Pith. sign in

REVIEW 4 major objections 4 minor 79 references

Attention in Krylov Space: Transformer-Based Extrapolation of Lanczos Coefficients

T0 review · 4 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 A transformer trained on short Lanczos prefixes extrapolates operator-growth coefficients with roughly ten times lower error than asymptotic fits, and transfers without retraining from small to larger systems.

desk verdict Solid in-horizon forecasting of Lanczos coefficients, but the 'deep Krylov space' claim is untested: the model never predicts beyond the largest index it saw in training. read the letter →

arxiv 2601.07937 v2 pith:NLXOMAAM submitted 2026-01-12 quant-ph cond-mat.stat-mech

Zihao Qi , Christopher Earls This is my paper
classification quant-phcond-mat.stat-mech
keywords LanczoscoefficientsKrylovcomplexityoperatorgrowthtransformerquantummany-bodychaossequenceextrapolationuniversalhypothesisautocorrelationfunction
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 claims that Lanczos coefficients—the hopping amplitudes governing operator growth in quantum and classical dynamics—form a causal sequence with long-range, history-dependent structure that standard asymptotic fits miss. It trains a small transformer on short prefixes of these coefficients and uses it to predict the long tail autoregressively. Across chaotic spin models, the learned extrapolation is roughly an order of magnitude more accurate than the standard linear-asymptotic fit, and the same model trained on one system size transfers to a larger size without retraining. If correct, this gives a practical surrogate for probing operator dynamics deep in Krylov space, where direct Lanczos iteration is numerically unstable and memory-bound. The paper also identifies which parts of the coefficient history matter most by inspecting learned attention weights and ablating them.

What carries the argument

The central object is a decoder-only transformer with masked multi-head self-attention—a sequence model that assigns learned weights to all past coefficients when predicting the next one—applied to the causal sequence of coefficient differences Δb_n = b_n − b_{n−1}. Predicting differences rather than raw b_n stabilizes the target scale, and the causal mask enforces that predictions depend only on earlier coefficients. Attention then captures long-range correlations that the iterative Lanczos recursion implicitly propagates.

What would settle it

Recompute the long-tail coefficients for the same L=8 Hamiltonians with arbitrary-precision arithmetic or via an independent moment-based recursion, then rerun the RMSE comparison; if the transformer's order-of-magnitude advantage over the asymptotic fit disappears or shrinks substantially, the central claim fails. Also check whether the L=12 'true' sequences used for zero-shot evaluation saturate before n=30—if they do, the benchmark is against truncated labels.

Watch

Extended reading notes

Core claim

The paper claims that Lanczos coefficients are not just an asymptotic ramp; they carry subleading, history-dependent structure—even-odd staggering, slow deviations, and early-coefficient anchors—that standard fits miss. A small decoder-only transformer, trained on the increment sequence Δb_n = b_n − b_{n−1} from thousands of Hamiltonian instances, predicts coefficients for n>10 autoregressively. On unseen instances of the classical XYZ spin top and a chaotic transverse-field Ising chain with L=8, the transformer's coefficient RMSE is consistently an order of magnitude smaller than the universal-operator-growth-hypothesis asymptotic fit with staggering; reconstructed Krylov complexity and aut

Load-bearing premise

The full-reorthogonalization Lanczos output used as ground truth is exact over the entire horizon (T=100 for XYZ, T=30 for TFIM); if loss of orthogonality or finite-size saturation corrupts the long-tail labels, the reported error reductions are measured against corrupted targets.

Editorial extensions

If this is right

  • Coefficient extrapolation can be treated as sequence prediction: a model trained on short prefixes reproduces long Lanczos tails with order-of-magnitude lower RMSE than linear asymptotic fits on chaotic systems.
  • Zero-shot transfer from L=8 to L=12 means training data can be collected at accessible sizes and deployed to larger systems where direct Lanczos iteration is prohibitively expensive.
  • Observable-level reconstruction—Krylov complexity and autocorrelation—inherits the extrapolation accuracy, with errors orders of magnitude smaller at late times.
  • Attention ablations show that removing long-range context, parity information, or the first few coefficients degrades accuracy by an order of magnitude, so Lanczos sequences should be modeled as long-range causal sequences rather than fitted to a local asymptotic form.

