pith. sign in

arxiv: 2606.10612 · v1 · pith:E7FBJ6JVnew · submitted 2026-06-09 · 💻 cs.CV

GaussTrace: Provenance Analysis of 3D Gaussian Splatting Models with Evidence-based LLM Reasoning

Haoliang Han , Ziyuan Luo , Renjie Wan This is my paper

Pith reviewed 2026-06-27 13:37 UTC · model grok-4.3

classification 💻 cs.CV
keywords 3D Gaussian Splattingprovenance analysisLLM reasoningchain-of-thoughtmodel editingforensic traceabilityintellectual property3D assets
0
0 comments X

The pith

GaussTrace constructs directed provenance graphs for 3D Gaussian Splatting models by feeding attribute-wise statistical profiles and simulated edits into an LLM for chain-of-thought inference.

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

The paper presents GaussTrace as a framework that treats provenance analysis of 3D Gaussian Splatting models as an evidence-based reasoning task. It profiles statistical attributes of the model's parameters to capture intrinsic properties and runs hypothesis-driven simulations of common editing operations to generate auxiliary evidence about possible transformation paths. These two sources of evidence are supplied to a large language model that performs structured chain-of-thought reasoning to infer the direction of each transformation and produce explainable edges in a directed graph. A reader would care because 3DGS models are now widely shared and iteratively modified on digital platforms, raising issues of intellectual property protection and forensic traceability that cannot be addressed by methods requiring training data or access to edit histories. If the approach holds, it enables accurate and interpretable provenance reconstruction on any released model without those prerequisites.

Core claim

GaussTrace formulates provenance analysis as an evidence-based reasoning problem. It builds upon attribute-wise statistical profiling of 3DGS parameters to capture intrinsic properties. Hypothesis-driven editing simulations of common operations provide auxiliary evidence for plausible transformation pathways. These statistical and simulated cues jointly enable a Large Language Model to perform structured Chain-of-Thought reasoning, yielding directional provenance inferences and explainable edge reasons. Experimental results demonstrate that GaussTrace effectively constructs evolutionary relationships among diverse 3DGS models, delivering accurate, interpretable, and robust provenance graphs

What carries the argument

Evidence-based LLM reasoning that combines attribute-wise statistical profiling of 3DGS parameters with hypothesis-driven simulations of editing operations to infer directional transformation edges.

If this is right

  • Produces directed provenance graphs that link diverse 3DGS models through inferred editing sequences.
  • Yields both the graph structure and explicit textual reasons for each inferred edge.
  • Operates without any model training step or access to original editing histories.
  • Maintains performance across varied 3DGS models and editing operation types.

Where Pith is reading between the lines

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

  • The same statistical-plus-simulation evidence pattern could be tested on other explicit 3D representations such as meshes or neural radiance fields.
  • If the LLM inferences prove stable, platforms could run lightweight provenance checks on uploaded 3D assets as a standard forensic step.
  • A natural next measurement would be to quantify how often the method correctly ranks multiple plausible edit orders when more than one sequence is possible.

Load-bearing premise

Attribute-wise statistical profiles plus simulated editing operations supply enough independent evidence for an LLM to produce reliable directional inferences about real-world transformation sequences.

What would settle it

Apply GaussTrace to a controlled collection of 3DGS models whose exact sequence of real edits is known in advance and measure whether the inferred directed graph recovers the correct chronological order at rates above chance.

Figures

Figures reproduced from arXiv: 2606.10612 by Haoliang Han, Renjie Wan, Ziyuan Luo.

