Pith. sign in

REVIEW 2 major objections 1 minor 50 references

Monotone alternating splines overcome structural limits of monotone neural networks for cumulative conditional intensity functions.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

MAS is proposed as a new parameterization for CCIFs in TPPs using distinct interpolation and extrapolation spline components to overcome convexity, saturation, and modeling requirement issues of MNNs.

T0 review reviewed 2026-07-03 challenge →

load-bearing objection The paper hinges on whether MNNs really face three structural deadlocks for CCIFs that MAS then fixes with its alternating spline split. the 2 major comments →

arxiv 2607.01752 v1 pith:ABCO65UG submitted 2026-07-02 cs.LG

Efficient Temporal Point Processes via Monotone Alternating Splines

Cheng Wan , Quyu Kong , Feng Zhou This is my paper
classification cs.LG
keywords temporal point processesmonotone neural networkscumulative conditional intensity functionmonotone splinesinterpolationextrapolationevent modeling
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 establishes that parameterizing the cumulative conditional intensity function of temporal point processes with Monotone Alternating Splines yields more accurate and generalizable models than monotone neural networks. The authors argue that monotone neural networks suffer from convexity restrictions, saturation limits, and violations of CCIF modeling requirements that restrict their capacity for complex temporal dynamics. MAS addresses this by using separate interpolation and extrapolation components. Interpolation delivers strong fitting accuracy while extrapolation supports robust generalization, which together reduce the approximation gaps that remain with neural networks. This change also supports more efficient TPP modeling by working directly with the cumulative function rather than the intensity.

Core claim

Monotone Alternating Splines resolve the structural deadlocks of Monotone Neural Networks by using distinct interpolation and extrapolation components, providing strong fitting accuracy and robust generalization for CCIF modeling in temporal point processes.

What carries the argument

Monotone Alternating Splines (MAS), a framework that separates interpolation for fitting from extrapolation for generalization when modeling CCIFs.

Load-bearing premise

Monotone neural networks are limited by convexity restrictions, saturation limits, and violations of CCIF modeling requirements that cannot be overcome within the neural network approach.

What would settle it

An experiment in which a monotone neural network matches or exceeds MAS on fitting accuracy, generalization error, and downstream TPP likelihood on the same synthetic and real datasets would falsify the central claim.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • CCIF modeling becomes feasible for temporal dynamics that exceed the representational capacity of monotone neural networks.
  • Numerical approximation errors from intensity-based TPP modeling are eliminated.
  • Generalization improves because extrapolation is handled explicitly rather than through saturated network outputs.
  • Overall computational efficiency rises since the model operates directly on the cumulative function.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The separation of interpolation and extrapolation could apply to other tasks that require strictly monotone functions, such as certain survival or reliability models.
  • Long-sequence or streaming event data might test whether the extrapolation component scales without retraining.
  • Hybrid architectures that embed MAS layers inside recurrent structures could handle marked point processes where event types also evolve over time.
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

2 major / 1 minor

Summary. The paper claims that Monotone Neural Networks (MNNs) used for parameterizing cumulative conditional intensity functions (CCIFs) in temporal point processes (TPPs) are limited by three structural deadlocks—convexity restrictions, saturation limits, and violations of CCIF modeling requirements—which restrict representational capacity. It introduces Monotone Alternating Splines (MAS) with separate interpolation and extrapolation components to overcome these, asserting that MAS achieves stronger fitting accuracy and generalization while reducing irreducible approximation gaps of MNNs, with superior performance on synthetic and real-world datasets.