Reading between the lines

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

  • One passage to weigh: the abstract credits the model with extrapolating integrable regimes, but the body's Sec. VI lists that as an open direction; a reader should treat the integrable-regime claim as stated but unsupported in the current text.
  • The success of the increment-sequence representation suggests that other spectral quantities derived from autocorrelation moments—not just Lanczos coefficients—may be forecastable by the same autoregressive scheme.
  • If attention weights on early coefficients act as global anchors, then an even shorter input prefix plus system-size metadata might suffice for transfer; this is a testable ablation.
  • A practical deployment would train on small chains and then predict coefficients for a Hamiltonian one cannot iterate; verifying against exactly solvable or sparse Hamiltonians at intermediate sizes would sharpen the zero-shot claim.
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.

Referee Report

4 major / 4 minor

Summary. The paper frames the Lanczos coefficients {b_n} of operator growth as a causal time series and trains a masked decoder-only transformer to autoregressively predict future coefficient differences Δb_n from a short prefix. The model is benchmarked on the classical XYZ top (T=100) and on a chaotic transverse-field Ising chain (L=8, T=30, with an L=12 transfer), and it is compared against the standard asymptotic UOGH fit with an even-odd staggering term. Reported results include RMSE reductions in coefficient prediction, much smaller reconstruction errors in Krylov complexity K(t) and autocorrelation C(t), zero-shot transfer from L=8 to L=12, and attention-map/ablation analyses. The paper's central claim is that a transformer can serve as a practical surrogate for probing operator dynamics deep in Krylov space where brute-force Lanczos iteration is prohibitive.

Significance. If fully substantiated, the work would provide a useful ML-based tool for extending Lanczos sequences, and the zero-shot system-size transfer is a genuinely valuable feature. The paper is clearly written, the architecture is standard and reproducible from the hyperparameter table, and the attention ablation is a nice mechanistic addition. The main strengths are the direct comparison to the standard asymptotic baseline, the validation on two different chaotic models, and the transfer experiment. However, the support for the 'deep Krylov space' claim is incomplete: all evaluations stay within the sequence length T used in training, and the abstract claims an integrable-regime result that the body of the paper does not report. These issues are fixable with additional experiments or appropriately reworded claims, but they are load-bearing for the paper's central contribution as stated.

major comments (4)
  1. [Sec. III, Eq. (24); Sec. IV, Figs. 3-6] The evaluation horizon equals the training horizon. The loss in Eq. (24) trains next-token predictions for all m from n_in to T-1, so every target index m+1 <= T is present in the training sequences. All reported coefficient RMSEs (Figs. 3(b), 4(b), 6) and observable reconstructions (Fig. 5) are for n <= T. Thus the model is tested on continuation within the length distribution seen in training, not on forecasting coefficients for n > T. The L=12 transfer in Fig. 6 is a system-size extrapolation, not a depth extrapolation: the n range is still 1..30. Because the abstract and Secs. I and VI claim the model can probe 'deep in Krylov space' and act 'as a practical surrogate' where direct Lanczos is prohibitive, this is a central unsupported claim. I request an explicit test in which the training sequences are truncated at T_train < T_max and the model is evaluated for T_train < n <= T_max,
  2. [Abstract; Sec. VI] The abstract states: 'The model also accurately extrapolates coefficients in integrable regimes, where no universal asymptotic fit exists.' The full text contains no integrable-regime experiment; Sec. VI lists testing the model on sub-linear asymptotics as a future direction ('Another interesting direction would be to test the model's ability to forecast Lanczos sequences that are not asymptotically linear'). This is an unsupported claim in the abstract and should be removed or backed by the missing experiments.
  3. [Sec. III, Eq. (4); Sec. IV] All ground-truth labels are generated by finite-precision Lanczos with full reorthogonalization, yet the introduction motivates the work by the numerical instability of the same procedure. No independent validation of the long-tail labels is reported (e.g., high-precision arithmetic, orthogonality residuals, or comparison with direct moment computations). If loss of orthogonality or finite-size saturation silently corrupts the labels for n near T (T=100 for XYZ, T=30 for TFIM), the RMSE improvements are measured against corrupted labels and the 'order-of-magnitude' claim would not reflect true coefficient extrapolation. Please add a validation of at least a subset of long sequences, reporting, for example, the norm of the reorthogonalization residual as a function of n.
  4. [Sec. IV, Eq. (27); Figs. 4-6] The RMSE curves are point estimates averaged over N_T=100 test sequences without error bars, bootstrap confidence intervals, or per-sequence quantiles. The central quantitative claims ('order-of-magnitude reduction,' 'orders of magnitude smaller errors') are therefore not accompanied by any uncertainty estimate. Given the large reported separations this is not necessarily fatal, but for a quantitative ML benchmark in a physics journal the authors should report at least bootstrap CIs or the spread of per-sequence errors.
minor comments (4)
  1. [Sec. III, Eq. (13)] The text says 'Under UOGH, the raw coefficients grow without bound, whereas Δb_n approaches an n-independent constant in the asymptotic limit.' This is true for the d != 1 linear asymptote, but for the d=1 case in Eq. (9), b_n ~ α n / log n, so Δb_n ~ α / log n, which tends to 0 rather than a constant. Since the TFIM benchmark uses d=1, please qualify this statement.
  2. [Sec. V, Eq. (20), Fig. 7] The triangular support in the attention maps is a direct consequence of the hard causal mask M=-∞ in Eq. (20), not a learned property of the trained model. The meaningful learned patterns are the diagonal concentration, the checkerboard pattern, and the early-token attention. Please rephrase the sentence 'The observed triangular structure... provides a direct check that the causal constraint is indeed being enforced in the trained model' to avoid implying this is learned.
  3. [Abstract] The abstract says the method achieves 'an order-of-magnitude reduction in error' for 'both classical and quantum chaotic systems.' In the classical XYZ case (Fig. 3(b)) the reported reduction is 'a few times,' not an order of magnitude. Please make the abstract's quantitative claim match the figures.
  4. [General] No code or data availability statement is included. For reproducibility of a numerical ML benchmark, please indicate whether training/test data and code will be released.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: supervised benchmark uses held-out test sequences and no central claim reduces to its inputs.

