Pith. sign in

REVIEW 3 major objections 6 minor 3 cited by

MATEY: multiscale adaptive foundation models for spatiotemporal physical systems

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Adaptive tokenization lets a vision transformer match fine-grid accuracy at roughly half the sequence length.

desk verdict Variance-based adaptive tokenization is a plausible win, but the headline 2x compute reduction rests on average sequence length, not measured cost. read the letter →

arxiv 2412.20601 v1 pith:W5PGXKZA submitted 2024-12-29 cs.LG cs.AIcs.CE

Pei Zhang , M. Paul Laiu , Matthew Norman , Doug Stefanski , John Gounley This is my paper · ORCID
classification cs.LGcs.AIcs.CE
keywords adaptivetokenizationvisiontransformerspatiotemporalattentionmultiscalephysicalsystemsfoundationmodelspretrainingfine-tuningMATEY
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

In a vision transformer, the number of tokens, the patches the attention mechanism processes, grows quickly with resolution, making accurate multiscale forecasting of physical fields expensive. MATEY instead starts from coarse patches and splits only the highest-variance patches into smaller sub-patches, producing a mixed-resolution token sequence of adjustable length. On the colliding-thermals test problem, the paper reports that this adaptive scheme (Adap Mix) reaches lower prediction error than a uniformly fine patch model while using roughly half the average token sequence length, and that the same tradeoff holds for a second coarse/fine patch pair. It further reports that a time-then-space attention scheme is more efficient and expressive than full spatiotemporal attention or fully axial attention, and that pretraining on a multi-physics benchmark transfers to new physics mainly when attention is frozen and fine-tuning data are scarce.

What carries the argument

The mechanism that carries the accuracy-compute claim is variance-threshold adaptive tokenization. For each coarse patch, the model computes the variance of the solution field inside it; patches whose variance exceeds $\gamma_{\mathrm{sts}}$ times the maximum patch variance are split into smaller sub-token-scale (STS) patches, per Eq. (4). Two fusion strategies are defined: Adap Mix inserts the refined STS tokens directly into the token sequence so coarse and fine tokens attend together, and Adap Mul keeps coarse and refined tokens as separate sequences, merging them only at reconstruction. The attention variants, full spatiotemporal attention (ViT), time-then-space attention (SViT), and fully axial attention (AViT), are the comparators, and the paper's cost indices $L_{\mathrm{lin}}$ and $L_{\mathrm{quad}}$ make the sequence-length tradeoff quantitative.

What would settle it

A concrete test: on a holdout set, build an oracle refinement mask from the actual prediction error of the coarse-patch model and compare it with the variance mask from Eq. (4); if the variance-selected model needs substantially more tokens than the error-selected oracle to reach the same NRMSE, or fails on a low-variance thin boundary layer or smooth wide-amplitude gradient, the central tradeoff collapses.

Watch

Extended reading notes

Core claim

On its own terms, the central finding is that adaptive tokenization is not merely a compute-saving approximation: the paper reports cases where it improves accuracy over the uniformly fine patch model while roughly halving the average token sequence length. For Adap Mix settings such as $(p_x^1,p_x^{\mathrm{sts}},\gamma_{\mathrm{sts}})=(32,16,0.2)$ and similar intermediate values, the final NRMSE falls below the $16\times16$ or $8\times8$ uniform cases at about half the sequence length. A second finding is about attention structure: SViT, which factors attention into a time pass followed by one full spatial pass, reaches lower test error than full spatiotemporal attention (ViT) and fully axial attention (AViT) for comparable parameter counts, while also costing less training time than AViT. The third finding is that pretraining transfer to genuinely new physical equations with new variables is real but muted: pretrained models beat randomly initialized models in low-data, frozen-attention fine-tuning, and the advantage largely disappears under full fine-tuning with sufficient data.

Load-bearing premise

The accuracy/compute benefit depends on the assumption that a coarse patch's solution variance is a reliable indicator of where smaller patches are needed, so that refining only the highest-variance patches preserves accuracy while shortening the token sequence.

