Pith. sign in

REVIEW 3 major objections 4 minor 41 references

DA-CASE: reusable measurements for adaptive quantum subspaces

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read DA-CASE shows that one cached Pauli bank can rebuild all subspace matrices, and that generator resolution alone cuts the retained bank from 7,371 to 2,240 words on H4.

desk verdict Careful, well-scoped measurement-accounting paper whose central bank-size result rests on private code; deserves peer review if the artifact snapshot actually gets provided. read the letter →

arxiv 2608.08739 v1 pith:L7KNPS2H submitted 2026-08-09 quant-ph

Ginanjar Utama , Hermawan Kresno Dipojono This is my paper · ORCID
classification quant-ph MSC 81P6865F15
keywords adaptivequantumsubspaceeigensolvermeasurementreusePauliexpectationbankgeneratorresolutionreference-conditionedsymmetrydyadiccommutinghierarchyfinite-shotcovarianceallocationoverlapregularization
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 tries to establish a measurement architecture for quantum subspace diagonalization in which every projected quantity—overlap, Hamiltonian, and observables—is reconstructed by linear combinations of one cached set of Pauli expectations on a single reference state. The paper argues that the dominant costs of such methods are not raw basis size, but measurement contexts, settings, shots, circuit depth, and post-selection retries, and it builds a resource ledger to keep those units separated. On a frozen eight-qubit H4 Hamiltonian, it shows that two generator resolutions, determinant-level and Pauli-word-level, converge to the same nine-dimensional subspace and the same energy to machine precision, while the retained measurement bank falls from 7,371 to 2,240 Pauli words. The contribution is a demonstration of measurement reuse with an explicit ledger, explicitly limited to small exact and Monte Carlo instances.

What carries the argument

The load-bearing object is the single-reference Pauli expectation bank: for a reference $\rho$, each matrix element of $S$, $H$, and $Q$ is expanded in the Hermitian Pauli basis, and all entries are assembled from expectations $\langle P_w\rangle$ on the same $\rho$. Three mechanisms carry the argument. Generator resolution is the choice between scoring a candidate as a complete determinant excitation or as its component Pauli words; this changes the retained word universe and setting count without changing the solved span. The reference-conditioned symmetry certificate, $\ell_B=\lambda_{\max}(L,S)$ with $L_{ij}=\operatorname{Tr}[\rho A_i^\dagger(1-P_q)A_j]$, certifies sector leakage of every normalized combination in the retained subspace, even when individual words fail an operator-global commutator test. The dyadic block-commuting hierarchy groups Pauli words that commute on each contiguous block, interpolating from qubit-wise commuting at block size 1 to full Pauli commutation at block size $n$, and trades setting count for logical-CX depth.

What would settle it

Take a reference with known sector weight below 1 (for instance, a superposition across particle-number sectors) and run the word-resolution pipeline: if the reconstructed $S$ and $H$ matrices deviate from the exact sector-projected matrices by more than numerical tolerance, or if the computed $\ell_B$ from Eq. (9) fails to match direct projection of the retained Ritz vectors, the single-bank exactness claim is refuted for sector-mixed references.

Watch

Extended reading notes

Core claim

The central claim is that for a fixed reference $\rho$, the projected matrices $S_{ij}=\operatorname{Tr}(\rho A_i^\dagger A_j)$, $H_{ij}=\operatorname{Tr}(\rho A_i^\dagger H A_j)$, and observable matrices $Q_{ij}=\operatorname{Tr}(\rho A_i^\dagger Q A_j)$ can all be reconstructed from the same cached Pauli expectations on $\rho$, with no separately prepared basis states and no Hadamard tests. The paper then claims that generator resolution is a real resource variable: resolving a determinant excitation into component Pauli words, rather than scoring it as one operator, reaches the same retained subspace and energy on H4—energies agree to about $4\times10^{-16}$ Ha—while the retained word universe drops from 7,371 to 2,240 and the qubit-wise-commuting setting count from 913 to 465. The narrower bank is licensed by a reference-conditioned sector certificate, $\ell_B=\lambda_{\max}(L,S)$, which bounds leakage of the whole retained span even when individual Pauli words fail the global commutator test. The paper also reports auxiliary trades: a dyadic block-commuting hierarchy reduces settings from 913 to 64 at the price of logical CX gates, and covariance-aware shot allocation lowers a projected-matrix variance target by 68.9% in a four-qubit finite-shot diagnostic.

