Pith. sign in

REVIEW 5 major objections 7 minor 43 references

Exploration Matters for Escaping the Blur Trap in 3D Gaussian Splatting

T0 review · 5 major / 7 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read 3D Gaussian Splatting's 'Blur Trap' is a gradient bias, and two random operators—Random Seeding and Random Splitting—escape it, improving rendering across five datasets.

desk verdict A correct orthogonality proof plus two dead-simple exploration operators that improve 3DGS on most benchmarks; the central mechanism is plausible but the empirical base needs error bars and the missing MCMC-style baselines make the claim of 'fundamental' a bit strong. read the letter →

arxiv 2607.17965 v1 pith:CM5PORNR submitted 2026-07-20 cs.CV

Chengbo Wang , Guozheng Ma , Jinhong Wu , Tie Ji , Yizhen Lao This is my paper · ORCID
classification cs.CV
keywords 3DGaussianSplattingBlurTrapgradientorthogonalityalpha-blendingattenuationadaptivedensificationstochasticexplorationRandomSeedingSplitting
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

The paper claims that persistent blur in 3D Gaussian Splatting reconstructions—in distant regions and in occluded near-field areas, even with abundant training views—is not a data or capacity problem but a systematic optimization bias it names the Blur Trap. The bias has two geometric sources: the dominant backpropagated gradient (the 2D projected-position branch) is proven strictly orthogonal to the viewing ray, so it can never push Gaussians along depth; and alpha-blending multiplies each primitive's gradient by a rapidly decaying transmittance, so occluded primitives rarely accumulate enough signal to trigger adaptive densification. The paper shows that injecting two deliberately minimal random operators—Random Seeding (new Gaussians at random 3D positions) and Random Splitting (splitting large Gaussians regardless of gradient)—bypasses both biases and improves rendering fidelity across Mip-NeRF 360, Tanks & Temples, Deep Blending, OMMO, and DL3DV with negligible overhead. The result reframes a common failure mode of a widely used renderer as a missing-exploration problem rather than a capacity or data limit, and both fixes are simple enough to drop into existing pipelines.

What carries the argument

The load-bearing result is the orthogonality identity (Eq. 16) for the 2D-position gradient branch, proved by factoring the chain rule through perspective division—the homogeneous coordinate vector lies in the null space of the perspective-division Jacobian—and through the viewport transform, whose Jacobian annihilates the remaining z-component. The second mechanism is the transmittance bound on the per-Gaussian 2D gradient: |∂L/∂α_m| ≤ T_{m-1}, monotonically decaying with depth-sorted index, which explains why occluded primitives rarely split. The proposed remedies are Random Seeding (uniformly sampling new Gaussian positions inside the bounding box, which bypasses the depth-orthogonality c

What would settle it

Train 3DGS on a standard benchmark scene and log the ratio ‖g_2d‖/‖g_cov2d‖ at every iteration, and separately compute the dot product (∂L_2D/∂P_3D)·(P_3D − P_cam) for a sample of Gaussians. If the covariance branch ever comes within an order of magnitude of the 2D branch, or if any nonzero depth-directed 2D-position gradient is measured, the core claim is falsified. A second check: run the same pipeline under a fisheye camera model; if far-side blur disappears without random seeding, the orthogonality theorem is specific to pinhole projection and the claimed bias is not intrinsic to 3DGS.

Watch

Extended reading notes

Core claim

The central discovery is an identity and an attenuation mechanism. For any Gaussian, the 3D position gradient derived from the 2D projected-position branch is strictly orthogonal to the camera-to-Gaussian ray: ∂L_2D/∂P_3D · (P_3D − P_cam) = 0. Because this branch dominates the total position gradient by two to three orders of magnitude, the optimizer effectively cannot move primitives along depth, leaving distant regions underdetermined (the Far-Side Blur Trap). At the same time, alpha-blending's compositing equation bounds each depth-sorted Gaussian's gradient magnitude by the transmittance accumulated from all Gaussians in front of it, so occluded primitives receive signals too weak to cro

Load-bearing premise

The argument rests on the measured claim that the screen-space position gradient is vastly larger than the other two gradient branches; if those branches ever contribute comparably, the optimizer could already move primitives along depth and the Far-Side Blur Trap would be much weaker.

