pith. machine review for the scientific record. sign in

arxiv: 2604.06358 · v1 · submitted 2026-04-07 · 💻 cs.GR · cs.AI

Recognition: 2 theorem links

· Lean Theorem

GS-Surrogate: Deformable Gaussian Splatting for Parameter Space Exploration of Ensemble Simulations

Ziwei Li , Rumali Perera , Angus Forbes , Ken Moreland , Dave Pugmire , Scott Klasky , Wei-Lun Chao , Han-Wei Shen
Authors on Pith no claims yet

Pith reviewed 2026-05-10 17:52 UTC · model grok-4.3

classification 💻 cs.GR cs.AI
keywords Gaussian splattingensemble simulationvisualization surrogateparameter space explorationdeformable modelsscientific visualizationreal-time renderingisosurface extraction
0
0 comments X

The pith

Deformable Gaussian splatting separates simulation variations from visualization changes for real-time ensemble exploration.

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

The paper introduces GS-Surrogate, which builds one canonical 3D Gaussian field from ensemble data and then applies sequential deformations driven by parameter values. This structure keeps simulation-driven changes distinct from adjustments needed for specific visualizations such as isosurface thresholds or transfer function edits. A reader would care because full ensemble data is costly to store and re-render, so an explicit surrogate that supports interactive changes to both inputs and display settings could make post-hoc analysis practical on ordinary hardware. The method is shown to deliver real-time performance while handling multiple datasets and tasks.

Core claim

GS-Surrogate constructs a canonical Gaussian field as a base 3D representation and adapts it through sequential parameter-conditioned deformations. By separating simulation-related variations from visualization-specific changes, this explicit formulation enables efficient and controllable adaptation to different visualization tasks, such as isosurface extraction and transfer function editing.

What carries the argument

Sequential parameter-conditioned deformations applied to a canonical Gaussian field, which isolate simulation variations from visualization adjustments in an explicit 3D representation.

Load-bearing premise

Sequential deformations conditioned on parameters can accurately separate simulation variations from visualization modifications while keeping fidelity and speed across varied datasets.

What would settle it

A new ensemble dataset where the deformed splats produce visibly incorrect isosurfaces or transfer-function results compared with direct rendering of the original runs at interactive frame rates.

Figures

Figures reproduced from arXiv: 2604.06358 by Angus Forbes, Dave Pugmire, Han-Wei Shen, Ken Moreland, Rumali Perera, Scott Klasky, Wei-Lun Chao, Ziwei Li.

Figure 2
Figure 2. Figure 2: Overview of the two-stage training pipeline of GS-Surrogate. The [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Architecture of the two deformation networks, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visual comparison on a training volume-rendering instance from [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visual comparison on a representative unseen ensemble member [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparative results of the top three methods on three unseen [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 9
Figure 9. Figure 9: Zoomed-in views highlighting the visual differences between the [PITH_FULL_IMAGE:figures/full_fig_p008_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Interface for interactive parameter space exploration with GS [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Predicted volume renderings varying h ∈ {0.56,0.66,0.76,0.86} on the Nyx dataset, with OmM and OmB fixed. function is defined through four interactive control points that specify a piecewise linear mapping between scalar values and opacity. To ensure controllable and meaningful exploration, we fix the endpoint opacities of the transfer function: the first control point, corresponding to the lowest scalar … view at source ↗
Figure 12
Figure 12. Figure 12: TF space exploration on the Nyx dataset with fixed simulation [PITH_FULL_IMAGE:figures/full_fig_p009_12.png] view at source ↗
read the original abstract

Exploring ensemble simulations is increasingly important across many scientific domains. However, supporting flexible post-hoc exploration remains challenging due to the trade-off between storing the expensive raw data and flexibly adjusting visualization settings. Existing visualization surrogate models have improved this workflow, but they either operate in image space without an explicit 3D representation or rely on neural radiance fields that are computationally expensive for interactive exploration and encode all parameter-driven variations within a single implicit field. In this work, we introduce GS-Surrogate, a deformable Gaussian Splatting-based visualization surrogate for parameter-space exploration. Our method first constructs a canonical Gaussian field as a base 3D representation and adapts it through sequential parameter-conditioned deformations. By separating simulation-related variations from visualization-specific changes, this explicit formulation enables efficient and controllable adaptation to different visualization tasks, such as isosurface extraction and transfer function editing. We evaluate our framework on a range of simulation datasets, demonstrating that GS-Surrogate enables real-time and flexible exploration across both simulation and visualization parameter spaces.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 0 minor

