited by

Transformer refined quantum sampling for strongly correlated electronic structure

T0 review · 1 major / 1 minor · reviewed 2026-06-30 · grok-4.3

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

Pith's one-line read A transformer reconstructs full electronic wavefunctions from sparse quantum samples on NISQ hardware.

desk verdict The abstract outlines a hybrid NISQ method that samples key configs with a custom USCI ansatz on Zuchongzhi then uses a transformer to reconstruct wavefunctions, claiming chemical accuracy on a 40-qubit ferredoxin model and 12 mHartree agreement with DMRG on the nitrogenase P-cluster. read the letter →

arxiv 2605.24617 v1 pith:53X4HPVC submitted 2026-05-23 quant-ph physics.chem-ph

Xiongzhi Zeng , Ming Gong , Bowen Kan , Yi Fan , Huan Ma , Jianbin Cai , Yancheng Liu , Naibin Zhou
show 13 more authors
Tao Jiang Shaojun Guo Zhijie Fan Zongkang Zhang Yuan Li Sirui Cao Kai Yan Xiaobo Zhu Yi Luo Honghui Shang Zhenyu Li Jian-Wei Pan Jinlong Yang
This is my paper · ORCID
classification quant-phphysics.chem-ph
keywords quantumsamplingtransformerneuralnetworkelectronicstructurestronglycorrelatedelectronsNISQdevicesconfigurationinteractioniron-sulfurclusterswavefunctionreconstruction
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 introduces QiankunNet-QSCI, a hybrid quantum-classical framework that pairs quantum sampling on a superconducting processor with a transformer neural network to handle strongly correlated electrons. An efficient unitary selected configuration interaction ansatz runs on the Zuchongzhi 3.1 processor to locate the most chemically relevant configurations. The transformer then learns from this limited data to reconstruct the complete wavefunction. The approach produces chemical accuracy for the 40-qubit ferredoxin active center and 12 milli-Hartree agreement with DMRG for the nitrogenase P-cluster in a 114-electron space.

What carries the argument

The USCI ansatz that performs quantum sampling of key configurations combined with the QiankunNet transformer that reconstructs the complete wavefunction from those samples.

What would settle it

Apply the complete pipeline to a smaller strongly correlated system whose exact wavefunction and energy are known classically and check whether the reconstructed energy deviates beyond chemical accuracy.

Watch

Extended reading notes

Core claim

The authors establish that configurations identified by the USCI ansatz on the quantum processor contain sufficient information for the QiankunNet transformer to infer the full electronic wavefunction with high fidelity, enabling accurate simulations of the [2Fe-2S] ferredoxin active center to chemical accuracy and the nitrogenase P-cluster to 12 milli-Hartree agreement with the best DMRG result.

Load-bearing premise

The sparse configurations identified by the USCI ansatz on the NISQ processor contain enough information for the transformer to reconstruct the full wavefunction across the Hilbert space.

Editorial extensions

If this is right

  • The [2Fe-2S] ferredoxin active center simulation reaches chemical accuracy on 40 qubits.
  • The nitrogenase P-cluster simulation in a 114-electron 73-orbital space reaches 12 milli-Hartree agreement with DMRG.
  • The framework supplies a practical route to accurate quantum-assisted electronic structure calculations on current NISQ devices.

Reading between the lines

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

  • The same sampling-plus-reconstruction pattern may apply to other strongly correlated molecular or material systems where full classical treatment is intractable.
  • Training the transformer on data collected from repeated quantum runs could improve tolerance to device noise.
  • The method could reduce the quantum circuit depth or shot count required for larger active spaces compared with purely quantum algorithms.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

1 major / 1 minor

Summary. The manuscript introduces QiankunNet-QSCI, a hybrid quantum-classical framework for simulating strongly correlated electronic systems on NISQ hardware. It proposes an efficient unitary selected configuration interaction (USCI) ansatz implemented on the Zuchongzhi 3.1 processor to identify sparse but chemically relevant electronic configurations, which are then fed to a transformer model (QiankunNet) that infers and reconstructs the full wavefunction. Numerical results are reported for the 40-qubit [2Fe-2S] ferredoxin active center (chemical accuracy) and the nitrogenase P-cluster (114 electrons, 73 orbitals, 12 mHa agreement with DMRG).