Editorial extensions

If this is right

  • At the demonstrated $2\times$ sequence-length reduction, adaptive tokenization makes ViT-based forecasting feasible for spatial grids where uniformly fine patch sizes would be computationally prohibitive.
  • By sweeping $\gamma_{\mathrm{sts}}$ from 1 to 0, Adap Mix interpolates continuously between a coarse-patch model and a fine-patch model, giving practitioners a single scalar knob to trade cost against accuracy.
  • The comparisons among attention schemes imply that fully decoupled axial attention should be avoided for spatiotemporal physics, while decoupling time only (SViT) preserves accuracy with lower cost than full spatiotemporal attention.
  • Pretrained representations transfer to new physical variables mainly when attention layers are frozen and fine-tuning data are scarce; full fine-tuning with enough data removes most of the pretraining advantage.

Reading between the lines

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

  • The variance criterion is the load-bearing choice; an editor would expect it to be tested against alternative refinement signals, such as local gradients, wavelet coefficients, or the actual error map of the coarse model.
  • If the $2\times$ reduction scales with resolution, the method's practical payoff is largest in high-resolution CFD and climate applications, where uniform fine tokenization is the bottleneck, not in the small grids shown here.
  • Adap Mix's advantage over Adap Mul suggests that letting coarse and fine tokens attend to each other, cross-scale correlation, matters more than just having more tokens; ablating variance selection against random refinement would isolate this.
  • The fine-tuning results imply that foundation-model benchmarks for physics should routinely include tasks with entirely new variables, such as thermal or electromagnetic fields, rather than only same-equation out-of-distribution flows, because that is where pretraining gains become visible and measurable.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper introduces MATEY, a vision-transformer-based framework for forecasting spatiotemporal physical systems, with two contributions: adaptive tokenization schemes (Adap Mul and Adap Mix) that refine patch sizes locally based on a variance threshold, and a comparison of three spatiotemporal attention mechanisms (ViT, SViT, and AViT). The authors evaluate MATEY on a colliding-thermals dataset, show that adaptive tokenization can reach lower NRMSE than uniform fine patch tokenization at smaller average sequence length, and demonstrate that models pretrained on PDEBench outperform randomly initialized models when fine-tuned on out-of-distribution colliding-thermals and MHD tasks, especially with frozen attention in low-data regimes. The paper also reports that fully decoupled axial attention (AViT) is less efficient and expressive than SViT or ViT in their experiments.

Significance. If the compute-efficiency claim is substantiated, adaptive tokenization is a useful contribution for applying ViTs to high-resolution physical simulations, where token counts become prohibitive. The paper offers a simple, interpretable refinement rule (local variance thresholding), guarantees a smooth interpolation between coarse and fine uniform tokenization for Adap Mix, and evaluates on genuinely out-of-distribution fine-tuning tasks that include variables absent from pretraining. The planned public release of data, code, and trained models is a strength. The main limitation is that the central '2x compute reduction' claim is currently supported only by an average-sequence-length proxy, not by measured cost; and the statistical evidence (error bars, multiple-run variability) is incomplete. With those gaps addressed, the paper would be a solid empirical contribution to scientific machine learning.