Load-bearing premise

The whole reconstruction and symmetry certificate assumes the reference state is a sharp eigenstate of the particle-number and $S_z$ sector projectors; when the reference is sector-mixed, exactness requires a QND post-selection whose ancilla, gate, and latency costs the paper does not price.

Editorial extensions

If this is right

  • A single cached bank can serve overlap, Hamiltonian, and multiple observable matrices, so subsequent solves or observable evaluations on the same reference reuse the same set of physical shots rather than requiring new prepared basis states.
  • Generator resolution becomes a practical tuning knob: word-level scoring can shrink the final measurement bank by roughly a factor of three on H4 while preserving the solved subspace, at the cost of more extensive selection work.
  • A reference-conditioned certificate can admit Pauli directions that a global symmetry filter rejects, provided the reference is sharp in the declared sector, and it certifies the full retained span rather than individual words.
  • The dyadic commuting hierarchy gives a monotone family of operating points from 913 settings to 64 settings on the retained H4 bank, with explicitly counted logical-CX gates and depths, so a hardware-specific optimum can be chosen once gate and noise data exist.
  • Covariance-aware shot allocation and mode-wise overlap regularization reduce catastrophic finite-shot outliers and RMSE while increasing median error, indicating a tunable bias–tail-risk trade-off rather than an unconditional accuracy improvement.

Reading between the lines

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

  • A testable extension is to reuse the wide selection cache across a family of related Hamiltonians, such as nearby geometries or embedding parameters, which the paper lists as a possible next step but does not test; the retained-bank ratios on H4 and the 2x2 Hubbard plaquette hint that the saving is system-dependent.
  • If QND post-selection for sector-mixed references turns out to be expensive on real hardware, the warm-start rows of the ledger would grow by more than the stated retry factor, potentially reversing the apparent advantage over multi-reference methods; the paper explicitly leaves that circuit unpriced.
  • Adopting the paper's accounting rule—keep contexts, settings, shots, depth, and retries as separate ledger columns—would make future subspace-method comparisons more informative than basis-dimension or energy-error comparisons alone.
  • The exact span coincidence between determinant and word resolutions is striking but only shown on small systems; a natural check is to test whether the two resolutions continue to converge to the same subspace on larger molecules, which would let cost alone decide the resolution choice.
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 / 4 minor

Summary. This paper introduces DA-CASE, a single-reference quantum subspace eigensolver in which basis states are virtual Clifford-algebra directions A_i|ψ⟩ and the projected overlap, Hamiltonian, and observable matrices are reconstructed from one cached set of Pauli expectations on |ψ⟩. The central numerical contract is a frozen eight-qubit H4 Hamiltonian: determinant-resolution and word-resolution generator families both reach a nine-dimensional subspace with energy error 3.019 mHa and energy difference 4×10^-16 Ha, while the retained Pauli bank shrinks from 7,371 to 2,240 words and the QWC setting count from 913 to 465. The paper also proposes a reference-conditioned sector-leakage certificate, a dyadic block-commuting measurement hierarchy that reduces QWC settings to 64 fully commuting settings at explicit logical-CX cost, and finite-shot diagnostics on a four-qubit TFIM bank in which covariance-aware allocation reduces a variance target by 68.9% and mode-wise overlap regularization trades median error for tail-risk control. The authors explicitly limit the claims to small-instance exact and Monte Carlo results and disclaim scaling, hardware, or quantum-advantage claims.