Significance. If the identified limitations of MNNs prove structural and irreducible, and if MAS demonstrably closes the approximation gaps while preserving monotonicity and efficiency, the work could improve computational tractability for TPPs by enabling direct CCIF modeling without numerical integration. The interpolation/extrapolation split offers a concrete architectural idea for monotone function approximation that may generalize beyond this application.

major comments (2)
  1. [Abstract] Abstract: The assertion that MNNs suffer from three structural deadlocks (convexity restrictions, saturation limits, and CCIF modeling violations) that 'fundamentally restrict their representational capacity' is presented as motivation without any derivation, error analysis, or comparison showing these cannot be mitigated inside the MNN class (e.g., via different monotone activations, regularizers, or architectures). This premise is load-bearing for the need to introduce MAS.
  2. [Abstract] Abstract: The theoretical statement that MAS 'reduc[es] the irreducible approximation gaps of MNNs' via its interpolation/extrapolation split is asserted without a supporting theorem, approximation bound, or quantitative comparison of representational capacity between the two families.
minor comments (1)
  1. [Abstract] The abstract refers to 'extensive experiments' on synthetic and real-world datasets but supplies no information on dataset characteristics, baselines, metrics, or statistical significance; a brief summary of these should appear even in the abstract for a methods paper.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the motivation and theoretical claims. We address each point below and will revise the manuscript to provide additional support.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The assertion that MNNs suffer from three structural deadlocks (convexity restrictions, saturation limits, and CCIF modeling violations) that 'fundamentally restrict their representational capacity' is presented as motivation without any derivation, error analysis, or comparison showing these cannot be mitigated inside the MNN class (e.g., via different monotone activations, regularizers, or architectures). This premise is load-bearing for the need to introduce MAS.

    Authors: Sections 3.1-3.3 derive these limitations directly from MNN architecture: monotonicity is enforced via non-negative weights, which (combined with common convex activations) produces convex CCIFs that cannot capture non-convex temporal dynamics; saturation follows from bounded activations such as sigmoids or softplus; and CCIF modeling violations arise because a single network cannot separately optimize interpolation on observed intervals and extrapolation beyond them. These properties are structural to the MNN class rather than artifacts of specific choices, as altering activations or adding regularizers does not remove the underlying convexity or single-function compromise. We will add a concise derivation of these points to the introduction to strengthen the abstract's motivation. revision: yes

  2. Referee: [Abstract] Abstract: The theoretical statement that MAS 'reduc[es] the irreducible approximation gaps of MNNs' via its interpolation/extrapolation split is asserted without a supporting theorem, approximation bound, or quantitative comparison of representational capacity between the two families.

    Authors: The split enables the interpolation spline to achieve exact fitting (zero approximation error) on observed data points by construction, while the independent extrapolation component handles tail behavior without back-propagating compromises into the fit region; a monolithic MNN cannot decouple these regimes. This is supported by the design analysis in Section 4 and the empirical gap reductions shown in Section 5 across synthetic and real datasets. We agree a formal statement would strengthen the claim and will add a proposition bounding the approximation advantage of the split relative to MNNs under standard Lipschitz and monotonicity assumptions. revision: yes

Circularity Check

0 steps flagged

No circularity; derivation is an independent construction.

full rationale

The paper motivates MAS by identifying limitations in existing MNN-based CCIF models and then constructs a new spline-based framework with distinct interpolation and extrapolation components. No load-bearing step reduces a claimed result or prediction to a fitted parameter, self-citation chain, or definitional equivalence; the central claims rest on the explicit design of MAS rather than on renaming or re-deriving inputs. The identification of MNN deadlocks functions as external motivation rather than an internal loop that forces the outcome.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

Based solely on abstract; the central claim rests on the domain assumption that CCIF modeling is preferable and that MNNs have fundamental structural limits, with MAS introduced as a new construction without independent evidence of its properties.

axioms (2)
  • domain assumption Modeling the cumulative conditional intensity function improves computational efficiency and eliminates numerical approximation errors compared to direct intensity modeling.
    Stated directly in the abstract as the motivation for focusing on CCIF.
  • ad hoc to paper Monotone neural networks suffer from convexity restrictions, saturation limits, and violations of CCIF modeling requirements that fundamentally restrict representational capacity.
    Identified in the abstract as the core motivation for proposing MAS.
invented entities (1)
  • Monotone Alternating Splines (MAS) no independent evidence
    purpose: Flexible parameterization of CCIFs using distinct interpolation and extrapolation components.
    Introduced in the abstract as the novel framework to resolve MNN limitations.