major comments (3)
  1. [Section 4.2, Eq. (10), Figure 5] The claimed '2x reduction in compute' is based on the average sequence length Lavg,mix, not on measured computational cost. Because the attention cost in ViT and SViT is quadratic in sequence length and because the manuscript states (Section 3, Figure 2) that variable-length samples are handled with padding masks, the actual per-step FLOPs and GPU time depend on the maximum sequence length in each batch, which can remain close to the fine-patch case even when Lavg,mix is halved. The paper should report wall-clock time per training step (and ideally FLOPs) for the adaptive versus uniform fine-patch runs, or provide a clear analysis of batch-level maximum lengths. Without this, the headline compute claim in the introduction and abstract is not established.
  2. [Figures 5-8, 10; Section 4.2] Most plots lack error bars: only the ViT panel in Figure 5 shows standard deviations from 3 runs, while SViT in Figure 5 and Figures 6, 7, 8, and 10 do not. The conclusion that Adap Mix 'achieves lower predictive errors than the more expensive ps=16x16 case despite requiring only half the average sequence length' is based on selected gamma values, and the gamma sweep is then presented with the best cases highlighted. The authors should provide repeated-seed variability for all central Pareto claims, and either report all gamma values in the sweep or justify the selection rule to avoid cherry-picking.
  3. [Section 3, Eq. (4)] The refinement criterion assumes that the variance of the solution field inside a patch is a sufficient proxy for where smaller patches are needed. This assumption is load-bearing for the entire accuracy/compute tradeoff, but it is not validated against alternative complexity measures or against error maps of the trained models. Physical features such as thin boundary layers, smooth wide-amplitude gradients, or near-uniform regions that still require fine resolution for accurate long-time prediction may not be refined, potentially collapsing the claimed benefit. The paper should either provide evidence that the selected patches correlate with prediction error on the test problem, or explicitly discuss this limitation and its consequences for generalization.
minor comments (6)
  1. [Section 3, Eq. (2)] The notation 'A ViT' is used inconsistently: the text and figures refer to 'AViT' and 'A ViT' interchangeably; please standardize.
  2. [Figure 7 caption] The caption contains a typo: 'hyperparamters (px1, pxsts, γsts)m' should be 'hyperparameters (px1, pxsts, gamma_sts)'.
  3. [Table A1] The number of colliding-thermals training trajectories is listed as [1, 6, 12, 24, 48] in Table A1, but Section 4.3 and Figure 8 report 1 to 24 trajectories; please reconcile the discrepancy.
  4. [Section 4.1, Figure 3] The conclusion that ViT is less expensive than SViT and AViT is drawn from a single dataset with a short token sequence (16x8x8); the text acknowledges this, but the generalized statement in the abstract and Section 5 should be couched as regime-specific.
  5. [Section 4.2] For Adap Mul, the 'dramatically more cost effective' conclusion is based on the linear and quadratic cost indices in Eqs. (11)-(12), not on measured time; the text should clarify that these are analytical estimates, not observed cost.
  6. [Related Work] The paper compares only against constant-patch baselines; a comparison with a recent adaptive tokenization method such as PMSViT, or at least a discussion of why it is not used as a baseline, would strengthen the claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: adaptive-tokenization accuracy is an empirical result; the gamma->1/0 convergence is an explicit design identity, and the only self-citation (miniWeather) is a dataset reference.

full rationale

The paper's central claims are empirical rather than derived from fitted inputs. The adaptive tokenization benefit (lower NRMSE at shorter average sequence length, Figure 5) is measured from trained models against a held-out test set; no parameter is fit to that test set and then reported as a prediction. The convergence of Adap Mix to the coarse patch solution at gamma_sts=1.0 and to the fine patch solution at gamma_sts=0.0 is explicitly built into the selection rule in Eq. (4) and the reconstruction/fusion operations in Eqs. (6)-(9), and the paper presents this as a design guarantee, not as a discovered prediction. The cost indices in Eqs. (10)-(12) are definitions used to plot empirical accuracy, not derived theoretical bounds. The only overlapping self-citation is Norman's miniWeather dataset, which is used as the source of the colliding-thermals data and is not load-bearing for any derivation; the pretraining baseline PDEBench and the MPP preprocessing are external citations without author overlap. No uniqueness theorem, ansatz, or fitted parameter is imported from the authors' prior work. The paper also explicitly acknowledges limitations of its fine-tuning transfer results. Thus there is no circular step requiring a nonzero score.