Figure 1
Figure 1. Figure 1: Illustration of our proposed scenario. A creator releases an original 3DGS model online, which may be modified and re￾distributed by malicious actors, leading to intellectual property infringement or harmful content propagation. Our GaussTrace can construct the evolutionary relationships among these models by performing evidence-based reasoning. The resulting provenance graph enables traceability of 3D ass… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of our GaussTrace framework. First, each 3DGS model is processed by attribute-wise statistical descriptors to capture intrinsic properties. Then, hypothesis-driven editing simulations are performed to generate plausible transformation signatures. These statistical and simulated cues are jointly encoded into a structured prompt. Finally, a Large Language Model (LLM) performs structured Chain-of… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of 3DGS provenance graph construction. From left to right: ground-truth, Our GaussTrace, ViT￾B (Dosovitskiy et al., 2020)+Integrity (Zhang et al., 2020) and Points CD (Butt & Maragos, 1998)+Integrity (Zhang et al., 2020). The green edges indicate correct connections with correct directions, pink edges denote correct connections with wrong directions, and red edges represent wrong con… view at source ↗
Figure 4
Figure 4. Figure 4: Impact of individual 3D Gaussian attributes on prove￾nance graph construction performance of GaussTrace. The Gaus￾sian attributes include position (µ), opacity (α), scale (s), rotation (r), and color (c). graph. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of explanations for provenance edges. Our method can generate human-readable justifications for each di￾rected relationship in the provenance graph, enabling traceable and interpretable model evolution. justment and rotation noise. GaussTrace correctly identifies this complex transformation and articulates the joint effect in its explanation. By grounding each inference in inter￾pretable cues,… view at source ↗
Figure 6
Figure 6. Figure 6: Example edge explanations produced by our method. Each directed relationship in the provenance graph is accompanied by a natural-language rationale that clarifies the underlying transformation, supporting transparent model evolution. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative comparison of 3DGS provenance graph construction. Top row, left to right: ground-truth, Our GaussTrace. Bottom row, left to right: ResNet (He et al., 2016)+Integrity (Zhang et al., 2020), Points CD (Butt & Maragos, 1998)+Integrity (Zhang et al., 2020). The green edges indicate correct connections with correct directions, pink edges denote correct connections with wrong directions, and red edges… view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative comparison of 3DGS provenance graph construction. Top row, left to right: ground-truth, Our GaussTrace. Bottom row, left to right: ResNet (He et al., 2016)+Integrity (Zhang et al., 2020), Points CD (Butt & Maragos, 1998)+Integrity (Zhang et al., 2020). The green edges indicate correct connections with correct directions, pink edges denote correct connections with wrong directions, and red edges… view at source ↗
read the original abstract

3D Gaussian Splatting (3DGS) is a powerful technique for creating high-fidelity 3D assets. However, the widespread sharing and iterative modification of 3DGS models across digital platforms create pressing challenges for intellectual property protection and forensic traceability. To address this, we propose GaussTrace, a novel framework for constructing directed provenance graphs for 3DGS models. GaussTrace formulates provenance analysis as an evidence-based reasoning problem. It builds upon attribute-wise statistical profiling of 3DGS parameters to capture intrinsic properties. Moreover, we introduce hypothesis-driven editing simulations of common operations to provide auxiliary evidence for plausible transformation pathways. These statistical and simulated cues jointly enable a Large Language Model (LLM) to perform structured Chain-of-Thought (CoT) reasoning, yielding directional provenance inferences and explainable edge reasons. Experimental results demonstrate that GaussTrace effectively constructs evolutionary relationships among diverse 3DGS models, delivering accurate, interpretable, and robust provenance graphs without requiring model training or access to editing histories. Project page: https://haolianghan.github.io/GaussTrace.

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 / 2 minor

Summary. The paper introduces GaussTrace, a framework for provenance analysis of 3D Gaussian Splatting (3DGS) models. It constructs directed provenance graphs by combining attribute-wise statistical profiling of 3DGS parameters with hypothesis-driven editing simulations. These provide evidence for a Large Language Model to perform Chain-of-Thought reasoning, producing directional inferences and explainable reasons for edges. The approach requires no model training or access to editing histories, and experimental results are claimed to show accurate, interpretable, and robust graphs for diverse 3DGS models.

