pith. sign in

arxiv: 2602.11626 · v2 · pith:DB574VACnew · submitted 2026-02-12 · 💻 cs.LG · cs.AI· physics.chem-ph· physics.comp-ph· physics.flu-dyn

ArGEnT: Arbitrary Geometry-encoded Transformer for Operator Learning

Wenqian Chen , Yucheng Fu , Michael Penwarden , Pratanu Roy , Panos Stinis This is my paper

Pith reviewed 2026-05-16 02:27 UTC · model grok-4.3

classification 💻 cs.LG cs.AIphysics.chem-phphysics.comp-phphysics.flu-dyn
keywords operator learningtransformergeometry encodingDeepONetpoint cloudarbitrary domainsscientific machine learning
0
0 comments X

The pith

A transformer encodes arbitrary geometries from point clouds and serves as the trunk in DeepONet to learn operators that depend on both geometry and other inputs without explicit geometry parametrization in the branch.

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

The paper introduces ArGEnT, a geometry-aware transformer that processes point-cloud representations of arbitrary domains using self-attention, cross-attention, or hybrid variants. When integrated as the trunk network in DeepONet, this setup learns solution operators for physical systems where geometry varies, without feeding explicit geometry parameters into the branch network. The approach targets many-query tasks such as design optimization and control by enabling flexible evaluation at arbitrary locations. A sympathetic reader would care because it simplifies surrogate modeling for systems with complex, changing shapes across fluid dynamics, solid mechanics, and electrochemical applications while improving accuracy and generalization over standard DeepONet.

Core claim

ArGEnT employs transformer attention mechanisms to encode geometric information directly from point-cloud representations of arbitrary domains, with three variants that incorporate geometric features differently; when used as the trunk in DeepONet, it produces a surrogate that maps both geometric and non-geometric inputs to solutions without requiring explicit geometry parametrization as a branch input.

What carries the argument

ArGEnT, the Arbitrary Geometry-encoded Transformer, which applies self-attention, cross-attention, or hybrid attention directly to point-cloud geometry representations to serve as the trunk network in DeepONet.

Load-bearing premise

Point-cloud representations processed by transformer attention reliably capture the geometric features needed for accurate operator learning across arbitrary domains without additional explicit parametrization or signed-distance inputs.

What would settle it

A benchmark case with a new geometry far outside the training distribution where ArGEnT predictions show errors no better than standard DeepONet or other geometry-aware baselines.

Figures

Figures reproduced from arXiv: 2602.11626 by Michael Penwarden, Panos Stinis, Pratanu Roy, Wenqian Chen, Yucheng Fu.

Figure 1
Figure 1. Figure 1: Arbitrary Geometry-encoded Transformer (ArGEnT). (a) Two-layer self-attention trans [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: ArGEnT DeepONet architecture. The ArGEnT model functions as the trunk network, [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Laminar airfoil flow: (a) Geometry setup. (b, d) Inputs to the cross-attention ArGEnT. [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Laminar airfoil flow: contour plots of predicted flow fields (left panels) and predicted [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Laminar airfoil flow: effect of sampling strategies on evaluation accuracy of the [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Turbulent airfoil flow: (a) Geometry setup. (b, d) Inputs to the cross-attention trans [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Turbulent airfoil flow: contour plots of predicted flow fields (left panels) and predicted [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Turbulent airfoil flow: effect of sampling strategies on evaluation accuracy of the [PITH_FULL_IMAGE:figures/full_fig_p024_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Lid driven flow: (a) Parametrized geometry. (b) Geometry setup. (c, e) Inputs to the [PITH_FULL_IMAGE:figures/full_fig_p029_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Lid driven flow: contour plots of predicted flow fields (upper panels) and predicted [PITH_FULL_IMAGE:figures/full_fig_p030_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Lid driven flow: contour plots of predicted flow fields (upper panels), reference (middle [PITH_FULL_IMAGE:figures/full_fig_p031_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Redox flow battery: (a) Geometry setup and boundary conditions. (b, c) Inputs to [PITH_FULL_IMAGE:figures/full_fig_p033_12.png] view at source ↗
Figure 14
Figure 14. Figure 14: The cross-attention ArGEnT is still able to produce reasonable pre [PITH_FULL_IMAGE:figures/full_fig_p036_14.png] view at source ↗
Figure 13
Figure 13. Figure 13: Redox flow battery: contour plots of predicted fields (left panels) and predicted absolute [PITH_FULL_IMAGE:figures/full_fig_p038_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Redox flow battery: contour plots of predicted fields (left panels) and predicted absolute [PITH_FULL_IMAGE:figures/full_fig_p039_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Jet engine bracket: (a) Isometric, (b) top, and (c) side views of the bracket geometry. [PITH_FULL_IMAGE:figures/full_fig_p042_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Jet engine bracket: contour plots of predicted structural response fields, ground truth [PITH_FULL_IMAGE:figures/full_fig_p046_16.png] view at source ↗
read the original abstract