Significance. If the central numerical contract is correct, the paper makes a useful conceptual point: generator resolution is an independent resource variable that can change the measurement bank without changing the solved subspace, and fixed-reference reconstruction separates state contexts from settings, shots, and depth. The strengths are real: exact projected eigensolves are checked against independent PySCF and OpenFermion constructions; the finite-shot study uses 200 Monte Carlo replicas with grouped resampling; evidence labels (exact, oracle-sampled, finite-sample, heuristic) are carried through the records; and the limitations are unusually candid, including no scaling result and no end-to-end advantage claim. The main weakness is that the headline H4 numbers are produced by a private codebase, with only the Hamiltonian energy and Pauli-term expansion cross-checked externally; the subspace-equality and retained-bank counts are not independently auditable as presented.

major comments (3)
  1. [§VII Reproducibility; §IV.B] The central H4 contract—identical nine-dimensional span, principal angles zero, retained banks 7,371 vs 2,240—is generated by the private clifford_qc repository. The internal digest checker verifies only that printed tables match committed JSON records; it does not certify that the records are correct, and the external cross-checks (PySCF/OpenFermion) cover the mapped energy and 185-term Hamiltonian, not the subspace-equality or bank-count claims. Since these counts are the paper's headline result, the frozen artifact should be made publicly available with a persistent identifier, or a minimal independent script should be included that builds the two generator families from the FCIDUMP and recomputes principal angles and retained word universes. Without such an audit path, the load-bearing numerical evidence is not independently verifiable.
  2. [§IV.B] The statement that the two resolutions reach the same subspace 'to machine precision' is not quantified: the nine principal angles are reported only as zero to numerical precision, and the claimed energy difference is 4×10^-16 Ha but no numerical tolerance or actual maximum angle is given. Please report the largest principal angle (or equivalently the smallest singular value of the overlap between the two spans), the exact energy difference, and the thresholds used for rank and overlap. The entire generator-resolution conclusion rests on this equality, so the evidence should be quantitative in the text rather than asserted qualitatively.
  3. [§II.C; §V] The reference-conditioned certificate in Eqs. (8)–(9) and the exactness of the reconstructed matrices presume a sector-sharp reference, or a post-selected ρ_q obtained via a quantum non-demolition measurement of N and Sz whose ancilla, gate, and latency costs are explicitly unpriced. The warm-start rows of Table I use algebraic post-selection, so their unchanged setting counts do not reflect the physical cost of realizing ρ_q. Please state explicitly that these rows are conditional on a QND sector measurement being available at negligible overhead, or extend the ledger with a lower-bound cost model for that measurement.
minor comments (4)
  1. [Table I] The header 'ArmMError' runs the arm and error columns together; separate the column headings to avoid ambiguity.
  2. [Table I] The abbreviation 'Rot.' in the table header is not defined in the text; define it in the caption, for example as preparation rotors corresponding to Dprep.
  3. [§II.A] In Eq. (5), the symbol W denotes both the word-universe set and its cardinality; use a distinct symbol for the set and write W=|W_set| to avoid confusion.
  4. [§II.B] In Eq. (7), the eigenvalue variable λ is introduced without a label; clarify that it is the smallest root of the generalized 2×2 pencil used in the candidate score.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DA-CASE's reconstruction equations, H4 subspace comparison, and symmetry certificate are exact algebraic computations with external energy checks; the self-citation [32] is methodological, not load-bearing.

full rationale

The central equations (2)-(4) define S, H, and Q as linear reconstructions from one cached set of Pauli expectations on a single reference; this is the method's architecture rather than a derived prediction. The headline H4 result (retained bank 7371 vs 2240, same nine-dimensional subspace and energy to machine precision) is obtained from exact generalized eigensolves and a direct principal-angle comparison, not from fitting any parameter to the claimed output. The reference-aware symmetry test (Eqs. 8-9) is an explicit certificate computed from rho and the projector Pq, and the finite-shot overlap rule is explicitly labeled a heuristic regularizer rather than a variational certificate. The one self-citation, [32], describes the A-CASE representation on which the implementation builds; it is not used as evidence for the present numerical results, which are cross-checked against PySCF and OpenFermion for the H4 energy. The private artifact is a reproducibility and verification limitation, not a circular derivation: the digest checker confirms internal consistency but does not establish external correctness. No step reduces a claimed output to its own input by construction.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central numerical results rest on standard linear algebra and the availability of exact Pauli expectations on a reference. The method-specific assumptions are the sector-sharpness of the reference and the extra cost of QND post-selection, both admitted in Section V.