Significance. If the central claims hold, this work could be significant for addressing intellectual property and traceability issues in the growing field of 3D asset creation and sharing using 3DGS. The method's training-free nature and use of LLM for structured reasoning on simulated evidence is a notable strength, offering interpretability valuable for forensic applications. It could inspire similar LLM-augmented approaches in other generative model domains.

major comments (2)
  1. [§3] §3 (Method): The assumption that attribute-wise statistical profiles augmented by editing simulations supply unique, disambiguating signatures for different transformation sequences is not validated. Multiple operations (pruning, densification, rotation, color shift) can induce statistically similar shifts in the same parameter distributions, leaving directional inferences dependent on untested LLM internal knowledge rather than provided evidence. This is load-bearing for the accuracy claim.
  2. [§4] §4 (Experiments): No quantitative metrics, error analysis, dataset details, ablation studies, or baseline comparisons are described to support the reported accuracy and robustness. Without these, it is impossible to determine whether results are affected by post-hoc choices in CoT prompts or simulation parameters, undermining the central experimental claim.
minor comments (2)
  1. [Abstract] Abstract: Specify the number and types of 3DGS models tested and the set of simulated edits to better contextualize the scope of the claimed results.
  2. [Notation] Notation: Introduce formal equations or definitions for the attribute-wise statistical profiles (means, covariances, opacities) to improve clarity and enable reproducibility.

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 validation of our method and the rigor of the experimental evaluation. We address each major comment below and commit to revisions that directly respond to the concerns raised.

read point-by-point responses

  1. Referee: [§3] §3 (Method): The assumption that attribute-wise statistical profiles augmented by editing simulations supply unique, disambiguating signatures for different transformation sequences is not validated. Multiple operations (pruning, densification, rotation, color shift) can induce statistically similar shifts in the same parameter distributions, leaving directional inferences dependent on untested LLM internal knowledge rather than provided evidence. This is load-bearing for the accuracy claim.

    Authors: We agree that explicit validation of signature uniqueness is necessary and currently insufficient in the manuscript. Section 3 describes how attribute-wise profiles compute multiple statistical moments (mean, variance, kurtosis) per parameter and how hypothesis-driven simulations generate operation-specific outcome tables (e.g., pruning reduces count while shifting opacity distributions in a characteristic way). The CoT prompt explicitly instructs the LLM to reason only from the supplied evidence. Nevertheless, the referee correctly notes that controlled tests distinguishing overlapping effects are absent. In revision we will add a new subsection with targeted experiments that isolate pairs of similar operations and quantify how often the joint evidence yields correct directional inference. revision: yes

  2. Referee: [§4] §4 (Experiments): No quantitative metrics, error analysis, dataset details, ablation studies, or baseline comparisons are described to support the reported accuracy and robustness. Without these, it is impossible to determine whether results are affected by post-hoc choices in CoT prompts or simulation parameters, undermining the central experimental claim.

    Authors: The referee is correct: the experimental section as written relies on qualitative demonstrations without the quantitative support required to substantiate the accuracy and robustness claims. We will revise §4 to include (1) dataset specifications and model counts, (2) quantitative metrics such as directional accuracy and graph edit distance against ground-truth provenance, (3) ablation studies on simulation parameters and CoT prompt variants, (4) error analysis across repeated runs, and (5) comparison against a simple statistical-matching baseline. These additions will be presented with tables and statistical significance tests. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper describes a framework that constructs provenance graphs via attribute-wise statistical profiling of 3DGS parameters combined with hypothesis-driven editing simulations, followed by LLM CoT reasoning. No equations, fitted parameters, self-definitional loops, or load-bearing self-citations appear in the abstract or described method. The central claim of accurate directed graphs is presented as validated by experiments on diverse models rather than reduced to the inputs by construction, rendering the derivation self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies insufficient technical detail to identify free parameters, axioms, or invented entities; ledger left empty.

pith-pipeline@v0.9.1-grok · 5729 in / 1045 out tokens · 18149 ms · 2026-06-27T13:37:22.699804+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

54 extracted references · 1 linked inside Pith

  1. [1]

    Langley , title =

    P. Langley , title =. Proceedings of the 17th International Conference on Machine Learning (ICML 2000) , address =. 2000 , pages =

    2000

  2. [2]

    T. M. Mitchell. The Need for Biases in Learning Generalizations. 1980

    1980

  3. [3]

    M. J. Kearns , title =

  4. [4]

    Machine Learning: An Artificial Intelligence Approach, Vol. I. 1983

    1983

  5. [5]

    R. O. Duda and P. E. Hart and D. G. Stork. Pattern Classification. 2000

    2000

  6. [6]

    Suppressed for Anonymity , author=

  7. [7]

    Newell and P

    A. Newell and P. S. Rosenbloom. Mechanisms of Skill Acquisition and the Law of Practice. Cognitive Skills and Their Acquisition. 1981