Learning solution operators for systems with complex, varying geometries and parametric physical settings is a central challenge in scientific machine learning. In many-query regimes such as design optimization, control and inverse problems, surrogate modeling must generalize across geometries while allowing flexible evaluation at arbitrary spatial locations. In this work, we propose Arbitrary Geometry-encoded Transformer (ArGEnT), a geometry-aware attention-based architecture for operator learning on arbitrary domains. ArGEnT employs Transformer attention mechanisms to encode geometric information directly from point-cloud representations with three variants-self-attention, cross-attention, and hybrid-attention-that incorporates different strategies for incorporating geometric features. By integrating ArGEnT into DeepONet as the trunk network, we develop a surrogate modeling framework capable of learning operator mappings that depend on both geometric and non-geometric inputs without the need to explicitly parametrize geometry as a branch network input. Evaluation on benchmark problems spanning fluid dynamics, solid mechanics and electrochemical systems, we demonstrate significantly improved prediction accuracy and generalization performance compared with the standard DeepONet and other existing geometry-aware saurrogates. In particular, the cross-attention transformer variant enables accurate geometry-conditioned predictions with reduced reliance on signed distance functions. By combining flexible geometry encoding with operator-learning capabilities, ArGEnT provides a scalable surrogate modeling framework for optimization, uncertainty quantification, and data-driven modeling of complex physical systems.

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

3 major / 2 minor

Summary. The paper proposes ArGEnT, a geometry-aware transformer architecture with self-, cross-, and hybrid-attention variants that encodes arbitrary geometries directly from point-cloud representations. Integrated as the trunk network in DeepONet, it learns solution operators depending on both geometric and non-geometric inputs without explicit geometry parametrization in the branch network. The authors claim significantly improved accuracy and generalization on benchmarks spanning fluid dynamics, solid mechanics, and electrochemical systems, with the cross-attention variant reducing reliance on signed-distance functions.

Significance. If the benchmark results prove robust under proper validation, ArGEnT would advance operator learning by providing a flexible point-cloud-based encoding for geometry-dependent PDE operators, enabling scalable surrogates for optimization, uncertainty quantification, and data-driven modeling without domain-specific parametrizations.

major comments (3)
  1. [Abstract] Abstract and evaluation section: the central claim of significantly improved prediction accuracy and generalization across arbitrary geometries rests entirely on benchmark results, yet no quantitative error values, error bars, training/validation data splits, or statistical significance tests are reported, preventing assessment of whether the gains are load-bearing or reproducible.
  2. [Evaluation on benchmarks] Evaluation on benchmarks: no ablation studies compare the three attention variants (self-, cross-, hybrid) or test the cross-attention variant with versus without signed-distance inputs, which is required to substantiate the claim that point-cloud attention alone captures the geometric features needed for accurate operator learning on unseen domains.
  3. [Methods] Methods section: the architecture description provides no explicit comparison of computational cost (e.g., FLOPs or wall-clock time) against standard DeepONet or other geometry-aware baselines, despite the quadratic scaling of transformer attention on point clouds, which directly affects the practicality of the claimed scalable framework for fine-resolution arbitrary geometries.
minor comments (2)
  1. [Abstract] Typo in Abstract: 'saurrogates' should read 'surrogates'.
  2. [Abstract] Abstract sentence structure: 'Evaluation on benchmark problems spanning fluid dynamics, solid mechanics and electrochemical systems, we demonstrate' is grammatically incomplete and should be revised for clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed review. We address each major comment point-by-point below and have revised the manuscript to incorporate the requested quantitative results, ablations, and computational analysis.

read point-by-point responses

  1. Referee: [Abstract] Abstract and evaluation section: the central claim of significantly improved prediction accuracy and generalization across arbitrary geometries rests entirely on benchmark results, yet no quantitative error values, error bars, training/validation data splits, or statistical significance tests are reported, preventing assessment of whether the gains are load-bearing or reproducible.