Summary. The paper introduces GS-Surrogate, a deformable Gaussian Splatting surrogate for exploring ensemble simulation parameter spaces. It constructs a canonical Gaussian field as a base 3D representation and adapts it via sequential parameter-conditioned deformations that purportedly separate simulation-related variations from visualization-specific changes (e.g., transfer functions or isosurface extraction). The authors claim this explicit formulation supports efficient, controllable, real-time adaptation across both parameter classes and evaluate the framework on multiple simulation datasets to demonstrate real-time flexible exploration.

Significance. If the separation of parameter classes holds with preserved 3D structure and real-time performance, the approach would offer a concrete advance over image-space surrogates and implicit NeRF-based methods by providing an explicit, editable 3D Gaussian representation suitable for downstream visualization tasks in scientific computing.

major comments (2)
  1. [Abstract / Evaluation] Abstract and Evaluation section: the central claim of real-time performance and flexible exploration across simulation and visualization parameter spaces is asserted without any reported metrics, baselines, error measures, timing results, or ablation studies. This absence makes it impossible to assess whether the sequential deformations actually achieve the required separation or accuracy.
  2. [Method] Method section (deformation formulation): the sequential parameter-conditioned deformations applied to the canonical Gaussian field are presented without any explicit disentanglement mechanism (e.g., orthogonality loss, independence regularizer, or staged training). Because deformations are applied in sequence and conditioned on parameters, cross-talk between simulation and visualization edits remains possible and directly undermines the separation claim that is load-bearing for the paper's contribution.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. The comments highlight important areas for strengthening the quantitative support and methodological clarity of our work. We address each point below and indicate the corresponding revisions.

read point-by-point responses

  1. Referee: [Abstract / Evaluation] Abstract and Evaluation section: the central claim of real-time performance and flexible exploration across simulation and visualization parameter spaces is asserted without any reported metrics, baselines, error measures, timing results, or ablation studies. This absence makes it impossible to assess whether the sequential deformations actually achieve the required separation or accuracy.

    Authors: We agree that the abstract and evaluation section would be strengthened by explicit quantitative results. The current evaluation provides qualitative demonstrations of real-time exploration and flexible parameter control across multiple datasets, but lacks the requested numerical benchmarks. In the revised manuscript we will update the abstract to reference key results and expand the evaluation section with: (i) timing measurements (e.g., frames per second on target hardware), (ii) reconstruction accuracy metrics such as PSNR and SSIM against ground-truth renderings, (iii) direct comparisons to image-space surrogates and NeRF-based baselines, and (iv) ablation studies that isolate the contribution of sequential deformations to separation accuracy. These additions will enable a rigorous assessment of both performance and the claimed separation. revision: yes

  2. Referee: [Method] Method section (deformation formulation): the sequential parameter-conditioned deformations applied to the canonical Gaussian field are presented without any explicit disentanglement mechanism (e.g., orthogonality loss, independence regularizer, or staged training). Because deformations are applied in sequence and conditioned on parameters, cross-talk between simulation and visualization edits remains possible and directly undermines the separation claim that is load-bearing for the paper's contribution.