reviewed 2026-07-03 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Temporal Point Processes via Monotone Alternating Splines." pith.science (2026). https://pith.science/paper/ABCO65UG

@misc{pith2026260701752,
  author       = {Pith},
  title        = {Pith review of: Efficient Temporal Point Processes via Monotone Alternating Splines},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ABCO65UG}},
  note         = {Machine review of arXiv:2607.01752}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Temporal point processes (TPPs) have widespread applications across various domains. Compared to modeling the conditional intensity of a TPP, modeling its cumulative conditional intensity function (CCIF) improves computational efficiency and eliminates numerical approximation errors. However, current CCIF parameterizations uniformly rely on Monotone Neural Networks (MNNs), which we identify as suffering from three structural deadlocks--convexity restrictions, saturation limits, and violations of CCIF modeling requirements--that fundamentally restrict their representational capacity for complex temporal dynamics. To resolve these bottlenecks, this paper proposes a novel framework called Monotone Alternating Splines (MAS). By leveraging distinct interpolation and extrapolation components, MAS provides a flexible and efficient framework for modeling CCIFs. Theoretically, MAS's interpolation provides strong fitting accuracy, while its extrapolation supports robust generalization, reducing the irreducible approximation gaps of MNNs. Extensive experiments show that MAS achieves superior performance on both synthetic and real-world datasets.

Figures

Figures reproduced from arXiv: 2607.01752 by Cheng Wan, Feng Zhou, Quyu Kong.

Figure 1
Figure 1. Figure 1: Performance comparison of MNNs and MAS on three classical TPPs [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Implementation of MAS interpolation. The history encoder maps past [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages · 1 internal anchor

  1. [1]

    Self-exciting point process modeling of crime,

    G. O. Mohler, M. B. Short, P. J. Brantingham, F. P. Schoenberg, and G. E. Tita, “Self-exciting point process modeling of crime,”Journal of the American Statistical Association, vol. 106, no. 493, pp. 100–108, 2011

    work page 2011

  2. [2]

    Efficient inference for nonparametric Hawkes processes using auxiliary latent variables,

    F. Zhou, Z. Li, X. Fan, Y . Wang, A. Sowmya, and F. Chen, “Efficient inference for nonparametric Hawkes processes using auxiliary latent variables,”Journal of Machine Learning Research, vol. 21, no. 241, pp. 1–31, 2020

    work page 2020

  3. [3]

    Marked self-exciting point process modelling of information diffusion on twitter,

    F. Chen and W. H. Tan, “Marked self-exciting point process modelling of information diffusion on twitter,” The Annals of Applied Statistics, vol. 12, no. 4, pp. 2175– 2196, 2018

    work page 2018

  4. [4]

    Interpretable transformer hawkes processes: Unveiling complex interactions in social networks,

    Z. Meng, K. Wan, Y . Huang, Z. Li, Y . Wang, and F. Zhou, “Interpretable transformer hawkes processes: Unveiling complex interactions in social networks,” inProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024, pp. 2200–2211

    work page 2024

  5. [5]

    Hawkes pro- cesses in finance,

    E. Bacry, I. Mastromatteo, and J.-F. Muzy, “Hawkes pro- cesses in finance,”Market Microstructure and Liquidity, vol. 1, no. 01, p. 1550005, 2015

    work page 2015

  6. [6]

    Hawkes processes and their applications to finance: a review,

    A. G. Hawkes, “Hawkes processes and their applications to finance: a review,”Quantitative Finance, vol. 18, no. 2, pp. 193–198, 2018

    work page 2018

  7. [7]

    Space-time point-process models for earth- quake occurrences,

    Y . Ogata, “Space-time point-process models for earth- quake occurrences,”Annals of the Institute of Statistical Mathematics, vol. 50, no. 2, pp. 379–402, 1998

    work page 1998

  8. [8]