    Authors: We agree that explicit quantitative metrics are necessary to support the accuracy claims. The revised manuscript now includes a dedicated table in Section 4 reporting mean relative L2 errors with standard deviations across five random seeds, the precise 80/20 training/validation splits used for each benchmark, and paired t-test p-values confirming statistical significance of improvements over DeepONet and other baselines. These numbers are also summarized in the abstract. revision: yes

  2. Referee: [Evaluation on benchmarks] Evaluation on benchmarks: no ablation studies compare the three attention variants (self-, cross-, hybrid) or test the cross-attention variant with versus without signed-distance inputs, which is required to substantiate the claim that point-cloud attention alone captures the geometric features needed for accurate operator learning on unseen domains.

    Authors: The referee correctly identifies the absence of intra-variant ablations and the SDF ablation for cross-attention. We have added these experiments to the revised evaluation section (new subsection 4.3). Results demonstrate that the hybrid-attention variant yields the lowest errors, while the cross-attention variant retains strong generalization on unseen geometries even when SDF inputs are removed, directly supporting the point-cloud encoding claim. revision: yes

  3. Referee: [Methods] Methods section: the architecture description provides no explicit comparison of computational cost (e.g., FLOPs or wall-clock time) against standard DeepONet or other geometry-aware baselines, despite the quadratic scaling of transformer attention on point clouds, which directly affects the practicality of the claimed scalable framework for fine-resolution arbitrary geometries.

    Authors: We acknowledge that the quadratic scaling of attention warrants explicit cost analysis. The revised Methods section (new subsection 3.4) now reports FLOPs counts and wall-clock inference times for all ArGEnT variants versus DeepONet and geometry-aware baselines at the point-cloud resolutions used in the benchmarks. The overhead is quantified and discussed, with notes on sparse-attention approximations for finer resolutions. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper presents ArGEnT as an independent architectural choice (transformer attention variants on point clouds) integrated into DeepONet as trunk network. Claims of improved operator learning for geometry-dependent mappings rest on external benchmark evaluations across fluid, solid, and electrochemical problems rather than any internal reduction. No equations, predictions, or uniqueness results are shown that reduce by construction to fitted parameters, self-citations, or ansatzes within the provided text. The central modeling decision remains an external modeling choice evaluated on independent data.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no explicit free parameters, axioms, or invented entities; the architecture is presented as a modeling choice whose performance is measured on benchmarks.

pith-pipeline@v0.9.0 · 5567 in / 1233 out tokens · 66258 ms · 2026-05-16T02:27:17.255279+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

42 extracted references · 42 canonical work pages · 1 internal anchor

  1. [1]

    Benner, S

    P. Benner, S. Gugercin, K. Willcox, A survey of projection-based model reduc- tion methods for parametric dynamical systems, SIAM review 57 (4) (2015) 63 483–531

    work page 2015

  2. [2]

    Jameson, Aerodynamic shape optimization using the adjoint method, Lec- tures at the Von Karman Institute, Brussels 6 (2003)

    A. Jameson, Aerodynamic shape optimization using the adjoint method, Lec- tures at the Von Karman Institute, Brussels 6 (2003)

    work page 2003

  3. [3]

    Y. Sun, U. Sengupta, M. Juniper, Physics-informed deep learning for simultane- ous surrogate modeling and pde-constrained optimization of an airfoil geometry, Computer Methods in Applied Mechanics and Engineering 411 (2023) 116042

    work page 2023

  4. [4]

    J.Sokolowski, J.-P.Zolésio, Introductiontoshapeoptimization, in: Introduction to Shape Optimization: Shape Sensitivity Analysis, Springer, 1992, pp. 5–12

    work page 1992

  5. [5]

    Samadian, I

    D. Samadian, I. B. Muhit, N. Dawood, Application of data-driven surrogate models in structural engineering: a literature review, Archives of Computational Methods in Engineering 32 (2) (2025) 735–784

    work page 2025

  6. [6]

    J. Wang, H. Jiang, G. Chen, H. Wang, L. Lu, J. Liu, L. Xing, Integration of multi-physics and machine learning-based surrogate modelling approaches for multi-objective optimization of deformed gdl of pem fuel cells, Energy and AI 14 (2023) 100261

    work page 2023

  7. [7]

    H.-W. Li, L. Wang, J.-N. Liu, Y. Yang, G.-L. Lu, Maximizing power density in proton exchange membrane fuel cells: An integrated optimization framework coupling multi-physics structure models, machine learning, and improved gray wolf optimizer, Fuel 358 (2024) 130351

    work page 2024

  8. [8]

    J. S. Hesthaven, S. Ubbiali, Non-intrusive reduced order modeling of nonlinear problems using neural networks, Journal of Computational Physics 363 (2018) 55–78. 64

    work page 2018

  9. [9]

