pith. sign in

arxiv: 2605.07098 · v2 · pith:FBFLWD2Bnew · submitted 2026-05-08 · 💻 cs.LG · physics.comp-ph

CarCrashNet: A Large-Scale Dataset and Hierarchical Neural Solver for Data-Driven Structural Crash Simulation

Mohamed Elrefaie , Dule Shu , Matthew Klenk , Faez Ahmed This is my paper

Pith reviewed 2026-05-20 22:42 UTC · model grok-4.3

classification 💻 cs.LG physics.comp-ph
keywords crash simulationfinite element analysismachine learningvehicle safetystructural mechanicsneural networksdataset benchmarkdata-driven modeling
0
0 comments X

The pith

CarCrashNet supplies a validated open dataset of vehicle crashes and a neural solver that predicts full-vehicle responses better than prior geometric deep learning methods.

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

The paper presents CarCrashNet as a public benchmark combining more than 14,000 component-scale bumper-beam pole-impact simulations with 825 full-vehicle crash simulations drawn from three industry-standard models of increasing complexity. It first confirms that the underlying open-source finite-element workflow produces results consistent with physical experiments and with the commercial Ansys LS-DYNA solver. The authors then introduce CrashSolver, a hierarchical machine-learning model trained on these high-resolution finite-element data, and show through extensive comparisons that it outperforms both geometric deep learning and transformer-based neural solvers on full-vehicle crash prediction tasks. A sympathetic reader would care because reliable data-driven crash simulation could cut the cost and time of physical prototyping while supporting virtual safety certification.

Core claim

CarCrashNet is introduced as a multi-modal public benchmark that pairs component-scale and full-vehicle finite-element crash data with validated open-source simulation results, and CrashSolver is presented as the hierarchical neural model that delivers superior accuracy on full-vehicle crash prediction relative to state-of-the-art geometric deep learning and transformer-based alternatives when evaluated on the released datasets.

What carries the argument

CrashSolver, the hierarchical neural model trained directly on high-resolution finite-element crash data from the CarCrashNet benchmark to predict full-vehicle structural responses.

If this is right

  • The benchmark enables direct, reproducible comparison of new neural solvers on nonlinear contact, large-deformation, and material-failure problems.
  • CrashSolver demonstrates that data-driven models can be applied to full-vehicle meshes while preserving accuracy relative to conventional finite-element methods.
  • The released multi-modal dataset supports development of virtual crash-testing pipelines that reduce reliance on physical prototypes.

Where Pith is reading between the lines

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

  • If the benchmark remains reliable, similar hierarchical architectures could be tested on higher-resolution meshes or on crash scenarios outside the current training distribution.
  • Widespread adoption might allow vehicle manufacturers to explore larger design spaces early in development before committing to physical builds.
  • A natural next measurement would be to assess how well CrashSolver generalizes to impact velocities or angles absent from the 825 full-vehicle cases.

Load-bearing premise

The open-source OpenRadioss finite-element simulations produce results that match experimental crash data and commercial Ansys LS-DYNA outputs for the tested vehicle models and impact conditions.

What would settle it

Physical crash tests performed on one of the included vehicle models that show large discrepancies in structural deformation, force histories, or energy absorption compared with the corresponding OpenRadioss simulations stored in CarCrashNet.

Figures

Figures reproduced from arXiv: 2605.07098 by Dule Shu, Faez Ahmed, Matthew Klenk, Mohamed Elrefaie.