    Seismicity analysis through point-process model- ing: A review,

    ——, “Seismicity analysis through point-process model- ing: A review,”Pure and applied geophysics, vol. 155, no. 2, pp. 471–507, 1999

    work page 1999

  9. [9]

    Bayesian methods for discovering structure in neural spike trains,

    S. W. Linderman, “Bayesian methods for discovering structure in neural spike trains,” Ph.D. dissertation, Har- vard University, 2016

    work page 2016

  10. [10]

    Efficient inference for dynamic flexible interactions of neural populations,

    F. Zhou, Q. Kong, Z. Deng, J. Kan, Y . Zhang, C. Feng, and J. Zhu, “Efficient inference for dynamic flexible interactions of neural populations,”Journal of Machine Learning Research, vol. 23, no. 211, pp. 1–49, 2022

    work page 2022

  11. [11]

    Recurrent marked temporal point processes: embedding event history to vector,

    N. Du, H. Dai, R. Trivedi, U. Upadhyay, M. Gomez- Rodriguez, and L. Song, “Recurrent marked temporal point processes: embedding event history to vector,” in International Conference on Knowledge Discovery and Data Mining, 2016

    work page 2016

  12. [12]

    The neural Hawkes process: A neurally self-modulating multivariate point process,

    H. Mei and J. Eisner, “The neural Hawkes process: A neurally self-modulating multivariate point process,” inAdvances in Neural Information Processing Systems, 2017

    work page 2017

  13. [13]

    Trans- former Hawkes process,

    S. Zuo, H. Jiang, Z. Li, T. Zhao, and H. Zha, “Trans- former Hawkes process,” inInternational Conference on Machine Learning, 2020

    work page 2020

  14. [14]

    Monte Carlo sampling methods using Markov chains and their applications,

    W. Hastings, “Monte Carlo sampling methods using Markov chains and their applications,”Biometrika, pp. 97–109, 1970

    work page 1970

  15. [15]

    Fully neural network based model for general temporal point processes,

    T. Omi, N. Ueda, and K. Aihara, “Fully neural network based model for general temporal point processes,” in Advances in Neural Information Processing Systems, 2019

    work page 2019

  16. [16]

    Automatic differentiation in machine learning: a survey,

    A. G. Baydin, B. A. Pearlmutter, A. A. Radul, and J. M. Siskind, “Automatic differentiation in machine learning: a survey,”Journal of machine learning research, vol. 18, 2018

    work page 2018

  17. [17]

    Cumulative hazard function based efficient mul- tivariate temporal point process learning,

    B. Liu, “Cumulative hazard function based efficient mul- tivariate temporal point process learning,”arXiv preprint PREPRINT 12 arXiv:2404.13663, 2024

    work page arXiv 2024

  18. [18]

    D. J. Daley and D. Vere-Jones,An introduction to the theory of point processes: volume II: general theory and structure. Springer Science & Business Media, 2007

    work page 2007

  19. [19]

    J. F. C. Kingman,Poisson processes. Clarendon Press, 1992, vol. 3

    work page 1992

  20. [20]

    Spectra of some self-exciting and mutu- ally exciting point processes,

    A. G. Hawkes, “Spectra of some self-exciting and mutu- ally exciting point processes,”Biometrika, vol. 58, no. 1, pp. 83–90, 1971

    work page 1971

  21. [21]

    A self-correcting point pro- cess,

    V . Isham and M. Westcott, “A self-correcting point pro- cess,”Stochastic processes and their applications, vol. 8, no. 3, pp. 335–347, 1979

    work page 1979

  22. [22]

    Self- attentive hawkes process,

    Q. Zhang, A. Lipani, O. Kirnap, and E. Yilmaz, “Self- attentive hawkes process,” inInternational conference on machine learning. PMLR, 2020, pp. 11 183–11 193

    work page 2020

  23. [23]

    Neural ordinary differential equations,

    R. T. Chen, Y . Rubanova, J. Bettencourt, and D. K. Duve- naud, “Neural ordinary differential equations,”Advances in neural information processing systems, vol. 31, 2018

    work page 2018

  24. [24]

