pith. sign in

arxiv: 2511.04271 · v1 · submitted 2025-11-06 · 🪐 quant-ph

Quantum time-marching algorithms for solving linear transport problems including boundary conditions

Sergio Bengoechea , Paul Over , Thomas Rung This is my paper

Pith reviewed 2026-05-18 01:11 UTC · model grok-4.3

classification 🪐 quant-ph
keywords quantum time-marchinglinear transportboundary conditionsmethod of imageslinear combination of unitariesheat equationquantum simulationsuccess probability
0
0 comments X

The pith

A quantum time-marching algorithm simulates multidimensional linear transport with arbitrary boundaries at linear time cost using one extra qubit per dimension.

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

The paper establishes a complete quantum time-marching method for linear transport problems that include arbitrary boundaries in any number of dimensions. It adapts the linear combination of unitaries approach to block-encode the dynamics and applies the method of images to enforce boundaries. Success probabilities remain high and intrinsic to the specific problem rather than falling with system size. Linear time complexity is preserved throughout, which supports running the routines on fault-tolerant quantum hardware. State-vector simulations of the heat equation verify the approach for Neumann, Dirichlet, and mixed boundary conditions.

Core claim

This article presents the first complete application of a quantum time-marching algorithm for simulating multidimensional linear transport phenomena with arbitrary boundaries, whereby the success probabilities are problem intrinsic. The method adapts the linear combination of unitaries algorithm to block encode the diffusive dynamics, while arbitrary boundary conditions are enforced by the method of images only at the cost of one additional qubit per spatial dimension. As an alternative to the non-periodic reflection, the direct encoding of Neumann conditions by the unitary decomposition of the discrete time-marching operator is proposed. All presented algorithms indicate optimal successprob

What carries the argument

Adaptation of linear combination of unitaries to block-encode diffusive dynamics together with the method of images for enforcing boundary conditions.

Load-bearing premise

Arbitrary boundary conditions can be enforced by reflecting the problem across the boundaries while adding only one qubit per spatial dimension and without lowering success probabilities or losing linear time scaling.

What would settle it

A state-vector run on the heat equation with mixed boundaries where the measured probability distribution after several time steps deviates from the known classical solution by more than the expected discretization error.

Figures

Figures reproduced from arXiv: 2511.04271 by Paul Over, Sergio Bengoechea, Thomas Rung.

Figure 1
Figure 1. Figure 1: FIG. 1: Three-qubit example for the implementation of [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2: Generic circuit for the temporal evolution of [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3: Exemplary quantum circuit implementations of ˆ [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Quantum circuit to implement one explicit time [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5: LCU circuit for the two-dimensional periodic heat equation. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6: Quantum circuit to implement one explicit time step of the one-dimensional non-periodic PDE defined by [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7: Quantum circuit to implement one explicit time step of the two-dimensional non-periodic PDE defined by [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8: Quantum circuit to implement one explicit time step of the two-dimensional non-periodic PDE defined by [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9: Three-qubit quantum circuits implementing [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10: Comparison of QM and FD results of [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗
Figure 7
Figure 7. Figure 7: Analogous to Sec. V A, the unitaries in Uc build- [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11: State-vector simulation of the two-dimensional heat equation (Eqn. (1)) with the homogeneous Neumann [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12: Benchmarking of QM results for the two-dimensional heat equation (Eqn. (1)) with homogeneous Dirichlet [PITH_FULL_IMAGE:figures/full_fig_p012_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13: QM results for the two-dimensional heat equation (Eqn. (1)) with homogeneous mixed boundary [PITH_FULL_IMAGE:figures/full_fig_p013_13.png] view at source ↗
read the original abstract

This article presents the first complete application of a quantum time-marching algorithm for simulating multidimensional linear transport phenomena with arbitrary boundaries, whereby the success probabilities are problem intrinsic. The method adapts the linear combination of unitaries algorithm to block encode the diffusive dynamics, while arbitrary boundary conditions are enforced by the method of images only at the cost of one additional qubit per spatial dimension. As an alternative to the non-periodic reflection, the direct encoding of Neumann conditions by the unitary decomposition of the discrete time-marching operator is proposed. All presented algorithms indicate optimal success probabilities while maintaining linear time complexity, thereby securing the practical applicability of the quantum algorithm on fault-tolerant quantum computers. The proposed time-marching method is demonstrated through state-vector simulations of the heat equation in combination with Neumann, Dirichlet, and mixed boundary conditions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript presents quantum time-marching algorithms for simulating multidimensional linear transport phenomena with arbitrary boundary conditions. It adapts the linear combination of unitaries (LCU) method to block-encode the diffusive dynamics and enforces boundary conditions either via the method of images (adding one qubit per spatial dimension) or by direct unitary decomposition of the discrete time-marching operator for Neumann conditions. The central claims are that these constructions achieve optimal success probabilities that depend only on intrinsic problem parameters and maintain linear time complexity, with validation through state-vector simulations of the heat equation under Neumann, Dirichlet, and mixed boundary conditions.