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

The central claims rest on variance-based refinement threshold and patch size choices, which are user-specified hyperparameters rather than fitted constants. There are no newly invented physical entities. The main unverified premises are the variance-as-complexity proxy and the out-of-distribution character of the fine-tuning tasks.

free parameters (3)
  • gamma_sts = swept: 0.0 to 0.8 in experiments
    Threshold fraction of the maximum patch variance used to select patches for refinement (Eq. 4). The headline 2x compute reduction with equal or better accuracy is shown at selected gamma values, so the result depends on this hyperparameter.
  • coarse patch size (px1, py1) = 32x32 or 16x16 in experiments
    Initial coarse patch size in adaptive tokenization. The accuracy/complexity tradeoff is measured relative to these base sizes.
  • STS patch size (pxsts, pysts) = 16x16 or 8x8 in experiments
    Refined sub-token-scale patch size. It determines the resolution of locally refined regions and the cost model in Eqs. (11)-(12).
assumptions (4)
  • domain assumption Local patch variance is a valid proxy for where smaller patches are needed to achieve accurate predictions.
    Used in Eq. (3)-(4) to define STS-IDs. If important features have low variance, such as smooth but wide gradients or thin boundary layers, the refinement criterion could miss them and the accuracy/compute benefit would not materialize.
  • domain assumption The colliding thermals and MHD datasets are sufficiently out-of-distribution relative to PDEBench to test transfer.
    Section 4.3 and Appendix A.1 argue the fine-tuning tasks include variables such as temperature and magnetic field that are absent in pretraining. The conclusions about pretraining value depend on this characterization.
  • standard math The standard Transformer components and the MPP preprocessing scheme behave as prior literature describes.
    MATEY inherits MHSA, MLP, InstanceNorm, and MPP-style pre/post processors without modification. Any failure of these assumptions would affect all compared models equally.
  • domain assumption The reported training settings are sufficient for fair comparison across attention schemes and tokenization variants.
    The attention comparison fixes the training budget across models. If some models need more steps to converge, the efficiency conclusions could be biased.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MATEY: multiscale adaptive foundation models for spatiotemporal physical systems." pith.science (2026). https://pith.science/paper/W5PGXKZA

@misc{pith2026241220601,
  author       = {Pith},
  title        = {Pith review of: MATEY: multiscale adaptive foundation models for spatiotemporal physical systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W5PGXKZA}},
  note         = {Machine review of arXiv:2412.20601}
}
read the original abstract

Accurate representation of the multiscale features in spatiotemporal physical systems using vision transformer (ViT) architectures requires extremely long, computationally prohibitive token sequences. To address this issue, we propose two adaptive tokenization schemes that dynamically adjust patch sizes based on local features: one ensures convergent behavior to uniform patch refinement, while the other offers better computational efficiency. Moreover, we present a set of spatiotemporal attention schemes, where the temporal or axial spatial dimensions are decoupled, and evaluate their computational and data efficiencies. We assess the performance of the proposed multiscale adaptive model, MATEY, in a sequence of experiments. The results show that adaptive tokenization schemes achieve improved accuracy without significantly increasing the length of the token sequence. Compared to a full spatiotemporal attention scheme or a scheme that decouples only the temporal dimension, we find that fully decoupled axial attention is less efficient and expressive, requiring more training time and model weights to achieve the same accuracy. Finally, we demonstrate in two fine-tuning tasks featuring different physics that models pretrained on PDEBench data outperform the ones trained from scratch, especially in the low data regime with frozen attention.

Figures

Figures reproduced from arXiv: 2412.20601 by the authors.

Figure 1
Figure 1. MATEY: multiscale adaptive foundation models for spatiotemporal physical systems. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Adaptive tokenization that dynamically adjusts patch sizes based on local features. There [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Learning efficiency of AViT, SViT, and ViT at three model sizes regarding final predictive [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Predicted temperature contours at t “ 590 from Ti-SViT models with constant patch sizes ps=16 ˆ 16 and ps=32 ˆ 32 and adaptive tokenization (Adap Mul with px1 “ py1 “ 32, pxsts “ pysts “ 16 , and γsts “ 0.2). Adap Mul predicts smoother, finer local structures that are …
Figure 5
Figure 5. Figure 5: Final NRMSE loss for Tiny ViT (left) and SViT (right) with adaptive tokenization — [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Final NRMSE loss for Tiny ViT (top) and SViT (bottom)with adaptive tokenization — [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Final NRMSE loss for Tiny AViT with adaptive tokenization — [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 9
Figure 9. Figure 9: For the limited fine-tuning test with the colliding thermals dataset, the pretrained models achieve significantly lower error than starting from scratch with randomly initialized parameters. Moreover, while this advantage persists as the number of fine-tuning data incr…
Figure 8
Figure 8. Figure 8: NRMSE loss for test set at different training data sizes in fine-tuning of colliding thermals [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Temperature contours of true solution vs predicted solutions from four fine-tuned models [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: NRMSE loss for test set at different training data sizes in fine-tuning of lid-driven cavity [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: Contours of true horizontal magnetic field values [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