Significance. If the central claims hold after detailed verification, the work would demonstrate a viable path to accurate electronic-structure calculations on current quantum processors by leveraging quantum sampling for configuration selection and neural-network reconstruction to compensate for sparsity. This combination could extend the reach of NISQ devices to bioinorganic active spaces that remain challenging for both classical methods and existing quantum algorithms.

major comments (1)
  1. [§3] The central claim that sparse USCI samples suffice for high-fidelity reconstruction of the full wavefunction across the Hilbert space (abstract and §3) is load-bearing; the manuscript must provide explicit validation (e.g., overlap or energy error versus full CI on smaller benchmark systems) to rule out that the reported accuracies arise from post-hoc selection or incomplete coverage of important configurations.
minor comments (1)
  1. Notation for the USCI ansatz and the precise definition of the transformer input/output (e.g., how configurations are encoded) should be clarified with explicit equations to aid reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive review and for highlighting the importance of validating the central claim regarding sparse USCI sampling and transformer reconstruction. We address the major comment below and will revise the manuscript to incorporate the requested validation.

read point-by-point responses

  1. Referee: [§3] The central claim that sparse USCI samples suffice for high-fidelity reconstruction of the full wavefunction across the Hilbert space (abstract and §3) is load-bearing; the manuscript must provide explicit validation (e.g., overlap or energy error versus full CI on smaller benchmark systems) to rule out that the reported accuracies arise from post-hoc selection or incomplete coverage of important configurations.

    Authors: We agree that explicit validation against full CI on smaller systems is necessary to substantiate the load-bearing claim. In the revised manuscript we will add a dedicated subsection (or appendix) presenting results on smaller benchmark molecules (e.g., H2O and N2 in minimal basis sets) where exact FCI is tractable. For these systems we will report both the wavefunction overlap with exact FCI and the energy error obtained using the identical USCI sampling protocol followed by QiankunNet reconstruction. This addition will directly address concerns about post-hoc selection or incomplete coverage of important configurations. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity detected

full rationale

The described framework combines quantum sampling via a USCI ansatz on the Zuchongzhi processor with subsequent transformer-based inference to reconstruct the wavefunction. No equations or steps in the abstract reduce by construction to fitted inputs, self-definitions, or load-bearing self-citations; results are presented as empirical outcomes benchmarked externally against DMRG. The derivation chain relies on hardware execution and network training rather than renaming or smuggling prior results into the target claim, making the method self-contained against the provided description.

Assumptions & free parameters 0 free parameters · 1 assumptions · 2 invented entities

Abstract-only; no explicit free parameters, ad-hoc axioms, or invented entities beyond the named method components are identifiable. Relies on standard quantum mechanics and electronic structure assumptions.

assumptions (1)
  • standard math Standard principles of quantum mechanics and configuration interaction theory apply to the electronic structure problem.
    The framework assumes the validity of the electronic Hamiltonian and configuration interaction expansion for the systems studied.
invented entities (2)
  • QiankunNet
    purpose: Transformer neural network to learn and reconstruct the full wavefunction from sparse quantum samples.
    Introduced as the machine learning component of the hybrid framework.
  • USCI ansatz
    purpose: Unitary selected configuration interaction ansatz designed for efficient quantum sampling of significant configurations.
    Proposed specifically for use on the quantum processor.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transformer refined quantum sampling for strongly correlated electronic structure." pith.science (2026). https://pith.science/paper/53X4HPVC

@misc{pith2026260524617,
  author       = {Pith},
  title        = {Pith review of: Transformer refined quantum sampling for strongly correlated electronic structure},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/53X4HPVC}},
  note         = {Machine review of arXiv:2605.24617}
}
read the original abstract