Editorial extensions

If this is right

  • Any 3DGS variant that relies only on 2D reprojection gradients will keep the far-side depth deficiency; recovering distant geometry requires a non-gradient mechanism (seeding, depth priors, or a different projection model).
  • Any densification criterion based solely on accumulated 2D gradient magnitude will under-densify occluded regions because transmittance bounds the gradient; random splitting restores density there, often with fewer total primitives.
  • Random Seeding and Random Splitting are complementary, and their combination gives the best or near-best metrics across the five datasets, supporting the claim that the two trap subtypes are distinct.
  • The same diagnosis transfers to 4D Gaussian Splatting: random splitting alone (N_split=5) substantially improves perceptual fidelity on Neu3D, so the bias is inherited by splatting-based dynamic renderers.
  • Lowering the split threshold to mimic random splitting does not produce the same quality and inflates primitive counts, indicating the gain comes from targeted exploration rather than brute-force densification.

Reading between the lines

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

  • An extension the paper leaves implicit: because the orthogonality proof uses the pinhole projection model, non-pinhole cameras (fisheye, catadioptric) may create depth-aligned 2D-gradient components and thereby weaken the Far-Side Blur Trap without any seeding; this is testable with a standard dataset.
  • Random Seeding is a coverage-driven explorer in the spirit of count-based reinforcement learning; a learned or novelty-weighted seed proposal could be far more sample-efficient than uniform sampling inside the Gaussian bounding box.
  • The paper's depth-regularization ablation suggests depth priors and random seeding are partly substitutable as depth sources; a hybrid that applies pseudo-depth regularization only where the prior is confident might beat either alone.
  • The transmittance-bound argument applies to any alpha-composited splatting or particle renderer, so the random-splitting fix may transfer beyond Gaussian primitives to surfel or point renderers.
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

5 major / 7 minor

Summary. The paper argues that standard 3D Gaussian Splatting optimization is prone to a 'Blur Trap' caused by gradient biases. In §3.1 it claims that the 2D screen-space position gradient branch dominates the other positional gradient branches by two to three orders of magnitude, reducing the 3D position update to a 2D reprojection error minimization. §3.2 and Appendix A prove that this g2d branch is orthogonal to the camera-to-Gaussian viewing ray (Eq. 16), implying no depth-directed signal from that branch; Appendix B shows that alpha-blending attenuates the densification gradient of occluded primitives by a transmittance envelope. The authors categorize the resulting artifacts into Far-Side and Near-Side Blur Traps and propose two minimal operators—Random Seeding and Random Splitting—to bypass gradient-based optimization. They report improved PSNR/SSIM/LPIPS on Mip-NeRF 360, Tanks & Temples, Deep Blending, OMMO, and DL3DV, plus a 4DGS extension.

Significance. The orthogonality proof in Appendix A is clean and is a useful formal observation about the g2d branch; the transmittance bound in Appendix B is also correct as a per-pixel bound. The proposed operators are refreshingly simple, and the experiments show consistent, if modest, improvements across five datasets and a transfer to 4DGS. The paper ships falsifiable predictions and gives a mechanistic explanation, which are clear strengths. However, the broad conclusion that 3DGS 'cannot' receive depth-directed optimization signals is established only for the g2d branch; the paper's own Table 2 concedes that the covariance branch has depth-aligned components. The dismissal of that branch rests on an unquantified empirical magnitude claim. The absence of comparisons to the two closest stochastic baselines, 3DGS-MCMC and Opt3DGS, also tempers the significance. If the magnitude dominance is properly established, this would be a valuable diagnostic and a strong argument for explicit exploration.