Significance. If the claims on success probabilities and complexity are rigorously established, the work would constitute a meaningful advance in quantum algorithms for PDEs by addressing non-periodic boundaries with only linear qubit overhead. The combination of LCU block-encoding with a quantum-adapted method of images could enable practical fault-tolerant simulations of transport problems where boundary handling has previously introduced prohibitive costs.

major comments (2)
  1. [§4.2, Eq. (17)] §4.2, Eq. (17): The LCU subnormalization factor for the reflected operators is stated to be identical to the periodic case, yet the construction incorporates 2^d image copies with sign flips; this appears to alter the effective operator norm, which would make the post-selection probability depend on dimension and boundary type rather than remaining strictly problem-intrinsic.
  2. [§5.3, Table 1] §5.3, Table 1: The reported success probabilities for the mixed-boundary heat-equation simulations match the periodic-case formula but do not isolate or tabulate the normalization shift induced by the method-of-images block encoding; without this separation it is impossible to confirm that optimality is preserved for general linear transport.
minor comments (2)
  1. [Figure 4] Figure 4: The success-probability curves for different boundary conditions are plotted on different scales, making direct visual comparison of dimension dependence difficult.
  2. [§2.1] §2.1: The discussion of prior LCU applications to transport equations omits the 2023 work on quantum linear systems for advection-diffusion (reference should be added for completeness).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading of our manuscript and the constructive comments. We address each major comment below. We have revised the manuscript to clarify the normalization analysis and to provide explicit separation of factors in the numerical results.

read point-by-point responses

  1. Referee: [§4.2, Eq. (17)] The LCU subnormalization factor for the reflected operators is stated to be identical to the periodic case, yet the construction incorporates 2^d image copies with sign flips; this appears to alter the effective operator norm, which would make the post-selection probability depend on dimension and boundary type rather than remaining strictly problem-intrinsic.

    Authors: We appreciate the referee highlighting this point. In the method-of-images construction, the 2^d reflected copies with sign flips are incorporated via a unitary extension of the diffusion operator to the doubled (or 2^d-extended) domain. The LCU coefficients are selected so that the subnormalization factor equals the spectral norm of the target block-encoded operator. Because the image reflections are isometries that preserve the norm of the original diffusive operator (the extended operator on the larger domain has the same maximum eigenvalue magnitude as the periodic case), the sum of absolute coefficients remains identical to the periodic construction. Consequently the post-selection probability depends only on the intrinsic parameters (time-step size and diffusion coefficient) and is independent of spatial dimension d and boundary type. We have added an explicit norm calculation and a short remark after Eq. (17) in the revised manuscript to make this invariance transparent. revision: yes

  2. Referee: [§5.3, Table 1] The reported success probabilities for the mixed-boundary heat-equation simulations match the periodic-case formula but do not isolate or tabulate the normalization shift induced by the method-of-images block encoding; without this separation it is impossible to confirm that optimality is preserved for general linear transport.

    Authors: We agree that an explicit separation would strengthen the presentation. The success probabilities reported in Table 1 were obtained directly from the state-vector simulations and coincide with the periodic-case formula, indicating that the method-of-images encoding introduces no additional subnormalization. To address the referee’s request, we have revised Table 1 to include a supplementary column that lists the measured subnormalization factor for each boundary-condition variant (periodic, Neumann, Dirichlet, mixed) alongside the theoretical value. This confirms that the factor remains unchanged across all cases and that optimality is preserved for general linear transport problems. revision: yes

Circularity Check

0 steps flagged

No significant circularity; new application of LCU and method of images

full rationale

The paper applies established LCU block-encoding for the time-marching operator and the method of images (or direct unitary decomposition) for boundary conditions. These are presented as extensions that preserve problem-intrinsic success probabilities and linear complexity, without any quoted reduction of a prediction to a fitted parameter defined by the output itself or a self-citation chain that bears the central load. The claim of being the 'first complete application' is a novelty statement rather than a definitional loop. State-vector simulations are offered as demonstration rather than as the source of the optimality result. This is the normal case of an independent algorithmic construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Review limited to abstract; no explicit free parameters, new entities, or non-standard axioms are stated. Relies on standard quantum computing primitives such as block encoding and unitary operators.