Figure 1
Figure 1. Figure 1: Overview of our CARCRASHNET framework. Left: the released datasets include a large-scale bumper-beam pole-impact dataset with more than 14k simulations and three full-vehicle crash datasets based on the Toyota Yaris sedan, Dodge Neon, and Chevrolet Silverado. Middle: the machine learning tasks considered in this work include full-field crash prediction using FIGConvUNet, Transolver, GeoTransolver, and our … view at source ↗
Figure 2
Figure 2. Figure 2: Representative low- and high-velocity crash cases for the three vehicle-scale models. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Structural regions edited by the vehicle-scale DoE. Highlighted front support and lower [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Overview of CrashSolver for full-vehicle crash prediction. Starting from the undeformed [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Scalar wall-force validation summary. NHTSA 5677 and 6221 values are digitized from [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison of post-impact deformation fields between OpenRadioss and LS [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Validation comparison of wall-force and energy responses for the same Yaris full-frontal [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Baseline vehicle models used for the dataset generation in [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Front-support components whose thicknesses are varied as design parameters for vehicle [PITH_FULL_IMAGE:figures/full_fig_p025_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Lower-rail and subframe components whose thicknesses are varied as design parameters [PITH_FULL_IMAGE:figures/full_fig_p025_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Representative deformed configurations from the Dodge Neon campaign shown in [PITH_FULL_IMAGE:figures/full_fig_p027_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Representative deformed configurations from the Dodge Neon campaign shown in side [PITH_FULL_IMAGE:figures/full_fig_p028_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Representative deformed configurations from the Toyota Yaris campaign shown in [PITH_FULL_IMAGE:figures/full_fig_p029_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Representative deformed configurations from the Toyota Yaris campaign shown in side [PITH_FULL_IMAGE:figures/full_fig_p030_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Representative deformed configurations from the Chevrolet Silverado campaign shown in [PITH_FULL_IMAGE:figures/full_fig_p031_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Representative deformed configurations from the Chevrolet Silverado campaign shown [PITH_FULL_IMAGE:figures/full_fig_p032_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Geometry and loading setup of the simplified bumper beam assembly used for dataset [PITH_FULL_IMAGE:figures/full_fig_p039_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Representative bumper-beam pole-impact configurations sampled from the design space. [PITH_FULL_IMAGE:figures/full_fig_p044_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Representative deformed configurations from the bumper-beam dataset used in [PITH_FULL_IMAGE:figures/full_fig_p045_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Column-normalized feature impor￾tance averaged over XGBoost, LightGBM, and CatBoost. The plotted values are model-internal importance scores averaged over the three boost￾ers, then normalized independently within each target so that every target column sums to one. AutoML versus hand-picked boosters. Auto￾Gluon’s best_quality preset reaches a mean R2 of 0.760, close to the bagged-tree baselines but below … view at source ↗
read the original abstract

Crash simulation is a cornerstone of modern vehicle development because it reduces the need for costly physical prototypes, accelerates safety-driven design iteration, and increasingly supports virtual testing workflows. At the same time, modeling structural crash mechanics remains exceptionally challenging: the response is governed by nonlinear contact, large deformation, material plasticity, failure, and complex multi-body interactions evolving over space and time on high-resolution finite-element meshes. In this work, we introduce CarCrashNet, a public high-fidelity open-source benchmark for data-driven structural crash simulation. CarCrashNet combines component-scale and full-vehicle simulations in a multi-modal format, including more than 14,000 bumper-beam pole-impact simulations with varying geometry, materials, and boundary conditions, together with 825 full-vehicle crash simulations built from three industry-standard vehicle models of increasing structural complexity: Dodge Neon, Toyota Yaris, and Chevrolet Silverado. To establish the reliability of the benchmark, we validate our open-source finite-element workflow based on OpenRadioss against both experimental crash data and the commercial solver Ansys LS-DYNA. We also introduce CrashSolver, a machine-learning model designed for full-vehicle crash prediction from high-resolution finite-element crash data. We further perform extensive benchmarking across the released datasets and evaluate CrashSolver against state-of-the-art geometric deep learning and transformer-based neural solvers. Our results position CarCrashNet as a foundation for reproducible research in structural simulation, crashworthiness modeling, and AI-driven virtual crash testing. The dataset is available at https://github.com/Mohamedelrefaie/CarCrashNet.

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

1 major / 3 minor

Summary. The manuscript introduces CarCrashNet, a public large-scale benchmark dataset for data-driven structural crash simulation. It comprises more than 14,000 component-scale bumper-beam pole-impact simulations with varied geometry, materials, and boundary conditions, plus 825 full-vehicle crash simulations using three industry-standard models (Dodge Neon, Toyota Yaris, Chevrolet Silverado). The authors validate their open-source OpenRadioss finite-element workflow against both experimental crash data and the commercial Ansys LS-DYNA solver. They further propose CrashSolver, a hierarchical neural model for full-vehicle crash prediction, and benchmark it against geometric deep learning and transformer-based neural solvers, reporting superior performance on held-out data.