    Authors: The separation is realized by construction: simulation parameters are conditioned exclusively on the geometric Gaussian attributes (means, covariances, and rotations) that encode ensemble structural variations, while visualization parameters (transfer functions, isosurface thresholds) are conditioned only on the appearance attributes (spherical harmonics coefficients and opacities). Because these two groups of parameters modify disjoint subsets of the Gaussian representation, cross-talk is structurally limited. Nevertheless, we acknowledge that an auxiliary regularizer would provide additional assurance. In the revised method section we will introduce a lightweight orthogonality regularizer on the deformation gradients produced by simulation versus visualization parameters, together with an ablation that quantifies any residual interference. This constitutes a partial revision that augments rather than replaces the existing explicit formulation. revision: partial

Circularity Check

0 steps flagged

No circularity: new constructive surrogate architecture

full rationale

The paper presents GS-Surrogate as a novel construction: a canonical Gaussian field is built as base representation, then adapted via sequential parameter-conditioned deformations that explicitly separate simulation variations from visualization changes. No equations, predictions, or central claims reduce to inputs by construction, fitted parameters renamed as outputs, or load-bearing self-citations. The derivation chain is self-contained as an explicit 3D surrogate method evaluated on datasets, with no self-definitional loops or ansatzes smuggled via prior author work. This aligns with the absence of any quoted reduction in the abstract or described framework.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

Review is abstract-only, so ledger entries are inferred at high level from the described method; full parameter counts and assumptions cannot be verified.

free parameters (1)
  • Deformation conditioning parameters
    Values that control how the canonical field adapts to each simulation parameter; likely learned or tuned during construction.
axioms (1)
  • domain assumption Gaussian Splatting can serve as an explicit, deformable 3D scene representation.
    The method builds directly on established Gaussian Splatting techniques from computer graphics.
invented entities (2)
  • Canonical Gaussian field no independent evidence
    purpose: Base 3D representation that is deformed to represent different simulation states.
    Introduced as the starting point for the surrogate.
  • Sequential parameter-conditioned deformations no independent evidence
    purpose: Mechanism to adapt the base field while separating simulation and visualization variations.
    Core novel component described in the abstract.

pith-pipeline@v0.9.0 · 5498 in / 1364 out tokens · 44572 ms · 2026-05-10T17:52:39.278721+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

43 extracted references · 37 canonical work pages

  1. [1]

    Ahrens, S

    J. Ahrens, S. Jourdain, P. O’Leary, J. Patchett, D. H. Rogers, and M. Pe- tersen. An image-based approach to extreme scale in situ visualization and analysis. InSC’14: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, pp. 424–434. IEEE, 2014. doi: 10.1109/SC.2014.40 2

    work page doi:10.1109/sc.2014.40 2014

  2. [2]

    A. S. Almgren, J. B. Bell, M. J. Lijewski, Z. Luki ´c, and E. Van Andel. Nyx: A massively parallel amr code for computational cosmology.The Astrophysical Journal, 765(1):39, 2013. doi: 10.1088/0004-637X/765/1/ 39 5

    work page doi:10.1088/0004-637x/765/1/ 2013

  3. [3]

    J. Bae, S. Kim, Y . Yun, H. Lee, G. Bang, and Y . Uh. Per-gaussian embedding-based deformation for deformable 3d gaussian splatting. In European Conference on Computer Vision, pp. 321–335. Springer, 2024. doi: 10.48550/arXiv.2404.03613 Focus to learn more 2

    work page doi:10.48550/arxiv.2404.03613 2024

  4. [4]

    Bartholomew, G

    P. Bartholomew, G. Deskos, R. A. Frantz, F. N. Schuch, E. Lamballais, and S. Laizet. Xcompact3d: An open-source framework for solving turbulence problems on a cartesian mesh.SoftwareX, 12:100550, 2020. doi: 10. 1016/j.softx.2020.100550 5

    work page arXiv 2020

  5. [5]

    A. C. Bauer, H. Abbasi, J. Ahrens, H. Childs, B. Geveci, S. Klasky et al. In situ methods, infrastructures, and applications on high performance computing platforms. InComputer Graphics Forum, vol. 35, pp. 577–597. Wiley Online Library, 2016. doi: 10.1111/cgf.12930 1

    work page doi:10.1111/cgf.12930 2016

  6. [6]

    Berger, J

    M. Berger, J. Li, and J. A. Levine. A generative model for volume rendering.IEEE transactions on visualization and computer graphics, 25(4):1636–1650, 2018. doi: 10.1109/TVCG.2018.2816059 2

    work page doi:10.1109/tvcg.2018.2816059 2018

  7. [7]