full rationale

The derivation chain is: generate Lanczos sequences by the recursion (Eq. 4) with full reorthogonalization; train a transformer with the next-step loss (Eq. 24) on 10,000 training sequences; evaluate RMSE (Eq. 27) on 100 unseen sequences; reconstruct observables via Eq. (30). Test sequences are not used in training, so the reported coefficient and observable errors are genuine out-of-sample measurements, not fitted inputs renamed as predictions. The UOGH asymptotic-fit baseline is fit per test sequence, but it is the comparator, not the paper's central claim. The attention ablation fixes the trained weights and alters inference-time masks; this is interpretability, not circular derivation. Self-citations ([25], [54]) appear only in motivating/outlook contexts and are not load-bearing for the transformer extrapolation result. No uniqueness theorem is imported from the authors' prior work. Two non-circular caveats should be weighed: (i) the test horizon is never longer than the training sequence length (T=100 for XYZ, T=30 for TFIM) — Eq. (24) supervises positions n_in..T-1, so 'deep Krylov space' beyond T is not actually tested; (ii) the abstract's claim of accurate extrapolation 'in integrable regimes' has no corresponding experiment in the body. These are support/scope gaps, not circular steps.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim is an empirical ML benchmark. It depends on the numerical exactness of the generated ground-truth coefficients, on the UOGH-based asymptotic baseline, and on distributional similarity across system sizes. No new physical entities are introduced; the transformer is a computational tool, not a postulated physical object.

free parameters (3)
  • Transformer parameters Θ = trained on 10,000 sequences; weights not released
    All learnable weights (embedding, attention, FFN, unembedding) are fit to training data and determine the extrapolation quality; no trained weights or training curves beyond a test-loss plot are provided.
  • Model hyperparameters (L=3, d_model=64, H=4, n_in=10, dropout=0.1, lr=1e-3, epochs=300) = listed in Table I; d_model=64 chosen empirically via Fig. 9
    These are hand- and empirically-chosen settings; the claimed performance depends on them, and no sensitivity analysis is reported for most of them.
  • Asymptotic fit parameters α, γ, γ* (baseline) = least-squares fit per test sequence to first n_in=10 coefficients
    The baseline's extrapolated coefficients are generated from these fitted values (Eqs. 25-26); the central comparison is against this fitted predictor.
assumptions (4)
  • domain assumption The Lanczos algorithm with full reorthogonalization produces numerically exact coefficients up to T=30/100
    Section III: 'Full reorthogonalization [39] is applied at each step to ensure numerical stability.' The entire training/test dataset relies on this; no independent verification against higher-precision arithmetic is given.
  • domain assumption The UOGH asymptotic forms (Eqs. 9, 25-26) are the correct standard baseline for these systems
    The baseline comparison assumes b_n ~ α n/log(n)+γ in d=1 and ~ α n+γ in d≠1, with an even-odd staggering term. If the asymptotic form were mis-specified for the sampled parameter ranges, the baseline would be artificially weak. This is supported only by citation to Ref. [15].
  • domain assumption Sampled TFIM parameters (g/J ∈ [1,2], h/J ∈ [0.1,1]) are all in the chaotic regime with asymptotically linear growth
    Section IV.B: 'Because h≠0 breaks integrability, all sampled Hamiltonians are in the chaotic regime.' The transformer is trained and tested under this assumption; if some samples were near-integrable, the sequence statistics would differ.
  • domain assumption Lanczos sequences across system sizes share learnable statistical structure sufficient for zero-shot transfer
    The zero-shot transfer claim (Fig. 6) assumes that a model trained on L=8 data can predict L=12 sequences; no theory is given, only the empirical figure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Attention in Krylov Space: Transformer-Based Extrapolation of Lanczos Coefficients." pith.science (2026). https://pith.science/paper/NLXOMAAM