major comments (5)
  1. [Section 3.1, Figure 2, Table 1, Table 2] The load-bearing premise of the Far-Side Blur Trap is the claimed 2–3 order-of-magnitude dominance of ||g2d|| over ||gcov2d|| and ||gsh||. This is only shown in a qualitative curve (Figure 2) from the authors' own runs, with no error bars, no multi-seed statistics, and no per-dataset or per-training-phase ratio. Table 2 explicitly states that gcov2d has 'more random direction' including depth-aligned components. Since Appendix A's orthogonality theorem (Eq. 16) applies only to g2d, the total 3D position gradient can still contain a depth-directed component if the covariance branch is not always negligible. The paper needs to report the ratio ||gcov2d||/||g2d|| over training phases and datasets, with multiple seeds, or otherwise bound it, before claiming a fundamental depth-gradient deficiency.
  2. [Section 5.2, Table 3] The experimental validation omits the two closest exploration baselines, 3DGS-MCMC [8] and Opt3DGS [29], both of which are discussed in Related Work. Since the paper's central claim is that explicit exploration is what 3DGS is missing, a comparison against these methods is necessary to show that the proposed operators are competitive or complementary. Without these numbers, Table 3 supports only 'better than 3DGS and HoGS' rather than the broader exploration claim.
  3. [Section 5.2, Tables 3 and 5] No error bars or multi-seed results are reported for any metric. The reported gains are often small (e.g., Mip-NeRF360 PSNR 27.52→27.96; DL3DV PSNR 27.16→28.43), and some combined results are slightly worse than the single operator (e.g., OMMO PSNR 31.27 vs 31.29; DB LPIPS 0.249 vs 0.248). The claim of consistent and complementary improvement therefore needs variance estimates or significance testing.
  4. [Section 3.2.1, Eq. (2), Appendix A.3] The statement 'We prove that the 3D positional update direction of any Gaussian primitive remains strictly orthogonal to the viewing ray' is stronger than what Appendix A proves. Eq. (16) shows orthogonality for the g2d branch only. The text should consistently qualify the theorem to the 2D-position branch and state explicitly that the total-gradient conclusion depends on the empirical dominance of that branch. As written, the paper risks overclaiming a mathematical impossibility where the actual result is a conditional, empirical bias.
  5. [Appendix B, Section 3.2.2] The transmittance bound in Eq. (21) is correct as a per-pixel, per-step bound, but it does not by itself establish densification failure, because gradients are accumulated over many pixels and many optimization steps, and the color-difference factor c_m - c_hat_{m+1:N} can be large. The conclusion that rear Gaussians are 'unable to surpass the densification threshold' is stronger than what the bound proves. The paper should either add an accumulation argument or rely primarily on the empirical profiling in Figure 5 rather than presenting the bound as the complete mechanism.
minor comments (7)
  1. [Throughout] Typos: 'prove' should be 'proof' in Appendix headings; 'dirsections' (Sec. 3.2.1); 'the burden of prove' (Sec. 1); 'Experientially' in Table 2; 'circled around the Blur Trap' (Sec. 1) is informal.
  2. [Figure 2] The axis labels and tick marks are garbled in the rendered figure; please provide a clean version with readable numeric scales so the claimed 2–3 orders of magnitude can be visually verified.
  3. [Section 5.1.2] The DL3DV subset is described as 'randomly select a subset'; no random seed or selection protocol is given, which hampers reproducibility.
  4. [Appendix C] The positional perturbation exploration is only illustrated qualitatively; adding the corresponding PSNR/SSIM/LPIPS numbers would make the comparison meaningful.
  5. [Table 4] The row 'Seed & Split Exp.' has values '2.53 2.11 0.79 1.77 2.11' without column header alignment; clarify which numbers correspond to which dataset.
  6. [Section 4.2] Random Seeding samples uniformly within the minimum bounding box of all Gaussians. In unbounded scenes this box can be very large; specify how seed positions are actually sampled (e.g., normalized volume, near-surface, or view frustum) and whether this choice affects the observed far-side improvements.
  7. [Section 6] The conclusion claims 'state-of-the-art fidelity with negligible overhead,' but Table 3 does not compare with any SOTA method besides HoGS, and no training-time or memory measurements are reported. Please soften or substantiate.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity: the central orthogonality theorem is derived self-containedly from the projection model; the only self-citation is a minor, non-load-bearing attribution in Appendix B.