Significance. If the OpenRadioss validation holds with quantitative fidelity, CarCrashNet would address a critical gap by providing the first open high-fidelity multi-modal dataset and solver for structural crash mechanics, enabling reproducible AI research in crashworthiness and virtual testing. The empirical outperformance claims and dataset release constitute concrete contributions that could accelerate progress in the field.

major comments (1)
  1. The central claim that CarCrashNet constitutes a reliable benchmark rests on the assertion that the OpenRadioss workflow faithfully reproduces structural crash physics. The abstract states validation against experimental data and LS-DYNA, but the manuscript must report quantitative agreement metrics (e.g., L2 errors on force-time histories, energy absorption, or nodal displacement fields) specifically for the full-vehicle cases across the three models; qualitative or bumper-beam-only comparisons would leave open the possibility of systematic biases in contact, plasticity, or failure modeling that undermine both the dataset value and the reported superiority of CrashSolver.
minor comments (3)
  1. Clarify the precise input/output tensor shapes and hierarchical processing steps in CrashSolver (e.g., how component-scale features are aggregated for full-vehicle prediction) to aid reproducibility.
  2. Add error bars or statistical significance tests to the benchmarking tables comparing CrashSolver against geometric deep learning and transformer baselines.
  3. Ensure the GitHub repository link includes the exact OpenRadioss input decks and post-processing scripts used for the validation studies.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive evaluation of our work and for the constructive major comment. We address it point by point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: The central claim that CarCrashNet constitutes a reliable benchmark rests on the assertion that the OpenRadioss workflow faithfully reproduces structural crash physics. The abstract states validation against experimental data and LS-DYNA, but the manuscript must report quantitative agreement metrics (e.g., L2 errors on force-time histories, energy absorption, or nodal displacement fields) specifically for the full-vehicle cases across the three models; qualitative or bumper-beam-only comparisons would leave open the possibility of systematic biases in contact, plasticity, or failure modeling that undermine both the dataset value and the reported superiority of CrashSolver.

    Authors: We agree that quantitative validation metrics for the full-vehicle simulations are essential to substantiate the benchmark's reliability. The current manuscript provides validation of the OpenRadioss workflow against experimental data primarily at the component (bumper-beam) scale and includes qualitative comparisons plus selected global metrics for the full-vehicle models (Dodge Neon, Toyota Yaris, Chevrolet Silverado). However, we acknowledge that comprehensive quantitative agreement metrics—such as L2 errors on force-time histories, energy absorption, and nodal displacement fields—specifically comparing OpenRadioss to LS-DYNA and experiments across all three full-vehicle cases are not reported in sufficient detail. In the revised manuscript we will add these quantitative metrics for the full-vehicle cases to close this gap and further support the claims regarding dataset fidelity and CrashSolver performance. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical dataset release and trained model evaluation on held-out splits

full rationale

The paper introduces CarCrashNet as a new public dataset of bumper-beam and full-vehicle crash simulations generated via OpenRadioss, validates the workflow against external experimental data and the independent commercial solver LS-DYNA, and trains CrashSolver on the data with performance reported on separate test splits. No derivation chain, equations, or first-principles results are presented that reduce to fitted parameters or self-citations by construction. The central claims rest on new data release plus empirical comparisons rather than any self-referential prediction step.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claims rest on the fidelity of the OpenRadioss simulations matching reality and the generalization of the trained neural model; no new physical entities are postulated.

free parameters (1)
  • CrashSolver network hyperparameters
    Architecture depth, learning rate, and other training choices are selected to fit the simulation data.
axioms (1)
  • domain assumption OpenRadioss finite-element models with chosen material and contact settings reproduce real crash mechanics to sufficient accuracy
    Invoked when claiming the dataset is high-fidelity after validation against experiments and LS-DYNA.

pith-pipeline@v0.9.0 · 5827 in / 1194 out tokens · 26294 ms · 2026-05-20T22:42:47.611051+00:00 · methodology

Review history (2 revisions) →

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

61 extracted references · 61 canonical work pages · 4 internal anchors

  1. [1]

    Transolver: A Fast Transformer Solver for PDEs on General Geometries

    Transolver: A fast transformer solver for pdes on general geometries , author=. arXiv preprint arXiv:2402.02366 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  2. [2]

    arXiv preprint arXiv:2512.20399 , year=

    GeoTransolver: Learning Physics on Irregular Domains Using Multi-scale Geometry Aware Physics Attention Transformer , author=. arXiv preprint arXiv:2512.20399 , year=

    work page arXiv

  3. [3]