Although quantum computing offers a promising solution for strongly correlated system simulation, existing algorithms face significant bottlenecks on current noisy intermediate-scale quantum (NISQ) devices. Here, we introduce QiankunNet-QSCI, a hybrid quantum-classical framework that addresses this challenge by combining efficient quantum-sampling with a transformer neural network. An efficient unitary selected configuration Interaction (USCI) ansatz especially designed for quantum sampling is proposed to identify the most chemically significant electronic configurations on the Zuchongzhi 3.1 quantum processor. Subsequently, the transformer model QiankunNet learns from these sparse yet critical quantum data to infer and reconstruct the complete electronic wavefunction with high fidelity. Simulation of the challenging 40-qubit [2Fe-2S] ferredoxin active center achieves chemical accuracy. Simulation of the nitrogenase P-cluster in a 114-electron 73-orbital active space also reaches 12 milli-Hartree-level agreement with the best density matrix renormalization group (DMRG) result. QiankunNet-QSCI thus offers a practical route to accurate quantum-assisted electronic structure calculations on current devices.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

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

  1. Embedded quantum computing for many-body surface reaction

    quant-ph 2026-07 conditional novelty 6.5 of 10

    Embedded quantum computing (QC-DFET) recovers experimental H2, CO, and formate energetics on Cu(111) with up to 28-qubit active spaces via DFET, MBECAS-SR, QSCI, and NEVPT2.

  2. Machine learning for sample-based quantum diagonalization: generative configuration recovery and the classical-simulability frontier

    quant-ph 2026-08 conditional novelty 6.0 of 10

    A critical review plus small exact-FCI experiments concludes that sample-based quantum diagonalization has not beaten classical selected CI and maps where, if anywhere, a quantum or generative advantage could survive.

  3. Enhanced Neural Quantum State via Annealed Gradient Descent

    quant-ph 2026-07 reject novelty 5.0 of 10

    AGD reweights sampled configurations with a temperature-annealed factor to avoid 'subspace trapping' in neural quantum state optimization, claimed to reach chemical accuracy on small molecules and high accuracy on fru...

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · cited by 3 Pith papers

  1. [1]

    https://github.com/xzzeng001/QiankunNet-QSCI qiankunnet-qsci

    xzzeng. https://github.com/xzzeng001/QiankunNet-QSCI qiankunnet-qsci. (2025). Materials and Methods

    work page 2025

  2. [2]

    Unitary Selected Configuration Interaction Ansatz The Unitary Selected Configuration Interaction (USCI) ansatz is designed to focus measurement probability on the chemically dominant determinant sector, rather than to approximate the full wavefunction uniformly across Hilbert space. In this sense, USCI sh ould be viewed as the front -end sampler of the Qi...

    work page

  3. [3]

    Quantum hardware implementation and experimental parameters We implemented the LUCJ and USCI ansatz on the Zuchongzhi 3.1 superconducting quantum processor. The purpose of the hardware stage is not to solve the full electronic problem by direct variational optimization alone, but to generate an informative determinant pool for subsequent QiankunNet refine...

    work page

  4. [4]

    Unitary Selected Configuration Interaction (USCI) Ansatz 1.1 Wavefunction formulation 1.2 Unitary factorization and determinant-pair rotations 1.3 Quantum Circuit Design and Configuration Selection 1.4 Circuit Implementation and Parameter optimization

    work page

  5. [5]

    USCI versus LUCJ: Theory, Circuit Construction, and Benchmark Results 2.1 Analytical comparison of unitary generators and gate counts 2.2 Numerical performance on representative molecules

    work page

  6. [6]

    General performance of determinant-focused sampling 3.1 Transferability of USCI across representative strongly correlated molecules 3.2 Sampling bias, noise robustness, and QiankunNet reweighting 3.3 H-Couple expansion for quantum-sampled determinant spaces

    work page

  7. [7]

    Unified resource-error model for focused quantum sampling on NISQ hardware 4.1 Deterministic subspace-truncation error 4.2 Gate-noise-induced probability redistribution 4.3 Finite‑shot uncertainty for a fixed determinant set 4.4 Uniform sampling versus structured quantum sampling

    work page

  8. [8]

    QiankunNet architecture, variational optimization, and computational settings 5.1 Wavefunction parameterization and training objective 5.2 Model hyperparameters for [2Fe-2S] and P-cluster 5.3 GPU hardware and computational cost

    work page

Show all 36 references
  1. [9]