    Biedert and C

    T. Biedert and C. Garth. Contour tree depth images for large data visualiza- tion. InEGPGV@ EuroVis, pp. 77–86, 2015. doi: 10.2312/pgv.20151158 2

    work page doi:10.2312/pgv.20151158 2015

  8. [8]

    Biswas and T

    A. Biswas and T. Turton. Cloverleaf data artifacts for artimis ldrd. Tech- nical report, Los Alamos National Laboratory (LANL), 2026. doi: 10. 25583/3022785 5

    2026

  9. [9]

    A. Bock, A. Pembroke, M. L. Mays, L. Rastaetter, T. Ropinski, and A. Ynnerman. Visual verification of space weather ensemble simulations. In2015 IEEE Scientific Visualization Conference (SciVis), pp. 17–24. IEEE, 2015. doi: 10.1109/SciVis.2015.7429487 2

    work page doi:10.1109/scivis.2015.7429487 2015

  10. [10]

    Sarikaya and M

    S. Bruckner and T. Möller. Result-driven exploration of simulation param- eter spaces for visual effects design.IEEE Transactions on Visualization and Computer Graphics, 16(6):1468–1476, 2010. doi: 10.1109/TVCG. 2010.190 2

    work page doi:10.1109/tvcg 2010

  11. [11]

    Feder Cooper, Katherine Lee, Madiha Zahrah Choksi, Solon Barocas, Christopher De Sa, James Grimmelmann, Jon Kleinberg, Siddhartha Sen, and Baobao Zhang

    A. Cao and J. Johnson. Hexplane: A fast representation for dynamic scenes. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 130–141, 2023. doi: 10.48550/arXiv.2301. 09632 2

    work page doi:10.48550/arxiv.2301 2023

  12. [12]

    Chao, H.-Y

    B. Chao, H.-Y . Tseng, L. Porzi, C. Gao, T. Li, Q. Li et al. Textured gaussians for enhanced 3d scene appearance modeling. InProceedings of the Computer Vision and Pattern Recognition Conference, pp. 8964–8974,

  13. [13]

    doi: 10.48550/arXiv.2411.18625 9

    work page doi:10.48550/arxiv.2411.18625

  14. [14]

    A. Chen, Z. Xu, A. Geiger, J. Yu, and H. Su. Tensorf: Tensorial radiance fields. InEuropean conference on computer vision, pp. 333–350. Springer,

  15. [15]

    doi: 10.48550/arXiv.2203.09517 1

    work page doi:10.48550/arxiv.2203.09517

  16. [16]

    H. Chen, S. Zhang, W. Chen, H. Mei, J. Zhang, A. Mercer et al. Uncertainty-aware multidimensional ensemble data visualization and ex- ploration.IEEE transactions on visualization and computer graphics, 21(9):1072–1086, 2015. doi: 10.1109/TVCG.2015.2410278 2

    work page doi:10.1109/tvcg.2015.2410278 2015

  17. [17]

    S. Frey, F. Sadlo, and T. Ertl. Explorable volumetric depth images from raycasting. In2013 XXVI Conference on Graphics, Patterns and Images, pp. 123–130. IEEE, 2013. doi: 10.1109/SIBGRAPI.2013.26 2

    work page doi:10.1109/sibgrapi.2013.26 2013

  18. [18]

    Fridovich-Keil, G

    S. Fridovich-Keil, G. Meanti, F. R. Warburg, B. Recht, and A. Kanazawa. K-planes: Explicit radiance fields in space, time, and appearance. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 12479–12488, 2023. doi: 10.48550/arXiv.2301.10241 2, 6

    work page doi:10.48550/arxiv.2301.10241 2023

  19. [19]

    Fridovich-Keil, A

    S. Fridovich-Keil, A. Yu, M. Tancik, Q. Chen, B. Recht, and A. Kanazawa. Plenoxels: Radiance fields without neural networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 5501–5510, 2022. doi: 10.48550/arXiv.2112.05131 2

    work page doi:10.48550/arxiv.2112.05131 2022

  20. [20]