@misc{pith2026260107937,
  author       = {Pith},
  title        = {Pith review of: Attention in Krylov Space: Transformer-Based Extrapolation of Lanczos Coefficients},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NLXOMAAM}},
  note         = {Machine review of arXiv:2601.07937}
}
read the original abstract

The Universal Operator Growth Hypothesis formulates time evolution of operators through Lanczos coefficients. In practice, however, numerical instability and memory cost limit the number of coefficients that can be exactly computed. In response to these challenges, the standard approach relies on fitting early coefficients to asymptotic forms, but such procedures can miss subleading, history-dependent structures in the coefficients that subsequently affect reconstructed observables. In this work, we treat the Lanczos coefficients as a causal time sequence and introduce a transformer-based model to autoregressively predict future Lanczos coefficients from short prefixes. For classical and quantum chaotic systems, our model outperforms asymptotic fits in both coefficient extrapolation and physical observable reconstruction, and achieves an order-of-magnitude reduction in error. The model also accurately extrapolates coefficients in integrable regimes, where no universal asymptotic fit exists. Remarkably, our model transfers across system sizes: it can be trained on smaller systems and then be used to extrapolate coefficients on a larger system \emph{without retraining}. By probing the learned attention patterns and performing targeted attention ablations, we identify portions of the coefficient history that are most influential for accurate forecasts. Our results demonstrate that modern sequence models can serve as practical surrogates for probing operator dynamics deep in Krylov space, where brute-force Lanczos iteration can be computationally prohibitive.

Figures

Figures reproduced from arXiv: 2601.07937 by the authors.

Figure 1
Figure 1. FIG. 1. Schematic illustration of a typical sequence of Lanczos [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Transformer performance on the classical XYZ model. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: FIG. 5. Root-mean-squared error (Eq [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 4
Figure 4. Figure 4: FIG. 4. Transformer performance on the quantum [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Zero-shot transferability of the transformer model. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Batch- and layer-averaged attention maps for the [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8. Ablation study on the attention weight matrix. Com [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9. Testing loss during training, for various choices of [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

79 extracted references · 1 canonical work pages

  1. [1]

    zero-shot transferability

    We use open boundary conditions and set J= 1. We sample the dimensionless parametersg/Jand h/Juniformly from [1.0,2.0] and [0.1,1.0], respectively. Becauseh̸= 0 breaks integrability, all sampled Hamil- tonians are in the chaotic regime. Throughout training, the system size is fixed to beL= 8 and the sequence length to beT= 30. We first evaluate performanc...

  2. [2]

    Srednicki, Chaos and quantum thermalization, Phys

    M. Srednicki, Chaos and quantum thermalization, Phys. Rev. E50, 888 (1994)

    1994

  3. [3]

    J. M. Deutsch, Quantum statistical mechanics in a closed system, Phys. Rev. A43, 2046 (1991)

  4. [4]

    Rigol, V

    M. Rigol, V. Dunjko, and M. Olshanii, Thermalization and its mechanism for generic isolated quantum systems, Nature452, 854 (2008)

    2008

  5. [5]

    D’Alessio, Y

    L. D’Alessio, Y. Kafri, A. Polkovnikov, and M. Rigol, From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics, Advances in Physics65, 239 (2016), arXiv:1509.06411 [cond-mat.stat- mech]. 10

    arXiv 2016

  6. [6]