  1. self citation load bearing [Appendix B.3, Eq. (20), derivation of ∂ĉ/∂α_m]
    "Differentiating Equation 19 w.r.t. α_m [33]: ... ∂ĉ/∂α_m = T_{m−1}·(c_m − ĉ_{m+1:N})."

    The derivative formula is attributed to the authors' own prior paper [33], but the same appendix derives it in full from the alpha-blending equation (Eqs. 19–20). The self-citation is therefore not load-bearing: the transmittance-attenuation conclusion (Eq. 21) is proven from the pipeline equations and does not depend on an unverified external result. This is a minor self-citation, not a reduction of the central claim to its inputs.

full rationale

The paper's load-bearing chain is: (i) the 3D position gradient decomposes into g2d, gcov2d, gsh; (ii) g2d dominates empirically; (iii) g2d is orthogonal to the camera-to-Gaussian ray (Eq. 16); (iv) alpha-blending attenuates densification gradients for occluded primitives (Eqs. 20–21); and (v) random seeding/splitting bypass these biases and improve held-out benchmarks. Step (iii) is derived self-contained in Appendix A from the projection and perspective-division Jacobians (Eqs. 5–15) and does not assume the target result; it is mathematically independent of the empirical dominance claim. Step (iv) is derived in Appendix B from the alpha-blending equation; the only self-citation [33] for one differentiation step is accompanied by a full derivation in the same appendix, so it is not load-bearing. The dominance claim (Figure 2, Table 1) is an empirical measurement without error bars or cross-dataset bounds, which is a potential correctness/robustness risk rather than circularity, because the paper does not fit parameters to the evaluation benchmarks and then call that a prediction. Random Seeding and Random Splitting are simple heuristics with fixed hyperparameters evaluated on held-out datasets; they are not obtained by inverting the evaluation metrics. Thus the derivation chain is self-contained and the central claims do not reduce to their inputs by construction.

Assumptions & free parameters 3 free parameters · 5 assumptions · 1 invented entities

No physical entities or fitted parameters beyond standard 3DGS hyperparameters. The central theoretical results are self-contained derivations; the method introduces two hand-set exploration intensities (N_seed, N_split). The Blur Trap is a conceptual label, not a new physical object.

free parameters (3)
  • N_seed = 20 (default)
    Number of random Gaussian seeds injected per densification iteration; hand-chosen, not optimized; controls far-side exploration intensity.
  • N_split = 20 (default), 5 for 4DGS
    Number of large-scale Gaussians randomly split per iteration; hand-chosen, not optimized; controls near-side exploration intensity.
  • seed sampling region = minimum bounding box of existing Gaussians
    Choice of sampling region for Random Seeding; affects which depth intervals are probed; no adaptive scheme.
assumptions (5)
  • domain assumption The 2D position gradient branch g2d dominates the total 3D position gradient by 2–3 orders of magnitude across the whole training process.
    Empirical claim from Section 3.1 (Figure 2, Table 1); if false, the covariance branch could carry depth-directed signal and the Far-Side Blur Trap would be less severe.
  • domain assumption 3DGS training uses deterministic, gradient-dominated updates with minibatch size one and no implicit regularization.
    Section 1 and 3.1; standard 3DGS setup, used to argue exploration is missing.
  • standard math The alpha-blending compositing model (Eq. 3/17) accurately describes the rendering and backpropagation of 3DGS.
    Standard 3DGS forward model [1]; used for the gradient attenuation proof in Appendix B.
  • standard math Perspective projection and viewport transformation as defined in Appendix A.1 are the correct differentiable rendering model for the 2D position path.
    Orthogonality proof rests on this projection model.
  • domain assumption Pseudo-depth maps from Depth Anything V2 are unreliable in distant regions.
    Used in Section 5.4 ablation to argue random seeding beats depth supervision; not central to main claim.
invented entities (1)
  • Blur Trap (conceptual construct) independent evidence
    purpose: Explanatory label for systematic blur in distant and occluded regions of 3DGS reconstructions.
    It predicts that bypassing gradient orthogonality and the densification gate should recover detail; this is tested in Tables 3/5 and Figures 8–9. It is a term, not a physical entity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploration Matters for Escaping the Blur Trap in 3D Gaussian Splatting." pith.science (2026). https://pith.science/paper/CM5PORNR