    Han and C

    J. Han and C. Wang. Coordnet: Data generation and visualization gen- eration for time-varying volumes via a coordinate-based neural network. IEEE Transactions on Visualization and Computer Graphics, 29(12):4951– 4963, 2022. doi: 10.1109/TVCG.2022.3197203 1

    work page doi:10.1109/tvcg.2022.3197203 2022

  21. [21]

    W. He, J. Wang, H. Guo, K.-C. Wang, H.-W. Shen, M. Raj et al. Insi- tunet: Deep image synthesis for parameter space exploration of ensemble simulations.IEEE transactions on visualization and computer graphics, 26(1):23–33, 2019. doi: 10.1109/TVCG.2019.2934312 1, 2, 6

    work page doi:10.1109/tvcg.2019.2934312 2019

  22. [22]

    Huynh-Thu and M

    Q. Huynh-Thu and M. Ghanbari. Scope of validity of psnr in image/video quality assessment.Electronics letters, 44(13):800–801, 2008. doi: 10. 1049/el:20080522 6

    2008

  23. [23]

    3d gaussian splatting for real-time radiance field rendering

    B. Kerbl, G. Kopanas, T. Leimkühler, G. Drettakis, et al. 3d gaussian splatting for real-time radiance field rendering.ACM Trans. Graph., 42(4):139–1, 2023. doi: 10.48550/arXiv.2308.04079 1, 2, 4

    work page doi:10.48550/arxiv.2308.04079 2023

  24. [24]

    Li, S.-S

    D. Li, S.-S. Huang, Z. Lu, X. Duan, and H. Huang. St-4dgs: Spatial- temporally consistent 4d gaussian splatting for efficient dynamic scene rendering. InACM SIGGRAPH 2024 Conference Papers, pp. 1–11, 2024. doi: 10.1145/3641519.3657520 2

    work page doi:10.1145/3641519.3657520 2024

  25. [25]

    Z. Lu, X. Guo, L. Hui, T. Chen, M. Yang, X. Tang et al. 3d geometry-aware deformable gaussian splatting for dynamic view synthesis. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8900–8910, 2024. doi: 10.48550/arXiv.2404.06270 2

    work page doi:10.48550/arxiv.2404.06270 2024

  26. [26]

    K.-L. Ma. In situ visualization at extreme scale: Challenges and oppor- tunities.IEEE Computer Graphics and Applications, 29(6):14–19, 2009. doi: 10.1109/MCG.2009.120 1, 2

    work page doi:10.1109/mcg.2009.120 2009

  27. [27]

    Matkovic, D

    K. Matkovic, D. Gracanin, B. Klarin, and H. Hauser. Interactive visual analysis of complex scientific data as families of data surfaces.IEEE Transactions on Visualization and Computer Graphics, 15(6):1351–1358,

  28. [28]

    doi: 10.1109/TVCG.2009.155 2

    work page doi:10.1109/tvcg.2009.155 2009

  29. [29]

    Mildenhall, P

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng. Nerf: Representing scenes as neural radiance fields for view synthesis.Communications of the ACM, 65(1):99–106, 2021. doi: 10. 48550/arXiv.2003.08934 1, 2

    work page arXiv 2021

  30. [30]

    Obermaier, K

    H. Obermaier, K. Bensema, and K. I. Joy. Visual trends analysis in time- varying ensembles.IEEE transactions on visualization and computer graphics, 22(10):2331–2342, 2015. doi: 10.1109/TVCG.2015.2507592 2

    work page doi:10.1109/tvcg.2015.2507592 2015

  31. [31]

    Moritz, C

    D. Orban, D. F. Keefe, A. Biswas, J. Ahrens, and D. Rogers. Drag and track: A direct manipulation interface for contextualizing data instances within a continuous parameter space.IEEE transactions on visualization and computer graphics, 25(1):256–266, 2018. doi: 10.1109/tvcg.2018. 2865051 2

    work page doi:10.1109/tvcg.2018 2018

  32. [32]