    International Journal of Impact Engineering , volume=

    Axial crushing of thin-walled high-strength steel sections , author=. International Journal of Impact Engineering , volume=. 2006 , doi=

    work page 2006

  4. [4]

    2026 , url =

    Crash Simulation Vehicle Models , howpublished =. 2026 , url =

    work page 2026

  5. [5]

    2026 , url =

    Finite Element Models , howpublished =. 2026 , url =

    work page 2026

  6. [6]

    2016 , doi =

    2010 Toyota Yaris Finite Element Model Validation: Coarse Mesh , howpublished =. 2016 , doi =

    work page 2010

  7. [7]

    2016 , doi =

    2010 Toyota Yaris Finite Element Model Validation: Detail Mesh , howpublished =. 2016 , doi =

    work page 2010

  8. [8]

    2011 , number =

    Development and Validation of a Finite Element Model for the 2010 Toyota Yaris Passenger Sedan , institution =. 2011 , number =

    work page 2010

  9. [9]

    2012 , number =

    Extended Validation of the Finite Element Model for the 2010 Toyota Yaris Passenger Sedan , institution =. 2012 , number =

    work page 2010

  10. [10]

    Computer Methods in Applied Mechanics and Engineering , volume=

    A machine learning framework for accelerating the design process using CAE simulations: An application to finite element analysis in structural crashworthiness , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2021 , publisher=

    work page 2021

  11. [11]

    Computer Methods in Applied Mechanics and Engineering , volume=

    Introducing finite element method integrated networks (FEMIN) , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2024 , publisher=

    work page 2024

  12. [12]

    Computer Methods in Applied Mechanics and Engineering , volume=

    Accelerating crash simulations with Finite Element Method Integrated Networks (FEMIN): Comparing two approaches to replace large portions of a FEM simulation , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2025 , publisher=

    work page 2025

  13. [13]

    The Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track , year=

    Bubbleformer: Forecasting Boiling with Transformers , author=. The Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track , year=

    work page

  14. [14]

    SIA Simulation num

    Comparing traditional surrogate modelling and neural fields for vehicle crash simulation data , author=. SIA Simulation num

    work page

  15. [15]

    arXiv preprint arXiv:2503.17386 , year=

    A new graph-based surrogate model for rapid prediction of crashworthiness performance of vehicle panel components , author=. arXiv preprint arXiv:2503.17386 , year=

    work page arXiv

  16. [16]

    International Journal of Crashworthiness , volume=

    Crashworthiness performance of thin-walled structures towards design configuration in vehicle crash boxes application: a review , author=. International Journal of Crashworthiness , volume=. 2025 , publisher=

    work page 2025

  17. [17]

    International Journal of Impact Engineering , volume=

    Neural network modelling of mechanical joints for the application in large-scale crash analyses , author=. International Journal of Impact Engineering , volume=. 2023 , publisher=

    work page 2023

  18. [18]

    International Journal of Crashworthiness , volume=

    A review on finite element modelling and experimental analysis of crashworthiness design of automotive body , author=. International Journal of Crashworthiness , volume=. 2025 , publisher=

    work page 2025

  19. [19]

    International Journal for Numerical Methods in Engineering , volume=

    Nonintrusive Local/Global Coupling With Local Deep Learning-Based Models for the Effective Simulation of Spotwelded Structures Under Impact , author=. International Journal for Numerical Methods in Engineering , volume=. 2025 , publisher=

    work page 2025

  20. [20]

    Thin-walled structures , volume=

    Design and analysis of an automotive bumper beam in low-speed frontal crashes , author=. Thin-walled structures , volume=. 2009 , publisher=

    work page 2009

  21. [21]

    The Journal of Engineering , volume=

    The Automotive Bumper Beam in the Era of the 4th Industrial Revolution , author=. The Journal of Engineering , volume=. 2025 , publisher=

    work page 2025

  22. [22]

    Inventions , volume=

    Intelligent Damage Prediction During Vehicle Collisions Based on Simulation Datasets , author=. Inventions , volume=. 2025 , publisher=

    work page 2025

  23. [23]

    PAMM , volume=

    Machine learning enhanced optimisation of crash box design for crashworthiness analysis , author=. PAMM , volume=. 2023 , publisher=

    work page 2023

  24. [24]