free parameters (3)
  • Overlap retention thresholds τS and λmax/κmax = not stated in the paper
    Equation (12) keeps mode k when λk > max(τS, λmax/κmax, rk). The thresholds are chosen heuristically and no sensitivity analysis is reported.
  • Pilot shot count for covariance-aware allocation = 200 shots per group
    Section IV.D fixes a 200-shot pilot per group before Neyman allocation (Eq. 13); no criterion for this number is given.
  • Mode-wise overlap cutoff rule = data-derived per-mode radius rk
    The calibrated arm selects rk from the same data used in the solve, so the retained rank is data-dependent; the authors state this is a heuristic regularizer, not a certificate.
assumptions (3)
  • domain assumption The input is an effective many-body Hamiltonian in FCIDUMP format, assumed to already incorporate prior DFT, Wannier, screening, or embedding steps.
    Section I defines the intended input boundary; DA-CASE does not construct this Hamiltonian.
  • domain assumption The reference state |ψ> is a sharp eigenstate of the sector projectors Pq for q=(N,Sz), or can be post-selected with a QND measurement whose cost is not priced.
    Section II.C uses Eq. (10) for sector-mixed references and Section V lists the QND circuit as an unpriced cost; the symmetry certificate Eq. (9) assumes a sharp reference.
  • domain assumption Pauli expectations on the reference can be measured with sufficient accuracy that noiseless reconstruction is representative.
    Exact projected solves are used for the main H4 claim; the finite-shot TFIM study is a separate diagnostic, and there is no hardware demonstration.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DA-CASE: reusable measurements for adaptive quantum subspaces." pith.science (2026). https://pith.science/paper/L7KNPS2H

@misc{pith2026260808739,
  author       = {Pith},
  title        = {Pith review of: DA-CASE: reusable measurements for adaptive quantum subspaces},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L7KNPS2H}},
  note         = {Machine review of arXiv:2608.08739}
}
abstract

Quantum subspace methods are often compared by basis dimension or energyerror, although their dominant experimental costs arise from different statepreparations, measurement settings, and shot allocations. We present theDyadic Adaptive Clifford-Algebra Subspace Eigensolver (DA-CASE), whose basisstates are virtual directions $A_i|\psi\rangle$ generated from one reference.Overlap, Hamiltonian, and observable matrices are reconstructed from onecached set of Pauli expectations on that reference. The method thereforetrades multiple prepared basis states for a potentially wide measurement bank.We make that trade explicit on a frozen eight-qubit H$_4$ Hamiltonian. Twogenerator resolutions reach the same nine-dimensional subspace and the sameenergy to machine precision, while the retained bank changes from 7371 to 2240Pauli words. A reference-conditioned symmetry test certifies the narrower spanwithout asserting that its individual Pauli words conserve the sector asabstract operators. Independently, a dyadic commuting hierarchy reduces thedeterminant bank from 913 qubit-wise-commuting settings to 64 fully commutingsettings, while exposing the added logical-CX cost. In a separate four-qubitfinite-shot diagnostic, covariance-aware allocation reduces theprojected-matrix variance target by 68.9%. Mode-wise overlap regularizationremoves the observed catastrophic energy estimates and lowers RMSE, butdoubles the median error relative to a fixed cutoff. These are small-instanceexact and Monte Carlo results, not a hardware demonstration, scaling result,or quantum advantage claim. The contribution is a single-referencemeasurement architecture and a resource ledger that keeps contexts, settings,shots, circuit depth, and post-selection retries in their proper units.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 31 canonical work pages

  1. [1]

    This removes separately prepared basis states but can enlarge the Pauli universe

    All projected matrices are linear reconstructions from one expectation bank on one reference. This removes separately prepared basis states but can enlarge the Pauli universe

    work page

  2. [2]

    On H4, coarse determinant excitations and their Pauli- word resolution reach the same retained subspace with widths 7371 and 2240

    Generator resolution is a real resource variable. On H4, coarse determinant excitations and their Pauli- word resolution reach the same retained subspace with widths 7371 and 2240

    work page

  3. [3]

    A reference-aware sector test accepts state-safe Pauli directions that fail a stronger operator-global test, and certifies the entire retained span

    work page

  4. [4]

    DA-CASE: reusable measurements for adaptive quantum subspaces

    Measurement grouping and overlap regularization expose separate depth–setting and bias–tail-risk trades; neither implies an end-to-end advantage. The intended input boundary is an effective many- body Hamiltonian, for example one obtained after density- functional, Wannier, screening, or embedding steps [ 15– 18]. DA-CASE is not a DFT method and does not ...