    Q. Wang, J. S. Hesthaven, D. Ray, Non-intrusive reduced order modeling of unsteady flows using artificial neural networks with application to a combustion problem, Journal of computational physics 384 (2019) 289–307

    work page 2019

  10. [10]

    Multilayer feedforward networks are universal approximators

    K. Hornik, M. Stinchcombe, H. White, Multilayer feedforward networks are universal approximators, Neural Networks 2 (5) (1989) 359–366. doi:https://doi.org/10.1016/0893-6080(89)90020-8. URLhttps://www.sciencedirect.com/science/article/pii/ 0893608089900208

    work page doi:10.1016/0893-6080(89)90020-8 1989

  11. [11]

    Zhang, C

    W. Zhang, C. Zhang, Y. Zhao, Z. Wang, Y. Liu, C. Zhou, Y. Hu, Convolu- tional neural networks-based surrogate model for fast computational fluid dy- namics simulations of indoor airflow distribution, Energy and Buildings 326 (2025) 115020

    work page 2025

  12. [12]

    Hua, C.-H

    Y. Hua, C.-H. Yu, Q. Zhao, M.-G. Li, W.-T. Wu, P. Wu, Surrogate modeling of heat transfers of nanofluids in absorbent tubes with fins based on deep con- volutional neural network, International Journal of Heat and Mass Transfer 202 (2023) 123736

    work page 2023

  13. [13]

    Kim, H.-r

    Y. Kim, H.-r. Kim, H. Jung, Prt-deeponet: Geometry-aware neural operator for efficient prediction of pore-scale concentration fields, Computers & Geosciences (2025) 106098

    work page 2025

  14. [14]

    L. Lu, P. Jin, G. Pang, Z. Zhang, G. E. Karniadakis, Learning nonlinear oper- ators via deeponet based on the universal approximation theorem of operators, Nature machine intelligence 3 (3) (2021) 218–229

    work page 2021

  15. [15]

    J. He, S. Koric, D. Abueidda, A. Najafi, I. Jasiuk, Geom-deeponet: A point- 65 cloud-based deep operator network for field predictions on 3d parameterized ge- ometries, Computer Methods in Applied Mechanics and Engineering 429 (2024) 117130

    work page 2024

  16. [16]

    Fusion-deeponet: A data-efficient neural operator for geometry-dependent hypersonic and supersonic flows

    A. Peyvan, V. Kumar, G. E. Karniadakis, Fusion-deeponet: A data-efficient neural operator for geometry-dependent hypersonic and supersonic flows, arXiv preprint arXiv:2501.01934 (2025)

    work page arXiv 2025

  17. [17]

    Shukla, V

    K. Shukla, V. Oommen, A. Peyvan, M. Penwarden, N. Plewacki, L. Bravo, A. Ghoshal, R. M. Kirby, G. E. Karniadakis, Deep neural operators as accu- rate surrogates for shape optimization, Engineering Applications of Artificial Intelligence 129 (2024) 107615

    work page 2024

  18. [18]

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, A.Anandkumar, Fourierneuraloperatorforparametricpartialdifferentialequa- tions, arXiv preprint arXiv:2010.08895 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  19. [19]

    Z.Li, D.Z.Huang, B.Liu, A.Anandkumar, Fourierneuraloperatorwithlearned deformations for pdes on general geometries, Journal of Machine Learning Re- search 24 (388) (2023) 1–26

    work page 2023

  20. [20]

    2806–2823

    B.Bonev, T.Kurth, C.Hundt, J.Pathak, M.Baust, K.Kashinath, A.Anandku- mar, Spherical fourier neural operators: Learning stable dynamics on the sphere, in: International conference on machine learning, PMLR, 2023, pp. 2806–2823

    work page 2023

  21. [21]

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, P. S. Yu, A comprehensive survey on graph neural networks, IEEE transactions on neural networks and learning systems 32 (1) (2020) 4–24. 66

    work page 2020

  22. [22]

    Horie, N

    M. Horie, N. Mitsume, Physics-embedded neural networks: Graph neural pde solvers with mixed boundary conditions, Advances in Neural Information Pro- cessing Systems 35 (2022) 23218–23229

    work page 2022

  23. [23]

    C. R. Qi, H. Su, K. Mo, L. J. Guibas, Pointnet: Deep learning on point sets for 3d classification and segmentation, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 652–660

    work page 2017

  24. [24]

    Kashefi, T

    A. Kashefi, T. Mukerji, Physics-informed pointnet: A deep learning solver for steady-state incompressible flows and thermal fields on multiple sets of irregular geometries, Journal of Computational Physics 468 (2022) 111510

    work page 2022

  25. [25]