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

  1. Automated Data Readiness for Scientific AI

    cs.AI 2026-07 conditional novelty 6.0 of 10

    REDI automates raw-to-AI-ready scientific data prep via a five-stage IPTSO pipeline with provenance, readiness scoring, and multi-domain validation matching expert references.

  2. Pixel-Resolved Long-Context Learning for Turbulence at Exascale: Resolving Small-scale Eddies Toward the Viscous Limit

    physics.flu-dyn 2025-07 conditional novelty 6.0 of 10

    A multiscale transformer with a new collective-based parallel attention method is claimed to be the first deep-learning model to reproduce small-scale turbulence statistics down to the viscous limit in 3D flow.

  3. PDE-Transformer: Efficient and Versatile Transformers for Physics Simulations

    cs.LG 2025-05 conditional novelty 6.0 of 10

    PDE-Transformer, a diffusion-transformer variant with shifted-window attention, multi-scale token processing, and per-channel tokens, outperforms leading transformer and operator baselines for PDE surrogate modeling a...

Reference graph

Works this paper leans on

14 extracted references · 4 canonical work pages · cited by 3 Pith papers

  1. [2]

    (18) The equations are solved by using a finite volume method with nx “ 256, ny“ 256 grid points in x and z directions, respectively

    at different locations following uniform distribution, xci„ Ur0.2L, 0.8Ls, zc1„ Ur0.2L, 0.3Ls, and zc2„ Ur0.7L, 0.8Ls, (16) with different elliptical shapes also following uniform distribution, rxi„ Ur0.1L, 0.2Ls and rzi„ Ur0.1L, 0.2Ls, (17) and with temperature amplitudes equally sampled from, T ci„ Ct10, 15, 20, 25u. (18) The equations are solved by usi...

    work page 2023

  2. [4]

    DPOT: Auto-regressive denoising operator transformer for large- scale PDE pre-training

    Zhongkai Hao, Chang Su, Songming Liu, Julius Berner, Chengyang Ying, Hang Su, Anima Anand- kumar, Jian Song, and Jun Zhu. DPOT: Auto-regressive denoising operator transformer for large- scale PDE pre-training. arXiv preprint arXiv:2403.03542,

    arXiv

  3. [5]

    Poseidon: Efficient Foundation Models for PDEs

    Maximilian Herde, Bogdan Raoni ´c, Tobias Rohner, Roger K ¨appeli, Roberto Molinaro, Emmanuel de B ´ezenac, and Siddhartha Mishra. Poseidon: Efficient Foundation Models for PDEs. arXiv preprint arXiv:2405.19101,

    arXiv

  4. [6]

    Axial attention in multidi- mensional transformers

    Jonathan Ho, Nal Kalchbrenner, Dirk Weissenborn, and Tim Salimans. Axial attention in multidi- mensional transformers. arXiv preprint arXiv:1912.12180,

    arXiv 1912

  5. [7]

    Yanghao Li, Chao-Yuan Wu, Haoqi Fan, Karttikeya Mangalam, Bo Xiong, Jitendra Malik, and Christoph Feichtenhofer

    URLhttps://arxiv.org/ abs/2309.14509. Yanghao Li, Chao-Yuan Wu, Haoqi Fan, Karttikeya Mangalam, Bo Xiong, Jitendra Malik, and Christoph Feichtenhofer. Mvitv2: Improved multiscale vision transformers for classification and detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recogni- tion, pp. 4804–4814,

    arXiv

  6. [8]