    Computers in Industry , volume=

    Explainable artificial intelligence for enhancing system understanding and interpretability of numerical crash simulations , author=. Computers in Industry , volume=. 2026 , publisher=

    work page 2026

  25. [25]

    arXiv preprint arXiv:2511.10821 , year=

    MECHBench: A Set of Black-Box Optimization Benchmarks originated from Structural Mechanics , author=. arXiv preprint arXiv:2511.10821 , year=

    work page arXiv

  26. [26]

    Journal of complexity , volume=

    Scrambling Sobol'and Niederreiter--Xing Points , author=. Journal of complexity , volume=. 1998 , publisher=

    work page 1998

  27. [27]

    Proceedings of the 7th International Symposium on Ballistics, Am

    A constitutive model and data for metals subjected to large strains, high strain rates and high temperatures , author=. Proceedings of the 7th International Symposium on Ballistics, Am. Def. Prep. Org.(ADPA), Netherlands, 1983 , pages=

    work page 1983

  28. [28]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Explicit algorithms for the nonlinear dynamics of shells , author =. Computer Methods in Applied Mechanics and Engineering , volume =. 1984 , doi =

    work page 1984

  29. [29]

    Technometrics , volume=

    A comparison of three methods for selecting values of input variables in the analysis of output from a computer code , author=. Technometrics , volume=. 2000 , publisher=

    work page 2000

  30. [30]

    Journal of statistical planning and inference , volume=

    Minimax and maximin distance designs , author=. Journal of statistical planning and inference , volume=. 1990 , publisher=

    work page 1990

  31. [31]

    TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Second

    Tabpfn: A transformer that solves small tabular classification problems in a second , author=. arXiv preprint arXiv:2207.01848 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  32. [32]

    AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data

    Autogluon-tabular: Robust and accurate automl for structured data , author=. arXiv preprint arXiv:2003.06505 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2003

  33. [33]

    A., Chavare, S., Akhare, D., Ranade, R., Cherukuri, R., and Tadepalli, S

    Automotive Crash Dynamics Modeling Accelerated with Machine Learning , author=. arXiv preprint arXiv:2510.15201 , year=

    work page arXiv

  34. [34]

    Journal of Machine Learning Research , volume=

    Neural operator: Learning maps between function spaces with applications to pdes , author=. Journal of Machine Learning Research , volume=

    work page

  35. [35]

    International Conference on Learning Representations , year =

    Pfaff, Tobias and Fortunato, Meire and Sanchez-Gonzalez, Alvaro and Battaglia, Peter , title =. International Conference on Learning Representations , year =

    work page

  36. [36]

    Advances in neural information processing systems , volume=

    Attention is all you need , author=. Advances in neural information processing systems , volume=

    work page

  37. [37]

    International Conference on Learning Representations , year =

    Fourier Neural Operator for Parametric Partial Differential Equations , author =. International Conference on Learning Representations , year =

    work page

  38. [38]

    Nature machine intelligence , volume=

    Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators , author=. Nature machine intelligence , volume=. 2021 , publisher=

    work page 2021

  39. [39]

    International conference on machine learning , pages=

    Learning to simulate complex physics with graph networks , author=. International conference on machine learning , pages=. 2020 , organization=

    work page 2020

  40. [40]

    Advances in Neural Information Processing Systems , volume=

    Towards foundation models for scientific machine learning: Characterizing scaling and transfer behavior , author=. Advances in Neural Information Processing Systems , volume=

    work page

  41. [41]

    Advances in Neural Information Processing Systems , volume=

    Drivaernet++: A large-scale multimodal car dataset with computational fluid dynamics simulations and deep learning benchmarks , author=. Advances in Neural Information Processing Systems , volume=

    work page

  42. [42]

    Advances in neural information processing systems , volume=

    Pdebench: An extensive benchmark for scientific machine learning , author=. Advances in neural information processing systems , volume=

    work page

  43. [43]

    Advances in Neural Information Processing Systems , volume=

    Airfrans: High fidelity computational fluid dynamics dataset for approximating reynolds-averaged navier--stokes solutions , author=. Advances in Neural Information Processing Systems , volume=

    work page

  44. [44]

    arXiv preprint arXiv:2310.05963 , year =

    Luo, Yining and Chen, Yingfa and Zhang, Zhen , title =. arXiv preprint arXiv:2310.05963 , year =

    work page arXiv

  45. [45]