    work page Pith review arXiv 2026

  5. [5]

    Reuse across observables or repeated solves is needed to amortize that cost

    Selection can dominate.The H 4 selection cache contains 14401–15803 words, much more than the retained bank. Reuse across observables or repeated solves is needed to amortize that cost

    work page

  6. [6]

    Conditioning remains physical.Exact nested sub- spaces are variational, but noisy pencils are not. In the auxiliary response test, increasing κS by four orders reduces successful bootstrap replicas from 200/200 to 137 /200 and widens a susceptibility in- terval by about fifteen times

    work page

  7. [7]

    Exact-ground-state samples are an oracle input, not an implementable preparation claim

    Sampling-fed selection is not an established gain.On the 2 × 3 Hubbard test at M = 7, operator dress- ing is indistinguishable from a sample-independent selected-CI control. Exact-ground-state samples are an oracle input, not an implementable preparation claim

    work page

  8. [8]

    Finite-shot adaptive growth at eight qubits remains costly and unreliable in the present implementation

    No scaling result is shown.All systems admit exact classical validation. Finite-shot adaptive growth at eight qubits remains costly and unreliable in the present implementation. These limitations prevent a total-resource comparison with ADAPT-VQE. Its physical gradient protocol and precision-matched group variances are not specified, while DA-CASE’s QND p...

    work page

Show all 41 references
  1. [9]

    O’Leary, L

    T. O’Leary, L. W. Anderson, D. Jaksch, and M. Kiffner, Quantum9, 1726 (2025)

    2025

  2. [10]

    J. R. McClean, M. E. Kimchi-Schwartz, J. Carter, and W. A. de Jong, Physical Review A95, 042308 (2017)

    2017

  3. [11]

    W. J. Huggins, J. Lee, U. Baek, B. O’Gorman, and K. B. Whaley, New Journal of Physics22, 073009 (2020)

    2020

  4. [12]

    N. H. Stair, R. Huang, and F. A. Evangelista, Journal of Chemical Theory and Computation16, 2236 (2020)

    2020

  5. [13]

    Bharti and T

    K. Bharti and T. Haug, Physical Review A104, L050401 (2021)

    2021

  6. [14]

    E. N. Epperly, L. Lin, and Y. Nakatsukasa, SIAM Journal on Matrix Analysis and Applications43, 1263 (2022)

    2022

  7. [15]

    G. Lee, D. Lee, and J. Huh, Quantum8, 1477 (2024)

    2024

  8. [16]

    Zhang, A

    Z. Zhang, A. Wang, X. Xu, and Y. Li, Quantum8, 1438 (2024)

    2024

  9. [17]

    N. V. Tkachenko, L. Cincio, A. I. Boldyrev, S. Tretiak, P. A. Dub, and Y. Zhang, Quantum Science and Technol- ogy9, 035012 (2024). 7

    2024

  10. [18]

    Georges, G

    A. Georges, G. Kotliar, W. Krauth, and M. J. Rozenberg, Reviews of Modern Physics68, 13 (1996)

    1996

  11. [19]

    Feniou, M

    C. Feniou, M. Hassan, D. Traore, E. Giner, Y. Maday, and J.-P. Piquemal, Overlap-adapt-vqe: Practical quan- tum chemistry on quantum computers via overlap-guided compact ans¨ atze (2023), arXiv:2301.10196 [quant-ph]

    2023 arXiv

  12. [20]

    Miura, Active sampling sample-based quantum di- agonalization from finite-shot measurements (2026), arXiv:2603.13536 [quant-ph]

    R. Miura, Active sampling sample-based quantum di- agonalization from finite-shot measurements (2026), arXiv:2603.13536 [quant-ph]

    2026 arXiv

  13. [21]

    Peruzzo, J

    A. Peruzzo, J. McClean, P. Shadbolt, M.-H. Yung, X.-Q. Zhou, P. J. Love, A. Aspuru-Guzik, and J. L. O’Brien, Nature Communications5, 4213 (2014)

    2014

  14. [22]