    J. Park, N. Kang, Point-deeponet: Predicting nonlinear fields on non-parametric geometries under variable load conditions, Neural Networks (2026) 108560

    work page 2026

  26. [26]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural infor- mation processing systems 30 (2017)

    work page 2017

  27. [27]

    Z. Li, K. Meidani, A. B. Farimani, Transformer for partial differential equations’ operator learning, arXiv preprint arXiv:2205.13671 (2022)

    work page arXiv 2022

  28. [28]

    S. Wen, A. Kumbhat, L. Lingsch, S. Mousavi, Y. Zhao, P. Chandrashekar, S. Mishra, Geometry aware operator transformer as an efficient and accurate neural surrogate for pdes on arbitrary domains, arXiv preprint arXiv:2505.18781 (2025)

    work page arXiv 2025

  29. [29]

    Q. Liu, W. Zhong, H. Meidani, D. Abueidda, S. Koric, P. Geubelle, Geometry- informed neural operator transformer for partial differential equations on ar- 67 bitrary geometries, Computer Methods in Applied Mechanics and Engineering 451 (2026) 118668

    work page 2026

  30. [30]

    P. Jin, S. Meng, L. Lu, Mionet: Learning multiple-input operators via tensor product, SIAM Journal on Scientific Computing 44 (6) (2022) A3490–A3514

    work page 2022

  31. [31]

    Bonnet, J

    F. Bonnet, J. Mazari, P. Cinnella, P. Gallinari, Airfrans: High fidelity computa- tional fluid dynamics dataset for approximating reynolds-averaged navier–stokes solutions, Advances in Neural Information Processing Systems 35 (2022) 23463– 23478

    work page 2022

  32. [32]

    Anand, Finite element and finite volume methods for heat transfer and fluid dynamics, Cambridge University Press, 2022

    N. Anand, Finite element and finite volume methods for heat transfer and fluid dynamics, Cambridge University Press, 2022

    work page 2022

  33. [33]

    S. Niu, Y. Liu, J. Wang, H. Song, A decade survey of transfer learning (2010– 2020), IEEE Transactions on Artificial Intelligence 1 (2) (2021) 151–166

    work page 2010

  34. [34]

    Goswami, K

    S. Goswami, K. Kontolati, M. D. Shields, G. E. Karniadakis, Deep transfer op- erator learning for partial differential equations under conditional shift, Nature Machine Intelligence 4 (12) (2022) 1155–1164

    work page 2022

  35. [35]

    G. I. Parisi, R. Kemker, J. L. Part, C. Kanan, S. Wermter, Continual lifelong learning with neural networks: A review, Neural networks 113 (2019) 54–71

    work page 2019

  36. [36]

    Hadsell, D

    R. Hadsell, D. Rao, A. A. Rusu, R. Pascanu, Embracing change: Continual learning in deep neural networks, Trends in cognitive sciences 24 (12) (2020) 1028–1040

    work page 2020

  37. [37]

    Y. Fu, A. Howard, C. Zeng, Y. Chen, P. Gao, P. Stinis, Physics-guided continual learning for predicting emerging aqueous organic redox flow battery material performance, ACS Energy Letters 9 (6) (2024) 2767–2774. 68

    work page 2024

  38. [38]

    Hollas, X

    A. Hollas, X. Wei, V. Murugesan, Z. Nie, B. Li, D. Reed, J. Liu, V. Spren- kle, W. Wang, A biomimetic high-capacity phenazine-based anolyte for aqueous organic redox flow batteries, Nature Energy 3 (6) (2018) 508–514

    work page 2018

  39. [39]

    C. Zeng, S. Kim, Y. Chen, Y. Fu, J. Bao, Z. Xu, W. Wang, Characterization of electrochemical behavior for aqueous organic redox flow batteries, Journal of The Electrochemical Society 169 (12) (2022) 120527

    work page 2022

  40. [40]

    S. Hong, Y. Kwon, D. Shin, J. Park, N. Kang, Deepjeb: 3d deep learning-based syntheticjetenginebracketdataset, JournalofMechanicalDesign147(4)(2025) 041703

    work page 2025

  41. [41]

    Cao, Choose a transformer: Fourier or galerkin, Advances in neural informa- tion processing systems 34 (2021) 24924–24940

    S. Cao, Choose a transformer: Fourier or galerkin, Advances in neural informa- tion processing systems 34 (2021) 24924–24940

    work page 2021

  42. [42]

    J. Su, M. Ahmed, Y. Lu, S. Pan, W. Bo, Y. Liu, Roformer: Enhanced trans- former with rotary position embedding, Neurocomputing 568 (2024) 127063. 69

    work page 2024