    J. Poco, A. Dasgupta, Y . Wei, W. Hargrove, C. R. Schwalm, D. N. Huntzinger et al. Visual reconciliation of alternative similarity spaces in climate modeling.IEEE transactions on visualization and computer graphics, 20(12):1923–1932, 2014. doi: 10.1109/TVCG.2014.2346755 2

    work page doi:10.1109/tvcg.2014.2346755 1923

  33. [33]

    Ringler, M

    T. Ringler, M. Petersen, R. L. Higdon, D. Jacobsen, P. W. Jones, and M. Maltrud. A multi-resolution approach to global ocean modeling.Ocean Modelling, 69:211–232, 2013. doi: 10.1016/j.ocemod.2013.04.010 5

    work page doi:10.1016/j.ocemod.2013.04.010 2013

  34. [34]

    N. Shi, J. Xu, H. Li, H. Guo, J. Woodring, and H.-W. Shen. Vdl-surrogate: A view-dependent latent-based model for parameter space exploration of ensemble simulations.IEEE Transactions on Visualization and Computer Graphics, 29(1):820–830, 2022. doi: 10.48550/arXiv.2207.13091 2, 5

    work page doi:10.48550/arxiv.2207.13091 2022

  35. [35]

    N. Shi, J. Xu, S. W. Wurster, H. Guo, J. Woodring, L. P. Van Roekel et al. Gnn-surrogate: A hierarchical and adaptive graph neural network for parameter space exploration of unstructured-mesh ocean simulations. IEEE Transactions on Visualization and Computer Graphics, 28(6):2301– 2313, 2022. doi: 10.1109/TVCG.2022.3165345 2, 3

    work page doi:10.1109/tvcg.2022.3165345 2022

  36. [36]

    J. Wang, X. Liu, H.-W. Shen, and G. Lin. Multi-resolution climate en- semble parameter analysis with nested parallel coordinates plots.IEEE transactions on visualization and computer graphics, 23(1):81–90, 2016. doi: 10.1109/TVCG.2016.2598830 2

    work page doi:10.1109/tvcg.2016.2598830 2016

  37. [37]

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli. Image quality assessment: from error visibility to structural similarity.IEEE transac- tions on image processing, 13(4):600–612, 2004. doi: 10.1109/TIP.2003. 819861 6

    work page doi:10.1109/tip.2003 2004

  38. [38]

    G. Wu, T. Yi, J. Fang, L. Xie, X. Zhang, W. Wei et al. 4d gaussian splatting for real-time dynamic scene rendering. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 20310–20320,

  39. [39]

    doi: 10.48550/arXiv.2310.08528 1, 2, 6

    work page doi:10.48550/arxiv.2310.08528

  40. [40]

    Z. Yang, X. Gao, W. Zhou, S. Jiao, Y . Zhang, and X. Jin. Deformable 3d gaussians for high-fidelity monocular dynamic scene reconstruction. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 20331–20341, 2024. doi: 10.48550/arXiv.2309.13101 1, 2

    work page doi:10.48550/arxiv.2309.13101 2024

  41. [41]

    Z. Yang, H. Yang, Z. Pan, and L. Zhang. Real-time photorealistic dynamic scene representation and rendering with 4d gaussian splatting.arXiv preprint arXiv:2310.10642, 2023. doi: 10.48550/arXiv.2310.10642 2

    work page doi:10.48550/arxiv.2310.10642 2023

  42. [42]

    S. Yao, Y . Lu, and C. Wang. Visnerf: Efficient multidimensional neu- ral radiance field representation for visualization synthesis of dynamic volumetric scenes. In2025 IEEE 18th Pacific Visualization Conference (PacificVis), pp. 235–245. IEEE, 2025. doi: 10.48550/arXiv.2502.16731 1, 2, 3, 6

    work page doi:10.48550/arxiv.2502.16731 2025

  43. [43]

    The Unreasonable Effectiveness of Deep Features as a Perceptual Metric

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang. The unreason- able effectiveness of deep features as a perceptual metric. InProceedings of the IEEE conference on computer vision and pattern recognition, pp. 586–595, 2018. doi: 10.48550/arXiv.1801.03924 6

    work page Pith review doi:10.48550/arxiv.1801.03924 2018