    Quantum experimental and QiankunNet-QSCI simulations 6.1 H10: hardware validation of determinant-focused sampling 6.2 [2Fe-2S]: from focused sampling to chemically meaningful convergence 6.3 P-cluster: spin-factorized hardware sampling at large scale

  2. [10]

    Wavefunction analysis of [2Fe-2S] 7.1 Spin-orbital entanglement entropy 7.2 Mutual information

  3. [11]

    The ansatz contains two conceptually distinct ingredients 𝑘̂ and 𝑅̂

    Unitary Selected Configuration Interaction (USCI) Ansatz 1.1 Wavefunction Formulation In the USCI formulation, the wavefunction is generated by applying the unitary operators to a reference determinant, typically the Hartree-Fock state: |Ψ⟩ =𝑒−𝑘̂ 𝑒−𝑅̂ |Φ0⟩ (1) This form is clo...

  4. [12]

    USCI versus LUCJ: Theory, Circuit Construction, and Benchmark Results To place USCI in context, we compare it with the Local Unitary Cluster -Jastrow (LUCJ) ansatz, which has emerged as an important hardware-aware variational form for near-term quantum chemistry. The purpose o...

  5. [13]

    The important distinction is that the noisy measurement histogram is not itself the final energy estimator

    Unified resource-error model for focused quantum sampling on NISQ hardware The accuracy of focused quantum sampling in QSCI is governed by three coupled factors: (i) the exact wavefunction weight retained in the selected determinant subspace, (ii) the redistribution of measure...

  6. [14]

    (38) and Eq. (36) reduces to the asymptotic estimate 𝐸𝑅 − 𝐸0 ≤ 2Λ𝐻√𝜂𝑅 (38) This bound formalizes the principle of focused determinant sampling: a compact determinant set can yield a small variational error if it retains most of the exact ground-state wavefunction weight. 4.2 G...

  7. [15]

    QiankunNet architecture, variational optimization, and computational settings 5.1 Wavefunction parameterization and training objective QiankunNet serves as the classical refinement module in the QiankunNet -QSCI workflow. After determinant-focused quantum sampling with the USC...

  8. [16]

    This system is large enough to exhibit a nontrivial determinant distribution, yet small enough that full configuration interaction (FCI) remains available as an exact benchmark

    Quantum experimental results and QiankunNet-QSCI simulations 6.1 H10: hardware validation of determinant-focused sampling We examine the H 10 ring in an active space of 10 electrons in 10 orbitals as a controlled test case for hardware-assisted determinant focusing. This syste...

  9. [17]

    Wavefunction analysis of [2Fe-2S] 7.1 Spin-orbital entanglement entropy We formulate the entropy and mutual -information measures at the spin -orbital level. For a given spin orbital i, we define the marginal occupation probability as the total probability of all determinants ...

  10. [18]

    & Olsen, J

    Helgaker, T., Jørgensen, P. & Olsen, J. Molecular Electronic -Structure Theory. (John Wiley & Sons, Ltd 2000)

    2000

  11. [19]

    Jordan-Wigner Transformation for Quantum-Spin Systems in Two Dimensions and Fractional Statistics

    Fradkin, E. Jordan-Wigner Transformation for Quantum-Spin Systems in Two Dimensions and Fractional Statistics. Phys. Rev. Lett. 63 (3), 322-325(1989)

    1989

  12. [20]

    A.; Tubman, N

    Holmes, A. A.; Tubman, N. M.; Umrigar, C. J. Heat-Bath Configuration Interaction: An Efficient Selected Configuration Interaction Algorithm Inspired by Heat -Bath Sampling. J. Chem. Theory Comput. 12, 3674- 3680(2016)

    2016

  13. [21]

    Arrazola, J. M. et al. Universal quantum circuits for quantum chemistry. Quantum 6, 742 (2022)

    2022

  14. [22]

    Gao, D. et al. Establishing a New Benchmark in Quantum Computational Advantage with 105 -qubit Zuchongzhi 3.0 Processor. Phys. Rev. Lett. 134, 090601 (2025)

    2025

  15. [23]