    1981

  8. [8]

    A. L. Samuel. Some Studies in Machine Learning Using the Game of Checkers. IBM Journal of Research and Development. 1959

    1959

  9. [9]

    3d gaussian splatting: Survey, technologies, challenges, and opportunities

    Bao, Y., Ding, T., Huo, J., Liu, Y., Li, Y., Li, W., Gao, Y., and Luo, J. 3d gaussian splatting: Survey, technologies, challenges, and opportunities. IEEE Transactions on Circuits and Systems for Video Technology, 2025

    2025

  10. [10]

    T., Mildenhall, B., Verbin, D., Srinivasan, P

    Barron, J. T., Mildenhall, B., Verbin, D., Srinivasan, P. P., and Hedman, P. Mip- N e RF 360: Unbounded anti-aliased neural radiance fields. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022

    2022

  11. [11]

    S URF : Speeded up robust features

    Bay, H., Tuytelaars, T., and Van Gool, L. S URF : Speeded up robust features. In European Conference on Computer Vision, 2006

    2006

  12. [12]

    U- P hylogeny: Undirected provenance graph construction in the wild

    Bharati, A., Moreira, D., Pinto, A., Brogan, J., Bowyer, K., Flynn, P., Scheirer, W., and Rocha, A. U- P hylogeny: Undirected provenance graph construction in the wild. In IEEE International Conference on Image Processing, 2017

    2017

  13. [13]

    Beyond pixels: Image provenance analysis leveraging metadata

    Bharati, A., Moreira, D., Brogan, J., Hale, P., Bowyer, K., Flynn, P., Rocha, A., and Scheirer, W. Beyond pixels: Image provenance analysis leveraging metadata. In IEEE Winter Conference on Applications of Computer Vision, 2019

    2019

  14. [14]

    J., de Rezende Rocha, A., Bowyer, K

    Bharati, A., Moreira, D., Flynn, P. J., de Rezende Rocha, A., Bowyer, K. W., and Scheirer, W. J. Transformation-aware embeddings for image provenance. IEEE Transactions on Information Forensics and Security, 2021

    2021

  15. [15]

    Butt, M. A. and Maragos, P. Optimum design of chamfer distance transforms. IEEE Transactions on Image Processing, 1998

    1998

  16. [16]

    Gaussian E ditor: Swift and controllable 3 D editing with gaussian splatting

    Chen, Y., Chen, Z., Zhang, C., Wang, F., Yang, X., Wang, Y., Cai, Z., Yang, L., Liu, H., and Lin, G. Gaussian E ditor: Swift and controllable 3 D editing with gaussian splatting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024 a

    2024

  17. [17]

    Splat F ormer: Point transformer for robust 3 D gaussian splatting

    Chen, Y., Mihajlovic, M., Chen, X., Wang, Y., Prokudin, S., and Tang, S. Splat F ormer: Point transformer for robust 3 D gaussian splatting. arXiv preprint arXiv:2411.06390, 2024 b

    arXiv 2024

  18. [18]

    Image phylogeny by minimal spanning trees

    Dias, Z., Rocha, A., and Goldenstein, S. Image phylogeny by minimal spanning trees. IEEE Transactions on Information Forensics and Security, 2011

    2011

  19. [19]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., and Houlsby, N. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020

    Pith/arXiv arXiv 2010

  20. [20]

    Trufor: Leveraging all-round clues for trustworthy image forgery detection and localization

    Guillaro, F., Cozzolino, D., Sud, A., Dufour, N., and Verdoliva, L. Trufor: Leveraging all-round clues for trustworthy image forgery detection and localization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023

    2023

  21. [21]

    G S-Checker: Tampering localization for 3D gaussian splatting

    Han, H., Luo, Z., Qi, J., Rocha, A., and Wan, R. G S-Checker: Tampering localization for 3D gaussian splatting. In Proceedings of the AAAI Conference on Artificial Intelligence, 2026

    2026

  22. [22]

    A., Holynski, A., and Kanazawa, A

    Haque, A., Tancik, M., Efros, A. A., Holynski, A., and Kanazawa, A. Instruct- NeRF2NeRF : Editing 3 D scenes with instructions. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023