@misc{pith2026260717965,
  author       = {Pith},
  title        = {Pith review of: Exploration Matters for Escaping the Blur Trap in 3D Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CM5PORNR}},
  note         = {Machine review of arXiv:2607.17965}
}
read the original abstract

3D Gaussian Splatting (3DGS) employs Gaussian primitives for explicit scene representation, facilitating real-time, high-fidelity reconstruction and novel view synthesis of complex scenes. However, the explicit modeling inherent in 3DGS introduces a gradient bias during optimization, rendering its non-convex optimization process highly susceptible to convergence toward local suboptimal solutions. This constitutes a fundamental limitation in 3DGS optimization, which we term the Blur Trap. To address this limitation, we integrate simple explicit exploration into the 3DGS optimization framework. First, through rigorous mathematical analysis of the 3DGS optimization formulation, we identify the underlying optimization bias responsible for the Blur Trap and categorize it into two distinct subtypes: the Far-Side Blur Trap and the Near-Side Blur Trap. Subsequently, we propose two highly straightforward exploration strategies (Random Seeding and Random Splitting) to mitigate the far-side and near-side blur traps, respectively. Experimental validation demonstrates that the incorporation of these exploration operators effectively and complementarily overcome the Blur Trap, achieving high-quality rendering performance across multiple datasets. Project page: https://chengbo-wang.github.io/ExploreGS/

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 2 linked inside Pith

  1. [8]

    3d gaus- sian splatting as markov chain monte carlo.Advances in Neural Information Processing Systems, 37:80965–80986,

    Shakiba Kheradmand, Daniel Rebain, Gopal Sharma, Weiwei Sun, Yang-Che Tseng, Hossam Isack, Abhishek Kar, Andrea Tagliasacchi, and Kwang Moo Yi. 3d gaus- sian splatting as markov chain monte carlo.Advances in Neural Information Processing Systems, 37:80965–80986,

  2. [29]

    Opt3dgs: Optimizing3dgaussiansplattingwithadaptive exploration and curvature-aware exploitation

    Ziyang Huang, Jiagang Chen, Jin Liu, and Shunping Ji. Opt3dgs: Optimizing3dgaussiansplattingwithadaptive exploration and curvature-aware exploitation. InPro- ceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 5230–5238, 2026. 3

    2026

  3. [1]

    3d gaussian splatting for real- time radiance field rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, George Drettakis, et al. 3d gaussian splatting for real- time radiance field rendering.ACM Trans. Graph., 42 (4):139–1, 2023. 1, 3, 4

    2023

  4. [2]

    Nerf: Representing scenes as neural radiance fields for view synthesis.Communications of the ACM, 65(1):99– 106, 2021

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis.Communications of the ACM, 65(1):99– 106, 2021. 1

    2021

  5. [3]

    Absgs: Recovering fine details in 3d gaussian splat- ting

    Zongxin Ye, Wenyu Li, Sidun Liu, Peng Qiao, and Yong Dou. Absgs: Recovering fine details in 3d gaussian splat- ting. InProceedings of the 32nd ACM international con- ference on multimedia, pages 1053–1061, 2024. 2, 3

    2024

  6. [4]

    Pixel-gs: Density control with pixel- aware gradient for 3d gaussian splatting

    Zheng Zhang, Wenbo Hu, Yixing Lao, Tong He, and Hengshuang Zhao. Pixel-gs: Density control with pixel- aware gradient for 3d gaussian splatting. InEuropean Conference on Computer Vision, pages 326–342. Springer,

  7. [5]

    Taming 3dgs: High-quality radiance fields with limited resources

    Saswat Subhajyoti Mallick, Rahul Goel, Bernhard Kerbl, Markus Steinberger, Francisco Vicente Carrasco, and Fer- nando De La Torre. Taming 3dgs: High-quality radiance fields with limited resources. InSIGGRAPH Asia 2024 Conference Papers, pages 1–11, 2024. 2