    Advances in Neural Information Processing Systems , volume=

    Lagrangebench: A lagrangian fluid mechanics benchmarking suite , author=. Advances in Neural Information Processing Systems , volume=

    work page

  46. [46]

    Advances in Neural Information Processing Systems , volume=

    The well: a large-scale collection of diverse physics simulations for machine learning , author=. Advances in Neural Information Processing Systems , volume=

    work page

  47. [47]

    Advances in Neural Information Processing Systems , volume=

    Apebench: A benchmark for autoregressive neural emulators of pdes , author=. Advances in Neural Information Processing Systems , volume=

    work page

  48. [48]

    Advances in Neural Information Processing Systems , volume=

    Pinnacle: A comprehensive benchmark of physics-informed neural networks for solving pdes , author=. Advances in Neural Information Processing Systems , volume=

    work page

  49. [49]

    Advances in Neural Information Processing Systems , volume=

    Poseidon: Efficient foundation models for pdes , author=. Advances in Neural Information Processing Systems , volume=

    work page

  50. [50]

    Journal of Advances in Modeling Earth Systems , volume=

    WeatherBench: a benchmark data set for data-driven weather forecasting , author=. Journal of Advances in Modeling Earth Systems , volume=. 2020 , publisher=

    work page 2020

  51. [51]

    0: A benchmark for data-driven climate projections , author=

    ClimateBench v1. 0: A benchmark for data-driven climate projections , author=. Journal of Advances in Modeling Earth Systems , volume=. 2022 , publisher=

    work page 2022

  52. [52]

    Scientific Data , year=

    WxC-Bench: A novel dataset for weather and climate downstream tasks , author=. Scientific Data , year=

    work page

  53. [53]

    arXiv preprint arXiv:2601.01829 , year =

    Hu, Peiyan and Feng, Haodong and Liu, Hongyuan and Yan, Tongtong and Deng, Wenhao and Gao, Tianrun and others , title =. arXiv preprint arXiv:2601.01829 , year =

    work page arXiv

  54. [54]

    Acs Catalysis , volume=

    Open catalyst 2020 (OC20) dataset and community challenges , author=. Acs Catalysis , volume=. 2021 , publisher=

    work page 2020

  55. [55]

    Tripnet: Learning large-scale high-fidelity 3d car aerodynamics with triplane networks.arXiv preprint arXiv:2503.17400, 2025

    Tripnet: Learning large-scale high-fidelity 3d car aerodynamics with triplane networks , author=. arXiv preprint arXiv:2503.17400 , year=

    work page arXiv

  56. [56]

    International Design Engineering Technical Conferences and Computers and Information in Engineering Conference , volume=

    Blendednet: A blended wing body aircraft dataset and surrogate model for aerodynamic predictions , author=. International Design Engineering Technical Conferences and Computers and Information in Engineering Conference , volume=. 2025 , organization=

    work page 2025

  57. [57]

    arXiv preprint arXiv:2512.07847 , year =

    Elrefaie, Mohamed and Shu, Dule and Klenk, Matt and Ahmed, Faez , title =. arXiv preprint arXiv:2512.07847 , year =

    work page arXiv

  58. [58]

    Journal of Mechanical Design , volume=

    Drivaernet: A parametric car dataset for data-driven aerodynamic design and prediction , author=. Journal of Mechanical Design , volume=. 2025 , publisher=

    work page 2025

  59. [59]

    International Design Engineering Technical Conferences and Computers and Information in Engineering Conference , volume=

    Drivaernet: A parametric car dataset for data-driven aerodynamic design and graph-based drag prediction , author=. International Design Engineering Technical Conferences and Computers and Information in Engineering Conference , volume=. 2024 , organization=

    work page 2024

  60. [60]

    BlendedNet++: A dataset and benchmark for field-resolved aerodynamics and inverse design of blended wing body aircraft

    BlendedNet++: A Large-Scale Blended Wing Body Aerodynamics Dataset and Benchmark , author=. arXiv preprint arXiv:2512.03280 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  61. [61]

    arXiv preprint arXiv:2502.04317 , year=

    Factorized implicit global convolution for automotive computational fluid dynamics prediction , author=. arXiv preprint arXiv:2502.04317 , year=

    work page arXiv