axioms (2)
  • domain assumption Linear combination of unitaries can be adapted to block-encode diffusive dynamics for time-marching.
    Implicit in the adaptation described for the core transport operator.
  • domain assumption Method of images extends to quantum setting with one qubit overhead per dimension.
    Central to the boundary-condition handling claim.

pith-pipeline@v0.9.0 · 5662 in / 1282 out tokens · 32641 ms · 2026-05-18T01:11:10.707834+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

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

  1. Resource Implications of Different Encodings for Quantum Computational Fluid Dynamics

    quant-ph 2026-04 unverdicted novelty 5.0

    Amplitude encoding in quantum CFD incurs circuit depths from Shende decomposition and requires an empirical ~n ln n runs for accurate readout, motivating a new encoding proposal for quantum LBM.

Reference graph

Works this paper leans on

57 extracted references · 57 canonical work pages · cited by 1 Pith paper · 2 internal anchors

  1. [1]

    , (4) using the unitaries identity I and the shift operator S0 =   0 1 0 · · · 0 0 1 ... ... ... ... 0 0 0 1 1 0 · · · 0   . (5) In Eqn. (4), S† 0 indicates the conjugate transpose of S0, and both of them can be efficiently implemented with the sequence of quantum gates depicted in Fig. 1. S0 X (a) Quantum circuit for S0. S† 0 X (b) Quantu...

    work page

  2. [2]

    To satisfy this condition, M is pre-scaled by a nor- malization factor α ∈ R+ such that α ≥ ∥ M ∥2. Con- sequently, for the non-unitary matrix M ∈ C2n×2n de- fined by n qubits, an accurate block encoding is achieved primarily by an optimal choice of the subnormalization factor, i.e., α = ∥M ∥2. For the sake of clarity, ˜M is as- sumed to be the optimal sc...

    work page

  3. [3]

    4 with κ0 = 1, κ1 = κ2 = rh/(1 − 2rh), and κ3 = 0

    in the LCU circuit of Fig. 4 with κ0 = 1, κ1 = κ2 = rh/(1 − 2rh), and κ3 = 0. Note that the ad- ditional constraint rh ≤ 1/(2 d) needs to be imposed to guarantee that all the LCU weights are κk ≥ 0. Here, the LCU scaling factor 1 /(κ0 + κ1 + κ2 + κ3) = 1 − 2rh cancels the initial subnormalization factor 1 /(1 − 2rh) in Eqn. (4) completely and results in t...

    work page

  4. [4]

    10 depict the QM solution of ϕϕϕt computed for Nt = 12000 explicit time steps with rh = 0 .2

    The results in Fig. 10 depict the QM solution of ϕϕϕt computed for Nt = 12000 explicit time steps with rh = 0 .2. The even (symmet- ric) reflection on the three additional quadrants II, III, and IV, without the register reset by the reflection part of the circuit in Fig. 7, is indicated in Fig. 10a, which corresponds to 1000 time steps. The full resolutio...

    work page

  5. [5]

    classical FD results

    5 1 ·10−12 t/T [− ] εt (c) Temporal evolution of the error measure εt w.r.t. classical FD results. Convergence to zero due to a vanishing steady solution. FIG. 13: QM results for the two-dimensional heat equation (Eqn. (1)) with homogeneous mixed boundary conditions enforced by the method of images. The horizontal direction corresponds to Neumann while th...

    work page doi:10.25592/uhhfdm.18048 2021

  6. [6]

    M. M. Aslam Bhutta, N. Hayat, M. H. Bashir, A. R. Khan, K. N. Ahmad, and S. Khan, Applied Thermal En- gineering 32, 1 (2012)

    work page 2012

  7. [7]

    Jameson, Journal of Scientific Computing 3, 233 (1988)

    A. Jameson, Journal of Scientific Computing 3, 233 (1988)

    work page 1988

  8. [8]

    Zhang, X

    Z. Zhang, X. Wang, and Y. Yan, e-Prime - Advances in Electrical Engineering, Electronics and Energy 1, 100009 (2021)

    work page 2021

  9. [9]

    G. K. Vallis, Atmospheric and Oceanic Fluid Dynam- ics: Fundamentals and Large-Scale Circulation , 2nd ed. (Cambridge University Press, 2017)

    work page 2017

  10. [10]

    H. N. Khan, D. A. Hounshell, and E. R. H. Fuchs, Nature Electronics 1, 14 (2018)

    work page 2018

  11. [11]