    2024

  8. [6]

    Mini-splatting: Repre- senting scenes with a constrained number of gaussians

    Guangchi Fang and Bing Wang. Mini-splatting: Repre- senting scenes with a constrained number of gaussians. InEuropean conference on computer vision, pages 165–

Show all 43 references
  1. [7]

    Hogs: Unifiednearandfarobjectreconstruc- tion via homogeneous gaussian splatting

    Xinpeng Liu, Zeyi Huang, Fumio Okura, and Yasuyuki Matsushita. Hogs: Unifiednearandfarobjectreconstruc- tion via homogeneous gaussian splatting. InProceedings oftheComputerVisionandPatternRecognitionConference, pages 26714–26722, 2025. 2, 3, 5

    2025

  2. [9]

    Bayesian learning via stochastic gradient langevin dynamics

    Max Welling and Yee W Teh. Bayesian learning via stochastic gradient langevin dynamics. InProceedings of the 28th international conference on machine learning (ICML-11), pages 681–688, 2011. 2, 3

    2011

  3. [10]

    How to escape saddle points effi- ciently

    Chi Jin, Rong Ge, Praneeth Netrapalli, Sham M Kakade, and Michael I Jordan. How to escape saddle points effi- ciently. InInternational conference on machine learning, pages 1724–1732. PMLR, 2017. 2, 3

    2017

  4. [11]

    Escap- ing from saddle points—online stochastic gradient for tensor decomposition

    Rong Ge, Furong Huang, Chi Jin, and Yang Yuan. Escap- ing from saddle points—online stochastic gradient for tensor decomposition. InConference on learning theory, pages 797–842. PMLR, 2015. 2, 3

    2015

  5. [12]

    Mip-nerf 360: Un- bounded anti-aliased neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Mip-nerf 360: Un- bounded anti-aliased neural radiance fields. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5470–5479, 2022. 2, 4, 8

    2022

  6. [13]

    Tanks and temples: Benchmarking large-scale scene reconstruction.ACM Transactions on Graphics (ToG), 36(4):1–13, 2017

    Arno Knapitsch, Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun. Tanks and temples: Benchmarking large-scale scene reconstruction.ACM Transactions on Graphics (ToG), 36(4):1–13, 2017. 2, 4, 8

    2017

  7. [14]

    Deep blendingforfree-viewpointimage-basedrendering.ACM Transactions on Graphics (ToG), 37(6):1–15, 2018

    Peter Hedman, Julien Philip, True Price, Jan-Michael Frahm, George Drettakis, and Gabriel Brostow. Deep blendingforfree-viewpointimage-basedrendering.ACM Transactions on Graphics (ToG), 37(6):1–15, 2018. 2, 4, 8

    2018

  8. [15]

    A large-scale outdoor multi- modal dataset and benchmark for novel view synthesis and implicit scene reconstruction

    Chongshan Lu, Fukun Yin, Xin Chen, Wen Liu, Tao Chen, Gang Yu, and Jiayuan Fan. A large-scale outdoor multi- modal dataset and benchmark for novel view synthesis and implicit scene reconstruction. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages...

    2023

  9. [16]

    Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision

    Lu Ling, Yichen Sheng, Zhi Tu, Wentian Zhao, Cheng Xin, Kun Wan, Lantao Yu, Qianyu Guo, Zixun Yu, Yawen Lu, et al. Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pag...

    2024

  10. [17]

    Gradient-direction- aware density control for 3d gaussian splatting

    Zheng Zhou, Yu-Jie Xiong, Jia-Chen Zhang, Chun-Ming Xia, Xihe Qiu, and Hongjian Zhan. Gradient-direction- aware density control for 3d gaussian splatting. InThe Fourteenth International Conference on Learning Repre- sentations, 2026. URL https://openreview.net/ forum?id=6qDxK4Gz7F. 3

    2026

  11. [18]

    Revising densification in gaussian splatting

    SamuelRotaBulò,LorenzoPorzi,andPeterKontschieder. Revising densification in gaussian splatting. InEuropean Conference on Computer Vision, pages 347–362. Springer,