    Gogolin and J

    C. Gogolin and J. Eisert, Equilibration, thermalisation, and the emergence of statistical mechanics in closed quantum systems, Reports on Progress in Physics79, 056001 (2016), arXiv:1503.07538 [quant-ph]

    arXiv 2016

  7. [7]

    Eisert, M

    J. Eisert, M. Friesdorf, and C. Gogolin, Quantum many- body systems out of equilibrium, Nature Physics11, 124 (2015), arXiv:1408.5148 [quant-ph]

    arXiv 2015

  8. [8]

    Polkovnikov, K

    A. Polkovnikov, K. Sengupta, A. Silva, and M. Vengalat- tore, Colloquium: Nonequilibrium dynamics of closed in- teracting quantum systems, Reviews of Modern Physics 83, 863 (2011), arXiv:1007.5331 [cond-mat.stat-mech]

    arXiv 2011

Show all 79 references
  1. [9]

    J. M. Deutsch, Eigenstate thermalization hypothesis, Re- ports on Progress in Physics81, 082001 (2018)

    2018

  2. [10]

    Maldacena, S

    J. Maldacena, S. H. Shenker, and D. Stanford, A bound on chaos, Journal of High Energy Physics2016, 1 (2016)

    2016

  3. [11]

    A. I. Larkin and Y. N. Ovchinnikov, Quasiclassical method in the theory of superconductivity, Soviet Physics JETP28, 1200 (1969)

    1969

  4. [12]

    Hayden and J

    P. Hayden and J. Preskill, Black holes as mirrors: quan- tum information in random subsystems, Journal of High Energy Physics2007, 120 (2007), arXiv:0708.4025 [hep- th]

    2007 arXiv

  5. [13]

    Sekino and L

    Y. Sekino and L. Susskind, Fast scramblers, Journal of High Energy Physics2008, 065 (2008), arXiv:0808.2096 [hep-th]

    2008 arXiv

  6. [14]

    Hosur, X.-L

    P. Hosur, X.-L. Qi, D. A. Roberts, and B. Yoshida, Chaos in quantum channels, Journal of High Energy Physics 2016, 004 (2016), arXiv:1511.04021 [hep-th]

    2016 arXiv

  7. [15]

    Swingle, Unscrambling the physics of out-of-time- order correlators, Nature Physics14, 988 (2018)

    B. Swingle, Unscrambling the physics of out-of-time- order correlators, Nature Physics14, 988 (2018)

    2018

  8. [16]

    D. E. Parker, X. Cao, A. Avdoshkin, T. Scaffidi, and E. Altman, A universal operator growth hypothesis, Phys. Rev. X9, 041017 (2019)

    2019

  9. [17]

    Caputa, J

    P. Caputa, J. M. Magan, and D. Patramanis, Geome- try of Krylov complexity, Physical Review Research4, 013041 (2022), arXiv:2109.03824 [hep-th]

    2022 arXiv

  10. [18]

    Nandy, A

    P. Nandy, A. S. Matsoukas-Roubeas, P. Mart ´ ınez- Azcona, A. Dymarsky, and A. del Campo, Quan- tum dynamics in Krylov space: Methods and applica- tions, Physics Reports1125, 1 (2025), arXiv:2405.09628 [quant-ph]

    2025 arXiv

  11. [19]

    J. L. F. Barb´ on, E. Rabinovici, R. Shir, and S. S. K. Ali, On the evolution of operator complexity in quantum field theory, Journal of High Energy Physics2019, 1 (2019)

    2019

  12. [20]

    Bhattacharya, A

    A. Bhattacharya, A. Bhattacharyya, P. Hunt, and N. Kundu, Krylov complexity in open quantum systems, Journal of High Energy Physics2022, 1 (2022)

    2022

  13. [21]

    Bhattacharjee, X

    B. Bhattacharjee, X. Cao, P. Nandy, and T. Pathak, Op- erator growth in open quantum systems: lessons from the dissipative syk, Journal of High Energy Physics2023, 10.1007/jhep03(2023)054 (2023)

    2023 doi

  14. [22]

    Bhattacharya, P

    A. Bhattacharya, P. Nandy, P. P. Nath, and H. Sahu, On krylov complexity in open systems: an approach via bi- lanczos algorithm, Journal of High Energy Physics2023, 10.1007/jhep12(2023)066 (2023)

    2023 doi

  15. [23]

    Bhattacharya, P

    A. Bhattacharya, P. P. Nath, and H. Sahu, Speed lim- its to the growth of krylov complexity in open quantum systems, Phys. Rev. D109, L121902 (2024)