    2023

  23. [23]

    Deep residual learning for image recognition

    He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016

    2016

  24. [24]

    Deep blending for free-viewpoint image-based rendering

    Hedman, P., Philip, J., Price, T., Frahm, J.-M., Drettakis, G., and Brostow, G. Deep blending for free-viewpoint image-based rendering. ACM Transactions on Graphics, 2018

    2018

  25. [25]

    Huang, G., Liu, Z., Van Der Maaten, L., and Weinberger, K. Q. Densely connected convolutional networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017

    2017

  26. [26]

    C., See, S., and Wan, R

    Huang, X., Li, R., Cheung, Y.-m., Cheung, K. C., See, S., and Wan, R. Gaussian M arker: Uncertainty-aware copyright protection of 3 D gaussian splatting. Advances in Neural Information Processing Systems, 2024

    2024

  27. [27]

    Marksplatter: Generalizable watermarking for 3d gaussian splatting model via splatter image structure

    Huang, X., Luo, Z., Song, Q., Wang, R., and Wan, R. Marksplatter: Generalizable watermarking for 3d gaussian splatting model via splatter image structure. In Proceedings of the 33rd ACM International Conference on Multimedia, 2025

    2025

  28. [28]

    I., Jang, M., Kim, J

    Jang, Y., Lee, D. I., Jang, M., Kim, J. W., Yang, F., and Kim, S. WateRF : Robust watermarks in radiance fields for protection of copyrights. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  29. [29]

    3 D-GSW : 3 D gaussian splatting for robust watermarking

    Jang, Y., Park, H., Yang, F., Ko, H., Choo, E., and Kim, S. 3 D-GSW : 3 D gaussian splatting for robust watermarking. In Proceedings of the Computer Vision and Pattern Recognition Conference, 2025

    2025

  30. [30]

    and Chang, S.-F

    Kennedy, L. and Chang, S.-F. Internet image archaeology: Automatically tracing the manipulation history of photographs on the web. In Proceedings of the ACM International Conference on Multimedia, 2008

    2008

  31. [31]

    3 D gaussian splatting for real-time radiance field rendering

    Kerbl, B., Kopanas, G., Leimk \"u hler, T., and Drettakis, G. 3 D gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 2023

    2023

  32. [32]

    Point-based neural rendering with per-view optimization

    Kopanas, G., Philip, J., Leimk \"u hler, T., and Drettakis, G. Point-based neural rendering with per-view optimization. In Computer Graphics Forum, 2021

    2021

  33. [33]

    Neural point catacaustics for novel-view synthesis of reflections

    Kopanas, G., Leimk \"u hler, T., Rainer, G., Jambon, C., and Drettakis, G. Neural point catacaustics for novel-view synthesis of reflections. ACM Transactions on Graphics, 2022

    2022

  34. [34]

    Kruskal, J. B. On the shortest spanning subtree of a graph and the traveling salesman problem. Proceedings of the American Mathematical society, 1956

    1956

  35. [35]

    Vastgaussian: Vast 3 D gaussians for large scene reconstruction

    Lin, J., Li, Z., Tang, X., Liu, J., Liu, S., Liu, J., Lu, Y., Wu, X., Xu, S., Yan, Y., et al. Vastgaussian: Vast 3 D gaussians for large scene reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  36. [36]

    Lowe, D. G. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 2004

    2004

  37. [37]

    Scaffold- GS : Structured 3 D gaussians for view-adaptive rendering

    Lu, T., Yu, M., Xu, L., Xiangli, Y., Wang, L., Lin, D., and Dai, B. Scaffold- GS : Structured 3 D gaussians for view-adaptive rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  38. [38]

    C., See, S., and Wan, R

    Luo, Z., Guo, Q., Cheung, K. C., See, S., and Wan, R. Copyrnerf: Protecting the copyright of neural radiance fields. In Proceedings of the IEEE/CVF international conference on computer vision, 2023

    2023

  39. [39]

    The nerf signature: Codebook-aided watermarking for neural radiance fields

    Luo, Z., Rocha, A., Shi, B., Guo, Q., Li, H., and Wan, R. The nerf signature: Codebook-aided watermarking for neural radiance fields. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    2025

  40. [40]