    Latent ordinary differential equations for irregularly-sampled time series,

    Y . Rubanova, R. T. Chen, and D. K. Duvenaud, “Latent ordinary differential equations for irregularly-sampled time series,”Advances in Neural Information Processing Systems, 2019

    work page 2019

  25. [25]

    Neural jump-diffusion temporal point processes,

    S. Zhang, C. Zhou, Y . Liu, P. Zhang, X. Lin, and Z.-M. Ma, “Neural jump-diffusion temporal point processes,” inInternational Conference on Machine Learning, 2024

    work page 2024

  26. [26]

    Calculation of Gauss quadrature rules,

    G. H. Golub and J. H. Welsch, “Calculation of Gauss quadrature rules,”Mathematics of Computation, vol. 23, no. 106, pp. 221–230, 1969

    work page 1969

  27. [27]

    Fast and flexible temporal point processes with triangular maps,

    O. Shchur, N. Gao, M. Bilo ˇs, and S. G ¨unnemann, “Fast and flexible temporal point processes with triangular maps,”Advances in Neural Information Processing Sys- tems, vol. 33, pp. 73–84, 2020

    work page 2020

  28. [28]

    Smurf-thp: score matching-based uncertainty quantification for transformer hawkes process,

    Z. Li, Y . Xu, S. Zuo, H. Jiang, C. Zhang, T. Zhao, and H. Zha, “Smurf-thp: score matching-based uncertainty quantification for transformer hawkes process,” inIn- ternational Conference on Machine Learning. PMLR, 2023, pp. 20 210–20 220

    work page 2023

  29. [29]

    Is score matching suitable for estimating point processes?

    H. Cao, Z. Meng, T. Ke, and F. Zhou, “Is score matching suitable for estimating point processes?”Advances in Neural Information Processing Systems, vol. 37, pp. 23 824–23 851, 2025

    work page 2025

  30. [30]

    Intensity-free learning of temporal point processes,

    O. Shchur, M. Bilos, and S. G ¨unnemann, “Intensity-free learning of temporal point processes,” inInternational Conference on Learning Representations, 2020

    work page 2020

  31. [31]

    Decomposable transformer point processes,

    A. Panos, “Decomposable transformer point processes,” inAnnual Conference on Neural Information Processing Systems, 2024

    work page 2024

  32. [32]

    Cumulative distribution function based general temporal point processes,

    M. Wang, Y . Pan, Z. Xu, R. Guo, X. Zhao, W. Wang, Y . Wang, Z. Liu, and L. Liu, “Cumulative distribution function based general temporal point processes,”arXiv preprint arXiv:2402.00388, 2024

    work page arXiv 2024

  33. [33]

    Monotonic networks,

    J. Sill, “Monotonic networks,”Advances in neural infor- mation processing systems, vol. 10, 1997

    work page 1997

  34. [34]

    Monotone regression splines in action,

    J. O. Ramsay, “Monotone regression splines in action,” Statistical science, pp. 425–441, 1988

    work page 1988

  35. [35]

    Monotone piecewise cu- bic interpolation,

    F. N. Fritsch and R. E. Carlson, “Monotone piecewise cu- bic interpolation,”SIAM Journal on Numerical Analysis, vol. 17, no. 2, pp. 238–246, 1980

    work page 1980

  36. [36]

    Invertible generative modeling using linear rational splines,

    H. M. Dolatabadi, S. Erfani, and C. Leckie, “Invertible generative modeling using linear rational splines,” in International Conference on Artificial Intelligence and Statistics. PMLR, 2020, pp. 4236–4246

    work page 2020

  37. [37]

    Piecewise rational quadratic interpolation to monotonic data,

    J. A. Gregory and R. Delbourgo, “Piecewise rational quadratic interpolation to monotonic data,”IMA Journal of Numerical Analysis, vol. 2, no. 2, pp. 123–130, 1982

    work page 1982

  38. [38]