  12. [19]

    Scaffold-gs: Structured 3d gaussians for view-adaptive rendering

    Tao Lu, Mulin Yu, Linning Xu, Yuanbo Xiangli, Limin Wang, Dahua Lin, and Bo Dai. Scaffold-gs: Structured 3d gaussians for view-adaptive rendering. InProceedings oftheIEEE/CVFconferenceoncomputervisionandpattern recognition, pages 20654–20664, 2024. 3

    2024

  13. [20]

    Proxy-gs: Unified occlusion priors for train- ing and inference in structured 3d gaussian splatting

    Yuanyuan Gao, Yuning Gong, Yifei Liu, Jingfeng Li, Dan Xu, Yanci Zhang, Dingwen Zhang, Xiao Sun, and Zhi- hang Zhong. Proxy-gs: Unified occlusion priors for train- ing and inference in structured 3d gaussian splatting. InProceedings of the IEEE/CVF Conference on Computer 14 Ex...

    2026

  14. [21]

    Optimization by simulated annealing.science, 220 (4598):671–680, 1983

    Scott Kirkpatrick, C Daniel Gelatt Jr, and Mario P Vec- chi. Optimization by simulated annealing.science, 220 (4598):671–680, 1983. 3

    1983

  15. [22]

    MIT press Cam- bridge, 1998

    Richard S Sutton, Andrew G Barto, et al.Reinforcement learning: An introduction, volume 1. MIT press Cam- bridge, 1998. 3

    1998

  16. [23]

    Pa- rameter space noise for exploration.arXiv preprint arXiv:1706.01905, 2017

    Matthias Plappert, Rein Houthooft, Prafulla Dhariwal, Szymon Sidor, Richard Y Chen, Xi Chen, Tamim As- four, Pieter Abbeel, and Marcin Andrychowicz. Pa- rameter space noise for exploration.arXiv preprint arXiv:1706.01905, 2017. 3

    2017 arXiv

  17. [24]

    Noisy networks for exploration

    Meire Fortunato, Mohammad Gheshlaghi Azar, Bilal Piot, Jacob Menick, Matteo Hessel, Ian Osband, Alex Graves, Volodymyr Mnih, Remi Munos, Demis Hass- abis, Olivier Pietquin, Charles Blundell, and Shane Legg. Noisy networks for exploration. InInternational Con- ference on Learni...

    2018

  18. [25]

    Unifying count-based exploration and intrinsic motivation.Ad- vances in neural information processing systems, 29, 2016

    Marc Bellemare, Sriram Srinivasan, Georg Ostrovski, Tom Schaul, David Saxton, and Remi Munos. Unifying count-based exploration and intrinsic motivation.Ad- vances in neural information processing systems, 29, 2016. 3

    2016

  19. [26]

    Curiosity-driven exploration by self- supervised prediction

    Deepak Pathak, Pulkit Agrawal, Alexei A Efros, and Trevor Darrell. Curiosity-driven exploration by self- supervised prediction. InInternational conference on machine learning, pages 2778–2787. PMLR, 2017. 3

    2017

  20. [27]

    Exploration by random network distillation

    Yuri Burda, Harrison Edwards, Amos Storkey, and Oleg Klimov. Exploration by random network distillation. arXiv preprint arXiv:1810.12894, 2018. 3

    2018 arXiv

  21. [28]

    Deep exploration via randomized value functions.Journal of machine learning research, 20(124): 1–62, 2019

    Ian Osband, Benjamin Van Roy, Daniel J Russo, and Zheng Wen. Deep exploration via randomized value functions.Journal of machine learning research, 20(124): 1–62, 2019. 3

    2019

  22. [30]

    Edgs: Eliminating densification for efficient convergence of 3dgs

    Dmytro Kotovenko, Olga Grebenkova, and Björn Ommer. Edgs: Eliminating densification for efficient convergence of 3dgs. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 41065– 41076, 2026. 4

    2026

  23. [31]