    J., Whaley, K

    Motta, M., Sung, K. J., Whaley, K. B., Head -Gordon, M., & Shee, J. Bridging physical intuition and hardware efficiency for correlated electronic states: the local unitary cluster Jastrow ansatz for electronic structure. Chem. Sci. 14, 11213-11227 (2023)

    2023

  16. [24]

    & Zapata-Rivera, J

    Varela Lambraño, R., Vivas-Reyes, R., Visbal, R. & Zapata-Rivera, J. Evaluation of the electronic structure and charge transfer in the Cu2O22+ core using multiconfigurational methods. Theor. Chem. Acc. 139, 56 (2020)

    2020

  17. [25]

    G., Feldt, M

    Keuter, J., Hepp, A., Daniliuc, C. G., Feldt, M. & Lips, F. Cycloadditions with a Stable Charge -Separated Cyclobutadiene-Type Amido-Substituted Silicon Ring Compound. Angew. Chem. Int. Ed. 60, 21761–21766 (2021)

    2021

  18. [26]

    & Kats, D

    Vitale, E., Alavi, A. & Kats, D. FCIQMC -Tailored Distinguishable Cluster Approach. J. Chem. Theory Comput. 16, 5621–5634 (2020)

    2020

  19. [27]

    F., Jørgensen, S., Johnson, M

    Jespersen, M. F., Jørgensen, S., Johnson, M. S. & Mikkelsen, K. V . Bypassing the multireference character of singlet molecular oxygen, part 1:1,4-cyclo-addition. Int. J. Quantum Chem. 121, e26523 (2021)

    2021

  20. [28]

    & Gagliardi, L

    Li Manni, G., Ma, D., Aquilante, F., Olsen, J. & Gagliardi, L. SplitGAS method for strong correlation and the challenging case of Cr2. J. Chem. Theory Comput. 9, 3375–3384 (2013)

    2013

  21. [29]

    Stair, N. H. & Evangelista, F. A. Exploring Hilbert space on a budget: Novel benchmark set and performance metric for testing electronic structure methods in the regime of strong correlation. J. Chem. Phys. 153, 104108 (2020)

    2020

  22. [30]

    Temme, K., Bravyi, S., & Gambetta, J. M. Error mitigation for short -depth quantum circuits. Phys. Rev. Lett. 119, 180509 (2017)

    2017

  23. [31]

    and Zhou, Y ., Hamiltonian -Guided Autoregressive Selected -Configuration Interaction Achieves Chemical Accuracy in Strongly Correlated Systems

    Zhang, H., Zeng, X., Li, Z. and Zhou, Y ., Hamiltonian -Guided Autoregressive Selected -Configuration Interaction Achieves Chemical Accuracy in Strongly Correlated Systems. J. Chem. Theory Comput. 21, 12622-12633 (2025)

    2025

  24. [32]

    Mitrushenkov, A. O. Second -order Epstein -Nesbet correction to dynamic configuration interaction energies. J. Chem. Phys. 105, 10487-10492 (1996)

    1996

  25. [33]

    Probability Inequalities for Sums of Bounded Random Variables

    Hoeffding, W. Probability Inequalities for Sums of Bounded Random Variables. J. Am. Stat. Assoc. 58, 13-30 (1963)

    1963

  26. [34]

    Chemistry beyond the scale of exact diagonalization on a quantum -centric supercomputer

    Javier Robledo-Moreno et al. Chemistry beyond the scale of exact diagonalization on a quantum -centric supercomputer. Sci. Adv. 11, eadu9991 (2025)

    2025

  27. [35]

    & Reiher, M

    Boguslawski, K., Tecmer, P., Barcza, G., Legeza, Ö. & Reiher, M. Orbital entanglement in bond-formation processes. J. Chem. Theory Comput. 9, 2959–2973 (2013)

    2013

  28. [36]

    & Tecmer, P

    Boguslawski, K. & Tecmer, P. Orbital entanglement in quantum chemistry. Int. J. Quantum Chem. 115, 1289–1295(2015)

    2015

Pith tools

Reviewed June 30, 2026 · model on record in the stance chip above.