    D EG auss: Defending against malicious 3 D editing for gaussian splatting

    Meng, L., Shao, M., Qiao, Y., and Lv, X. D EG auss: Defending against malicious 3 D editing for gaussian splatting. Advances in Neural Information Processing Systems, 2025

    2025

  41. [41]

    P., Tancik, M., Barron, J

    Mildenhall, B., Srinivasan, P. P., Tancik, M., Barron, J. T., Ramamoorthi, R., and Ng, R. Nerf: Representing scenes as neural radiance fields for view synthesis. In European Conference on Computer Vision, 2020

    2020

  42. [42]

    W., Flynn, P

    Moreira, D., Bharati, A., Brogan, J., Pinto, A., Parowski, M., Bowyer, K. W., Flynn, P. J., Rocha, A., and Scheirer, W. J. Image provenance analysis at scale. IEEE Transactions on Image Processing, 2018

    2018

  43. [43]

    C., See, S., and Wan, R

    Song, Q., Luo, Z., Cheung, K. C., See, S., and Wan, R. Geometry C loak: Preventing tgs-based 3 D reconstruction from copyrighted images. Advances in Neural Information Processing Systems, 2024 a

    2024

  44. [44]

    C., See, S., and Wan, R

    Song, Q., Luo, Z., Cheung, K. C., See, S., and Wan, R. Protecting nerfs’ copyright via plug-and-play watermarking base model. In European Conference on Computer Vision, 2024 b

    2024

  45. [45]

    C., See, S., and Wan, R

    Song, Q., Luo, Z., Cheung, K. C., See, S., and Wan, R. Align 3d representation and text embedding for 3d content personalization. In Proceedings of the 33rd ACM International Conference on Multimedia, 2025

    2025

  46. [46]

    V., Zhou, D., et al

    Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q. V., Zhou, D., et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 2022

    2022

  47. [47]

    Physgaussian: Physics-integrated 3 D gaussians for generative dynamics

    Xie, T., Zong, Z., Qiu, Y., Li, X., Feng, Y., Yang, Y., and Jiang, C. Physgaussian: Physics-integrated 3 D gaussians for generative dynamics. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  48. [48]

    Deformable 3 D gaussians for high-fidelity monocular dynamic scene reconstruction

    Yang, Z., Gao, X., Zhou, W., Jiao, S., Zhang, Y., and Jin, X. Deformable 3 D gaussians for high-fidelity monocular dynamic scene reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  49. [49]

    Differentiable surface splatting for point-based geometry processing

    Yifan, W., Serena, F., Wu, S., \"O ztireli, C., and Sorkine-Hornung, O. Differentiable surface splatting for point-based geometry processing. ACM Transactions On Graphics, 2019

    2019

  50. [50]

    Mip- S platting: Alias-free 3 D gaussian splatting

    Yu, Z., Chen, A., Huang, B., Sattler, T., and Geiger, A. Mip- S platting: Alias-free 3 D gaussian splatting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  51. [51]

    Image provenance analysis via graph encoding with vision transformer

    Zhang, K., Kong, C., Wang, S., Rocha, A., and Li, H. Image provenance analysis via graph encoding with vision transformer. IEEE Transactions on Information Forensics and Security, 2025

    2025

  52. [52]

    H., Karaman, S., and Chang, S.-F

    Zhang, X., Sun, Z. H., Karaman, S., and Chang, S.-F. Discovering image manipulation history by pairwise relation and forensics tools. IEEE Journal of Selected Topics in Signal Processing, 2020

    2020

  53. [53]

    GS-H ider: Hiding messages into 3 D gaussian splatting

    Zhang, X., Meng, J., Li, R., Xu, Z., Zhang, Y., and Zhang, J. GS-H ider: Hiding messages into 3 D gaussian splatting. Advances in Neural Information Processing Systems, 2024

    2024

  54. [54]

    Triplane meets gaussian splatting: Fast and generalizable single-view 3d reconstruction with transformers

    Zou, Z.-X., Yu, Z., Guo, Y.-C., Li, Y., Liang, D., Cao, Y.-P., and Zhang, S.-H. Triplane meets gaussian splatting: Fast and generalizable single-view 3d reconstruction with transformers. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024

    2024