    Esmaeilzadeh, E

    H. Esmaeilzadeh, E. Blem, R. S. Amant, K. Sankar- alingam, and D. Burger, in 2011 38th Annual Inter- national Symposium on Computer Architecture (ISCA) (2011) pp. 365–376

    work page 2011

  12. [12]

    J. P. Slotnick, A. Khodadoust, J. Alonso, D. Darmofal, W. Gropp, E. Lurie, and D. J. Mavriplis, CFD vision 2030 study: a path to revolutionary computational aero- sciences, Tech. Rep. (NASA, 2014)

    work page 2030

  13. [13]

    Jaksch, P

    D. Jaksch, P. Givi, A. J. Daley, and T. Rung, AIAA Journal 61, 1885 (2023)

    work page 2023

  14. [14]

    P. Givi, A. J. Daley, D. Mavriplis, and M. Malik, AIAA Journal 58, 3715 (2020)

    work page 2020

  15. [15]

    M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information (Cambridge University Press, 2010)

    work page 2010

  16. [16]

    Preskill, Quantum 2, 79 (2018)

    J. Preskill, Quantum 2, 79 (2018)

    work page 2018

  17. [17]

    Cerezo, A

    M. Cerezo, A. Arrasmith, R. Babbush, S. C. Benjamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio, and P. J. Coles, Nature Reviews Physics 3, 2522 (2021)

    work page 2021

  18. [18]

    Lubasch, J

    M. Lubasch, J. Joo, P. Moinier, M. Kiffner, and D. Jaksch, Physical Review A 101 (2020)

    work page 2020

  19. [19]

    P. Over, S. Bengoechea, T. Rung, F. Clerici, L. Scan- durra, E. de Villiers, and D. Jaksch, Computers & Fluids 288, 106508 (2025)

    work page 2025

  20. [20]

    Bengoechea, P

    S. Bengoechea, P. Over, D. Jaksch, and T. Rung, AIAA Journal (in-press) (2025)

    work page 2025

  21. [21]

    Y. Sato, R. Kondo, I. Hamamura, T. Onodera, and N. Yamamoto, Phys. Rev. Res. 6, 033246 (2024)

    work page 2024

  22. [22]

    Alipanah, F

    H. Alipanah, F. Zhang, Y. Yao, R. Thompson, N. Nguyen, J. Liu, P. Givi, B. J. McDermott, and J. J. Mendoza-Arenas, Phys. Rev. Res. , (2025)

    work page 2025

  23. [23]

    Brearley and S

    P. Brearley and S. Laizet, Phys. Rev. A 110, 012430 (2024)

    work page 2024

  24. [24]

    P. Over, S. Bengoechea, P. Brearley, S. Laizet, and T. Rung, Phys. Rev. A 112, L010401 (2025)

    work page 2025

  25. [25]

    An, J.-P

    D. An, J.-P. Liu, and L. Lin, Physical Review Letters 131, 150603 (2023)

    work page 2023

  26. [26]

    P. C. S. Costa, S. Jordan, and A. Ostrander, Phys. Rev. A 99, 012323 (2019)

    work page 2019

  27. [27]

    Esmaeilifar, D

    E. Esmaeilifar, D. Ahn, and R. S. Myong, Physics of Fluids 36, 106110 (2024)

    work page 2024

  28. [28]

    A. J. Pool, A. D. Somoza, C. Mc Keever, M. Lubasch, and B. Horstmann, Phys. Rev. Res. 6, 033257 (2024)

    work page 2024

  29. [29]

    Sanavio, R

    C. Sanavio, R. Scatamacchia, C. de Falco, and S. Succi, Physics of Fluids 36, 057143 (2024)

    work page 2024

  30. [30]

    S. S. Bharadwaj and K. R. Sreenivasan, Phys. Rev. Res. 7, 023262 (2025)

    work page 2025

  31. [31]

    D. Fang, L. Lin, and Y. Tong, Quantum 7, 955 (2023)

    work page 2023

  32. [32]

    Gily´ en, Y

    A. Gily´ en, Y. Su, G. H. Low, and N. Wiebe, in Pro- ceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing , STOC 2019 (Association for Computing Machinery, New York, NY, USA, 2019) p. 193–204

    work page 2019

  33. [33]

    Camps, L

    D. Camps, L. Lin, R. Van Beeumen, and C. Yang, SIAM Journal on Matrix Analysis and Applications 45, 801 (2024)

    work page 2024

  34. [34]

    S. Jin, N. Liu, and Y. Yu, Phys. Rev. A 108, 032603 (2023)

    work page 2023

  35. [35]

    S. Jin, X. Li, N. Liu, and Y. Yu, Journal of Computa- tional Physics 498, 112707 (2024)

    work page 2024

  36. [36]