    2024

  16. [24]

    N. S. Srivatsa and C. von Keyserlingk, Operator growth hypothesis in open quantum systems, Phys. Rev. B109, 125149 (2024)

    2024

  17. [25]

    D. J. Yates, A. G. Abanov, and A. Mitra, Lifetime of al- most strong edge-mode operators in one-dimensional, in- teracting, symmetry protected topological phases, Phys. Rev. Lett.124, 206803 (2020)

    2020

  18. [26]

    Z. Qi, T. Scaffidi, and X. Cao, Surprises in the deep hilbert space of all-to-all systems: From superexponen- tial scrambling to slow entanglement growth, Phys. Rev. B108, 054301 (2023)

    2023

  19. [27]

    Cao, A statistical mechanism for operator growth, Journal of Physics A: Mathematical and Theoretical54, 144001 (2021)

    X. Cao, A statistical mechanism for operator growth, Journal of Physics A: Mathematical and Theoretical54, 144001 (2021)

    2021

  20. [28]

    Yi-Thomas, B

    S. Yi-Thomas, B. Ware, J. D. Sau, and C. D. White, Comparing numerical methods for hydrodynamics in a one-dimensional lattice spin model, Phys. Rev. B110, 134308 (2024)

    2024

  21. [29]

    Y. Zhou, W. Xia, L. Li, and W. Li, Diagnosing quantum many-body chaos in non-hermitian quantum spin chain via krylov complexity, Phys. Rev. Res.7, 033281 (2025)

    2025

  22. [30]

    Bartsch, A

    C. Bartsch, A. Dymarsky, M. H. Lamann, J. Wang, R. Steinigeweg, and J. Gemmer, Estimation of equili- bration time scales from nested fraction approximations, Phys. Rev. E110, 024126 (2024)

    2024

  23. [31]

    Teretenkov, F

    A. Teretenkov, F. Uskov, and O. Lychkovskiy, Pseu- domode expansion of many-body correlation functions, Phys. Rev. B111, 174308 (2025)

    2025

  24. [32]

    Pinna, O

    G. Pinna, O. Lunt, and C. von Keyserlingk, Approxi- mation theory for green’s functions via the lanczos algo- rithm, Phys. Rev. B112, 054435 (2025)

    2025

  25. [33]

    Ballar Trigueros and C.-J

    F. Ballar Trigueros and C.-J. Lin, Krylov complexity of many-body localization: Operator localization in krylov basis, SciPost Physics13, 10.21468/scipostphys.13.2.037 (2022)

    2022 doi

  26. [34]

    D. J. Yates, F. H. L. Essler, and A. Mitra, Almost strong (0, π) edge modes in clean interacting one-dimensional floquet systems, Phys. Rev. B99, 205419 (2019)

    2019

  27. [35]

    Tausendpfund, A

    N. Tausendpfund, A. Mitra, and M. Rizzi, Almost strong zero modes at finite temperature, Phys. Rev. Res.7, 023245 (2025)

    2025

  28. [36]

    D. J. Yates and A. Mitra, Strong and almost strong modes of floquet spin chains in krylov subspaces, Phys. Rev. B104, 195121 (2021)

    2021

  29. [37]

    G. F. Scialchi, A. J. Roncaglia, and D. A. Wisniacki, Integrability-to-chaos transition through the krylov ap- proach for state evolution, Phys. Rev. E109, 054209 (2024)

    2024

  30. [38]

    Heveling, J

    R. Heveling, J. Wang, and J. Gemmer, Numerically prob- ing the universal operator growth hypothesis, Physical Review E106, 014152 (2022), arXiv:2203.00533 [cond- mat.stat-mech]

    2022 arXiv

  31. [39]

    Eckseler, M

    J. Eckseler, M. Pieper, and J. Schnack, Escaping the krylov space during the finite-precision lanczos algo- rithm, Physical Review E112, 10.1103/k94p-vls8 (2025)

    2025 doi

  32. [40]

    Rabinovici, A

    E. Rabinovici, A. S´ anchez-Garrido, R. Shir, and J. Son- ner, Operator complexity: a journey to the edge of krylov space, Journal of High Energy Physics2021, 10.1007/jhep06(2021)062 (2021)

    2021 doi

  33. [41]

    H. D. Simon, The lanczos algorithm with partial re- orthogonalization, Mathematics of computation42, 115 (1984)

    1984

  34. [42]