    3dgs2: Near second-order converg- ing 3d gaussian splatting

    Lei Lan, Tianjia Shao, Zixuan Lu, Yu Zhang, Chenfanfu Jiang, and Yin Yang. 3dgs2: Near second-order converg- ing 3d gaussian splatting. InProceedings of the Special Interest Group on Computer Graphics and Interactive Tech- niques Conference Conference Papers, pages 1–10, 2025. 4

    2025

  24. [32]

    Seele: A unified acceleration framework for real-time gaussian splatting on mobile devices

    He Zhu, Xiaotong Huang, Zihan Liu, Weikai Lin, Xiao- hong Liu, Zhezhi He, Jingwen Leng, Minyi Guo, and Yu Feng. Seele: A unified acceleration framework for real-time gaussian splatting on mobile devices. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern...

    2026

  25. [33]

    Faster and better 3d splatting via group training

    Chengbo Wang, Guozheng Ma, Yifei Xue, and Yizhen Lao. Faster and better 3d splatting via group training. InProceedings of the IEEE/CVF International Conference on Computer Vision, 2025. 6, 20

    2025

  26. [34]

    Speeding up the learning of 3d gaussians with much shorter gaussian lists

    Jiaqi Liu and Zhizhong Han. Speeding up the learning of 3d gaussians with much shorter gaussian lists. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1231–1240, 2026. 6

    2026

  27. [35]

    Rade-gs: Raster- izing depth in gaussian splatting.ACM Transactions on Graphics, 45(2):1–14, 2026

    Baowen Zhang, Chuan Fang, Rakesh Shrestha, Yixun Liang, Xiao-Xiao Long, and Ping Tan. Rade-gs: Raster- izing depth in gaussian splatting.ACM Transactions on Graphics, 45(2):1–14, 2026. 6

    2026

  28. [36]

    Pgsr: Planar-based gaussian splat- ting for efficient and high-fidelity surface reconstruction

    Danpeng Chen, Hai Li, Weicai Ye, Yifan Wang, Weijian Xie, Shangjin Zhai, Nan Wang, Haomin Liu, Hujun Bao, and Guofeng Zhang. Pgsr: Planar-based gaussian splat- ting for efficient and high-fidelity surface reconstruction. IEEE Transactions on Visualization and Computer Graph- i...

    2024

  29. [37]

    4d gaussian splatting for real-time dynamic scene rendering

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xi- aopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xing- gang Wang. 4d gaussian splatting for real-time dynamic scene rendering. InProceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition, pages 20310–2...

    2024

  30. [38]

    Neural 3d video synthesis from multi-view video

    Tianye Li, Mira Slavcheva, Michael Zollhoefer, Simon Green, Christoph Lassner, Changil Kim, Tanner Schmidt, Steven Lovegrove, Michael Goesele, Richard Newcombe, et al. Neural 3d video synthesis from multi-view video. InProceedings of the IEEE/CVF conference on computer vision ...

    2022

  31. [39]

    Depth anything v2.Advances in Neural Information Processing Systems, 37:21875–21911, 2024

    Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xi- aogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything v2.Advances in Neural Information Processing Systems, 37:21875–21911, 2024. 10 15 Exploration Matters for Escaping the Blur Trap in 3D Gaussian Splatting Appendix A...

    2024

  32. [40]

    The view matrixVmaps world coordinates to camera space:[pT cam,1] T =V[P T 3D,1] T

  33. [41]

    The projection matrixPmaps camera coordinates to homogeneous clip space:ℎp=P[p T cam,1] T

  34. [42]

    Perspective division yields Normalized Device Coordinates (NDC):pndc = [ 𝑝𝑥 𝑝𝑤 , 𝑝𝑦 𝑝𝑤 , 𝑝𝑧 𝑝𝑤 ]T

  35. [43]

    The composite matrixM=PVencodes the combined camera–projection mapping

    The viewport transformation with image dimensions𝑊×𝐻produces 2D pixel coordinatesp2D. The composite matrixM=PVencodes the combined camera–projection mapping. These symbols are summarized below. 16 Exploration Matters for Escaping the Blur Trap in 3D Gaussian Splatting P3D = [𝑥...

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.