    H. R. Grimsley, S. E. Economou, E. Barnes, and N. J. Mayhall, Nature Communications10, 3007 (2019)

    2019

  15. [23]

    Zheng, B

    M. Zheng, B. Peng, A. Li, X. Yang, and K. Kowalski, npj Quantum Information10, 127 (2024)

    2024

  16. [24]

    Marzari, A

    N. Marzari, A. A. Mostofi, J. R. Yates, I. Souza, and D. Vanderbilt, Reviews of Modern Physics84, 1419 (2012)

    2012

  17. [25]

    Pizzi, V

    G. Pizzi, V. Vitale, R. Arita,et al., Journal of Physics: Condensed Matter32, 165902 (2020)

    2020

  18. [26]

    Aryasetiawan, M

    F. Aryasetiawan, M. Imada, A. Georges, G. Kotliar, S. Biermann, and A. I. Lichtenstein, Physical Review B70, 195104 (2004)

    2004

  19. [27]

    Kanno, M

    K. Kanno, M. Kohda, R. Imai, S. Koh, K. Mitarai, W. Mizukami, and Y. O. Nakagawa, Physical Review Research8, 023268 (2026)

    2026

  20. [28]

    P. J. Knowles and N. C. Handy, Computer Physics Com- munications54, 75 (1989)

    1989

  21. [29]

    Q. Sun, T. C. Berkelbach, N. S. Blunt, G. H. Booth, S. Guo, Z. Li, J. Liu, J. D. McClain, E. R. Sayfutyarova, S. Sharma, S. Wouters, and G. K.-L. Chan, WIREs Com- putational Molecular Science8, e1340 (2018)

    2018

  22. [30]

    J. R. McClean, N. C. Rubin, K. J. Sung,et al., Quantum Science and Technology5, 034014 (2020)

    2020

  23. [31]

    Verteletskyi, T.-C

    V. Verteletskyi, T.-C. Yen, and A. F. Izmaylov, Journal of Chemical Physics152, 124114 (2020)

    2020

  24. [32]

    Crawford, B

    O. Crawford, B. van Straaten, D. Wang, T. Parks, E. Campbell, and S. Brierley, Quantum5, 385 (2021)

    2021

  25. [33]

    Miller, L

    D. Miller, L. E. Fischer, K. Levi, E. J. Kuehnke, I. O. Sokolov, P. K. Barkoutsos, J. Eisert, and I. Tavernelli, npj Quantum Information10, 122 (2024)

    2024

  26. [34]

    Efron, The Annals of Statistics7, 1 (1979)

    B. Efron, The Annals of Statistics7, 1 (1979)

    1979

  27. [35]

    Y. O. Nakagawa, M. Kamoshita, W. Mizukami, S. Sudo, and Y.-y. Ohnishi, Journal of Chemical Theory and Com- putation20, 10817 (2024)

    2024

  28. [37]

    Gaberle and M

    C. Gaberle and M. S. Jattana, A critical assessment of the sample-based quantum diagonalization for heisenberg and hubbard models (2026), arXiv:2605.02494 [quant-ph]

    2026 arXiv

  29. [38]

    J. I. Colless, V. V. Ramasesh, D. Dahlen, M. S. Blok, M. E. Kimchi-Schwartz, J. R. McClean, J. Carter, W. A. de Jong, and I. Siddiqi, Physical Review X8, 011021 (2018)

    2018

  30. [39]

    Umeano, F

    C. Umeano, F. Jamet, L. P. Lindoy, I. Rungger, and O. Kyriienko, Physical Review Materials9, 034401 (2025)

    2025

  31. [40]

    Patel, P

    S. Patel, P. Jayakumar, R. Huang, T. Zeng, and A. F. Izmaylov, Journal of Chemical Theory and Computation 22, 3937 (2026)

    2026

  32. [41]

    Utama and H

    G. Utama and H. K. Dipojono, arXiv preprint arXiv:2607.17443 10.48550/arXiv.2607.17443 (2026)

    2026 doi arXiv

  33. [256]

    The same checker verifies balanced environments, references, citations, and table column counts

    Every table fragment this manuscript inputs is gen- erated from committed JSON records and carries a blob digest of each source record and of the generator itself, which a structural checker re-derives; a record regener- ated without rerunning the generator therefore fails rat...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.