    Van der Veen and K

    H. Van der Veen and K. Vuik, Bi-lanczos with partial or- thogonalization, Computers & structures56, 605 (1995)

    1995

  35. [43]

    B. N. Parlett and D. S. Scott, The Lanczos algorithm with selective orthogonalization, Mathematics of Com- putation33, 217 (1979)

    1979

  36. [44]

    Capizzi, L

    L. Capizzi, L. Mazza, and S. Murciano, Universal prop- erties of the many-body lanczos algorithm at finite size (2025), arXiv:2507.17424 [quant-ph]. 11

    2025

  37. [45]

    O. Lunt, T. Kriecherbauer, K. T.-R. McLaughlin, and C. von Keyserlingk, Emergent random matrix universality in quantum operator dynamics (2025), arXiv:2504.18311 [quant-ph]

    2025

  38. [46]

    Uskov and O

    F. Uskov and O. Lychkovskiy, Quantum dynamics in one and two dimensions via the recursion method, Phys. Rev. B109, L140301 (2024)

    2024

  39. [47]

    Dymarsky and M

    A. Dymarsky and M. Smolkin, Krylov complexity in con- formal field theory, Phys. Rev. D104, L081702 (2021)

    2021

  40. [48]

    J. Wang, M. H. Lamann, R. Steinigeweg, and J. Gemmer, Diffusion constants from the recursion method, Physical Review B110, 10.1103/physrevb.110.104413 (2024)

    2024 doi

  41. [49]

    L. E. Herrera Rodr ´ ıguez and A. A. Kananenka, Convolu- tional neural networks for long time dissipative quantum dynamics, The Journal of Physical Chemistry Letters12, 2476 (2021)

    2021

  42. [50]

    L. E. Herrera Rodr ´ ıguez and A. A. Kananenka, A short trajectory is all you need: A transformer-based model for long-time dissipative quantum dynamics, The Journal of Chemical Physics161, 171101 (2024)

    2024

  43. [51]

    Kaneko, M

    R. Kaneko, M. Imada, Y. Kabashima, and T. Ohtsuki, Forecasting long-time dynamics in quantum many-body systems by dynamic mode decomposition, Physical Re- view Research7, 013085 (2025)

    2025

  44. [52]

    Pescia, J

    G. Pescia, J. Nys, J. Kim, A. Lovato, and G. Carleo, Message-passing neural quantum states for the homoge- neous electron gas, Phys. Rev. B110, 035108 (2024)

    2024

  45. [53]

    Denis and G

    Z. Denis and G. Carleo, Accurate neural quantum states for interacting lattice bosons, Quantum9, 1772 (2025)

    2025

  46. [54]

    Rende, L

    R. Rende, L. L. Viteritti, F. Becca, A. Scardicchio, A. Laio, and G. Carleo, Foundation neural-networks quantum states as a unified ansatz for multiple hamil- tonians, Nature Communications16, 7213 (2025)

    2025

  47. [55]

    Z. Qi, Y. Peng, and C. Earls, Fourier neural operators for time-periodic quantum systems: Learning floquet hamiltonians, observable dynamics, and operator growth (2025), arXiv:2509.07084 [quant-ph]

    2025 arXiv

  48. [56]

    Carleo and M

    G. Carleo and M. Troyer, Solving the quantum many- body problem with artificial neural networks, Science 355, 602 (2017)

    2017

  49. [57]

    Carrasquilla and R

    J. Carrasquilla and R. G. Melko, Machine learning phases of matter, Nature Physics13, 431 (2017)

    2017

  50. [58]

    D. Luo, Z. Chen, K. Hu, Z. Zhao, V. M. Hur, and B. K. Clark, Gauge-invariant and anyonic-symmetric au- toregressive neural network for quantum lattice models, Phys. Rev. Res.5, 013216 (2023)

    2023

  51. [59]

    Viteritti, R

    L. Viteritti, R. Rende, and S. Goldt, Transformer quan- tum states: A new class of expressive ansatzes for quan- tum many-body systems, Physical Review Letters130, 096401 (2023)

    2023

  52. [60]

    Zhang and M

    Y.-H. Zhang and M. Di Ventra, Transformer quantum state: A multipurpose model for quantum many-body problems, Phys. Rev. B107, 075147 (2023)

    2023

  53. [61]

    P. Cha, P. Ginsparg, F. Wu, J. Carrasquilla, P. L. McMa- hon, and E.-A. Kim, Attention-based quantum tomog- raphy, Machine Learning: Science and Technology3, 01LT01 (2021)