    Monotonicity- preserving c2 rational cubic spline for monotone data,

    M. Abbas, A. A. Majid, and J. M. Ali, “Monotonicity- preserving c2 rational cubic spline for monotone data,” Applied Mathematics and Computation, vol. 219, no. 6, pp. 2885–2895, 2012

    work page 2012

  39. [39]

    Neural spline flows,

    C. Durkan, A. Bekasov, I. Murray, and G. Papamakarios, “Neural spline flows,”Advances in neural information processing systems, vol. 32, 2019

    work page 2019

  40. [40]

    Hollander, D

    M. Hollander, D. A. Wolfe, and E. Chicken,Nonpara- metric statistical methods. John Wiley & Sons, 2013

    work page 2013

  41. [41]

    Dropout: a simple way to pre- vent neural networks from overfitting,

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: a simple way to pre- vent neural networks from overfitting,”The journal of machine learning research, vol. 15, no. 1, pp. 1929–1958, 2014

    work page 1929

  42. [42]

    Improving Generalization Performance by Switching from Adam to SGD

    N. S. Keskar and R. Socher, “Improving generalization performance by switching from adam to sgd,”arXiv preprint arXiv:1712.07628, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  43. [43]

    Hypro: A hybridly normalized probabilistic model for long-horizon predic- tion of event sequences,

    S. Xue, X. Shi, J. Zhang, and H. Mei, “Hypro: A hybridly normalized probabilistic model for long-horizon predic- tion of event sequences,”Advances in Neural Information Processing Systems, vol. 35, pp. 34 641–34 650, 2022

    work page 2022

  44. [44]

    Foiling nyc’s taxi trip data,

    C. Whong, “Foiling nyc’s taxi trip data,”FOILing NYCs Taxi Trip Data. Np, vol. 18, p. 14, 2014

    work page 2014

  45. [45]

    Easytpp: Towards open benchmarking temporal point processes,

    S. Xue, X. Shi, Z. Chu, Y . Wang, H. Hao, F. Zhou, C. JIANG, C. Pan, J. Y . Zhang, Q. Wenet al., “Easytpp: Towards open benchmarking temporal point processes,” inThe Twelfth International Conference on Learning Representations, 2023

    work page 2023

  46. [46]

    Learning triggering kernels for multi-dimensional hawkes processes,

    K. Zhou, H. Zha, and L. Song, “Learning triggering kernels for multi-dimensional hawkes processes,” inIn- ternational conference on machine learning. PMLR, 2013, pp. 1301–1309

    work page 2013

  47. [47]

    A. W. Van der Vaart,Asymptotic statistics. Cambridge university press, 2000, vol. 3

    work page 2000

  48. [48]

    Rademacher penalties and structural risk minimization,

    V . Koltchinskii, “Rademacher penalties and structural risk minimization,”IEEE Trans. Inf. Theory, vol. 47, pp. 1902–1914, 2001. PREPRINT 13 APPENDIXA NOTATIONS All notations used in the paper are listed in the Table V. TABLE V LIST OF NOTATIONS. Symbol Definition Symbol Definition am, bm,· · ·, e m parameters off m(t) C1,· · ·, C 4, Bgeneric constants δm...

    work page 1902

  49. [49]

    Cross Validation:To rigorously demonstrate that MAS indeed outperforms other baselines, we conduct a cross-validation study comparing MAS with a classical CCIF-based method, FullyNN. We evaluate FullyNN under different network depths, widths, activation functions, and training data sizes to ensure that MAS’s superiority is not merely due to accidental hyp...

    work page

  50. [50]

    sup f∈F 1 Z ZX z=1 σzf(x z) # ≤E σ

    Training Time:We align the parameter sizes of five models—THP, RMTPP, MAS, EMTPP, and FullyNN—and train them for 100 epochs on the Hawkes1 dataset. Experiments are rerun three times under the same conditions. We then measure the mean and the standard deviation of their running time (in minutes), and the results are summarized in Table VII. A visualized ve...

    work page

This paper was first reviewed by grok-4.3 on July 3, 2026.