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo

    URL https://arxiv.org/abs/2402.17177. Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision , pp. 10012–10022,

    arXiv

  7. [9]

    Multiple physics pretraining for physical surrogate models

    Michael McCabe, Bruno R ´egaldo-Saint Blancard, Liam Holden Parker, Ruben Ohana, Miles Cranmer, Alberto Bietti, Michael Eickenberg, Siavash Golkar, Geraud Krawezik, Francois Lanusse, et al. Multiple physics pretraining for physical surrogate models. arXiv preprint arXiv:2310.02994,

    arXiv

  8. [10]

    original-date: 2018-07-24T02:29:06Z

    URL https://github.com/ mrnorman/miniWeather. original-date: 2018-07-24T02:29:06Z. Matthew R Norman. miniWeather. Technical report, Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States),

    work page 2018

Show all 14 references
  1. [11]

    300.0` T10px, zq` T20px, zq, (13) with one hot T10 and cold T20 thermals being T10px, zq“

    URL https://arxiv.org/abs/2404.09707. 14 A A PPENDIX A.1 D ATASETS Three datasets were used in the work: PDEBench (Takamoto et al., 2022), colliding thermals (Nor- man, 2024), and lid-driven cavity MHD flows. • PDEBench ( https://github.com/pdebench/PDEBench) consists of diver...

    2024 arXiv

  2. [13]

    A.3 P RETRAINING AND FINE -TUNING A.3.1 P RETRAINING Five 2D datasets from PDEBench Takamoto et al

    for the two spatiotemporal attention schemes (ViT and SViT) in a single colliding thermals trajectory. A.3 P RETRAINING AND FINE -TUNING A.3.1 P RETRAINING Five 2D datasets from PDEBench Takamoto et al. (2022) were used for pretraining, including shal- low water, diffusion rea...

    2022

  3. [14]

    Table A1: Cases and datasets Pretraining: PDEBench Takamoto et al

    for A ViT in a single colliding thermals trajectory. Table A1: Cases and datasets Pretraining: PDEBench Takamoto et al. (2022) Dataset Variables (C) Spatiotemporal res. ( Tˆ Hˆ W ) Ntraj trajectories Shallow-water h 101ˆ 128ˆ 128 1,000 Diffusion-reaction [diffre2d] ξ, ϕ 101ˆ 1...

    2022

  4. [2021]

    Cristian Bodnar, Wessel P Bruinsma, Ana Lucic, Megan Stanley, Johannes Brandstetter, Patrick Garvan, Maik Riechert, Jonathan Weyn, Haiyu Dong, Anna Vaughan, et al

    URLhttps://arxiv.org/abs/2103.15691. Cristian Bodnar, Wessel P Bruinsma, Ana Lucic, Megan Stanley, Johannes Brandstetter, Patrick Garvan, Maik Riechert, Jonathan Weyn, Haiyu Dong, Anna Vaughan, et al. Aurora: A foundation model of the atmosphere. arXiv preprint arXiv:2405.13063,

    arXiv

  5. [2023]

    doi: https://doi.org/10.1016/j.jcp.2023.112493

    ISSN 0021-9991. doi: https://doi.org/10.1016/j.jcp.2023.112493. URL https: //www.sciencedirect.com/science/article/pii/S0021999123005880. Qihang Fan, Quanzeng You, Xiaotian Han, Yongfei Liu, Yunzhe Tao, Huaibo Huang, Ran He, and Hongxia Yang. ViTAR: Vision Transformer with Any...

    2023

  6. [2024]

    Zhou Hang, Yuezhou Ma, Haixu Wu, Haowen Wang, and Mingsheng Long

    URL https: //arxiv.org/abs/2403.18361. Zhou Hang, Yuezhou Ma, Haixu Wu, Haowen Wang, and Mingsheng Long. Unisolver: PDE- Conditional Transformers Are Universal PDE Solvers. arXiv preprint arXiv:2405.17527,

    arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.