    2021

  54. [62]

    Carleo, I

    G. Carleo, I. Cirac, K. Cranmer, L. Daudet, M. Schuld, N. Tishby, L. Vogt-Maranto, and L. Zdeborov´ a, Machine learning and the physical sciences, Reviews of Modern Physics91, 045002 (2019)

    2019

  55. [63]

    Carrasquilla, Machine learning for quantum matter, Advances in Physics: X5, 1797528 (2020)

    J. Carrasquilla, Machine learning for quantum matter, Advances in Physics: X5, 1797528 (2020)

    2020

  56. [64]

    H. Kim, A. Kumar, Y. Zhou, Y. Xu, R. Vasseur, and E.-A. Kim, Learning measurement-induced phase tran- sitions using attention (2025), arXiv:2508.15895 [quant- ph]

    2025 arXiv

  57. [65]

    Y. Zhou, C. Wan, Y. Xu, J. P. Zhou, K. Q. Weinberger, and E.-A. Kim, Learning to decode logical circuits, Na- ture Computational Science5, 1158 (2025)

    2025

  58. [66]

    H. Kim, Y. Zhou, Y. Xu, K. Varma, A. H. Karamlou, I. T. Rosen, J. C. Hoke, C. Wan, J. P. Zhou, W. D. Oliver, Y. D. Lensky, K. Q. Weinberger, and E.-A. Kim, Attention to quantum complexity, Science Advances11, eadu0059 (2025)

    2025

  59. [67]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, At- tention is all you need, Advances in Neural Information Processing Systems30(2017)

    2017

  60. [68]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, BERT: Pre-training of deep bidirectional trans- formers for language understanding, arXiv preprint arXiv:1810.04805 (2018)

    2018 arXiv

  61. [69]

    H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, and W. Zhang, Informer: Beyond efficient transformer for long sequence time-series forecasting, inProceedings of the AAAI Conference on Artificial Intelligence, Vol. 35 (2021) pp. 11106–11115

    2021

  62. [70]

    Liu and G

    J.-M. Liu and G. M¨ uller, Infinite-temperature dynamics of the equivalent-neighbor xyz model, Phys. Rev. A42, 5854 (1990)

    1990

  63. [71]

    Pfeuty, The transverse-field ising model, Annals of Physics57, 79 (1970)

    P. Pfeuty, The transverse-field ising model, Annals of Physics57, 79 (1970)

    1970

  64. [72]

    Roca-Jerat, M

    S. Roca-Jerat, M. Gallego, F. Luis, J. Carrete, and D. Zueco, Transformer wave function for quantum long- range models, Physical Review B110, 10.1103/phys- revb.110.205147 (2024)

    2024 doi

  65. [73]

    Hashimoto, K

    K. Hashimoto, K. Murata, N. Tanahashi, and R. Watan- abe, Krylov complexity and chaos in quantum me- chanics, Journal of High Energy Physics2023, 10.1007/jhep11(2023)040 (2023)

    2023 doi

  66. [74]

    B. L. Espa˜ nol and D. A. Wisniacki, Assessing the satu- ration of krylov complexity as a measure of chaos, Phys. Rev. E107, 024217 (2023)

    2023

  67. [75]

    J. D. Noh, Operator growth in the transverse-field ising spin chain with integrability-breaking longitudinal field, Phys. Rev. E104, 034112 (2021)

    2021

  68. [76]

    Y. Xian, B. Schiele, and Z. Akata, Zero-shot learning - the good, the bad and the ugly, inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)(2017)

    2017

  69. [77]

    D. P. Jatkar, S. Mahato, S. Mondkar, and P. Thalore, Violation of universal operator growth hypothesis inw 3 conformal field theories (2025), arXiv:2506.01957 [hep- th]

    2025 arXiv

  70. [78]

    Sprague and S

    K. Sprague and S. Czischek, Variational monte carlo with large patched transformers, Communications Physics7, 10.1038/s42005-024-01584-y (2024)

    2024 doi

  71. [79]

    Loshchilov and F

    I. Loshchilov and F. Hutter, Decoupled weight decay reg- ularization (2019), arXiv:1711.05101 [cs.LG]. 12 Appendix A: Lanczos Algorithm for Classical Chaotic Systems In Sec. II, we review the Lanczos algorithm for quantum operators acting on finite-dimensional Hilbert spaces. ...

    2019 arXiv

Pith tools

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