    S. Jin, X. Li, N. Liu, and Y. Yu, SIAM Journal on Sci- entific Computing 46, B403 (2024)

    work page 2024

  37. [37]

    S. Jin, N. Liu, and Y. Yu, Phys. Rev. Lett. 133, 230602 (2024)

    work page 2024

  38. [38]

    Lu and Y

    Z. Lu and Y. Yang, Proceedings of the Combustion In- stitute 40, 105440 (2024)

    work page 2024

  39. [39]

    A. M. Childs, J.-P. Liu, and A. Ostrander, Quantum 5, 574 (2021)

    work page 2021

  40. [40]

    D. J. Griffiths and D. F. Schroeter, Introduction to Quan- tum Mechanics , 3rd ed. (Cambridge University Press, 2018)

    work page 2018

  41. [41]

    Ran, Phys

    S.-J. Ran, Phys. Rev. A 101, 032310 (2020)

    work page 2020

  42. [42]

    P.-F. Zhou, R. Hong, and S.-J. Ran, Phys. Rev. A 104, 042601 (2021)

    work page 2021

  43. [43]

    Termanova, A

    A. Termanova, A. Melnikov, E. Mamenchikov, N. Be- lokonev, S. Dolgov, A. Berezutskii, R. Ellerbrock, C. Mansell, and M. R. Perelshtein, New Journal of Physics 26, 123019 (2024)

    work page 2024

  44. [44]

    I. V. Oseledets, Constructive Approximation 37, 1 (2013)

    work page 2013

  45. [45]

    D. W. Berry, A. M. Childs, and R. Kothari, in 2015 IEEE 56th Annual Symposium on Foundations of Com- puter Science (2015) pp. 792–809

    work page 2015

  46. [46]

    A. M. Childs, D. Maslov, Y. Nam, N. J. Ross, and Y. Su, Proceedings of the National Academy of Sciences 115, 9456 (2018)

    work page 2018

  47. [47]

    Hatano and M

    N. Hatano and M. Suzuki, Finding exponential prod- uct formulas of higher orders, in ”Quantum Annealing and Other Optimization Methods (Springer Berlin Hei- 15 delberg, 2005) pp. 37–68

    work page 2005

  48. [48]

    D. W. Berry, A. M. Childs, R. Cleve, R. Kothari, and R. D. Somma, Phys. Rev. Lett. 114, 090502 (2015)

    work page 2015

  49. [49]

    G. H. Low and I. L. Chuang, Phys. Rev. Lett. 118, 010501 (2017)

    work page 2017

  50. [50]

    R. M. Gingrich and C. P. Williams, in Proceedings of the Winter International Synposium on Information and Communication Technologies, Vol. 58 (2004) pp. 1–6

    work page 2004

  51. [51]

    Lapworth and C

    L. Lapworth and C. S¨ underhauf, Quantum Science and Technology 10, 045064 (2025)

    work page 2025

  52. [52]

    S.-H. Lin, R. Dilip, A. G. Green, A. Smith, and F. Poll- mann, PRX Quantum 2, 010342 (2021)

    work page 2021

  53. [53]

    A. M. Childs and N. Wiebe, Quantum Information & Computation 12, 901 (2012)

    work page 2012

  54. [54]

    Hantzko, L

    L. Hantzko, L. Binkowski, and S. Gupta, Physica Scripta 99, 085128 (2024)

    work page 2024

  55. [55]

    Sano and I

    Y. Sano and I. Hamamura, Phys. Rev. Appl. 24, 034062 (2025)

    work page 2025

  56. [56]

    Quantum-Inspired Tensor-Network Fractional-Step Method for Incompressible Flow in Curvilinear Coordinates

    N.-L. van H¨ ulst, P. Siegl, P. Over, S. Ben- goechea, T. Hashizume, M. G. Cecile, T. Rung, and D. Jaksch, Quantum-inspired tensor-network fractional- step method for incompressible flow in curvilinear coor- dinates (2025), arXiv:2507.05222

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  57. [57]

    Tensor-Programmable Quantum Circuits for Solving Differential Equations

    P. Siegl, G. S. Reese, T. Hashizume, N.-L. van H¨ ulst, and D. Jaksch, Tensor-programmable quantum circuits for solving differential equations (2025), arXiv:2502.04425. Appendix A: LCU circuit to matrix As an example of state-vector simulation, the matrix- vector operations executed by the LCU circuit of Fig. 4 are illustrated here. Starting from the init...

    work page internal anchor Pith review Pith/arXiv arXiv 2025