Research map / 3D vision / Gaussian splatting
Compact and efficient Gaussian splatting
Prunes, sparsifies and optimizes Gaussian primitives for faster, smaller 3DGS rendering.
269 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2024–2026) · an approach in Gaussian splatting, keyword cluster “3d gaussian · gaussian splatting · splatting 3dgs”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 3 of 6 tracked benchmarks.
29 papers in 2023–24 and 240 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 23.8% to 33.4%.
Open questions the data points to
- Mip-NeRF 360 (PSNR): best went from 28.01 (up to 2024) to 29.5 in 2025 — Topology-Aware 3D Gaussian Splatting: Leveraging Persistent Homology for Optimized Structural Integrity (AAAI 2025 · 0 citations)
- Mip-NeRF360 (SSIM): best went from 0.833 (up to 2024) to 0.874 in 2025 — Topology-Aware 3D Gaussian Splatting: Leveraging Persistent Homology for Optimized Structural Integrity (AAAI 2025 · 0 citations)
- Mip-NeRF360 (LPIPS): best went from 0.189 (up to 2024) to 0.172 in 2025 — Perceptual-GS: Scene-adaptive Perceptual Densification for Gaussian Splatting (ICML 2025 · 0 citations)
- Papers report 3DGS on Mip-NeRF 360 (PSNR) anywhere from 16.06 to 27.72: comparisons against it depend on whose number you take.
- Papers report 3DGS on Mip-NeRF 360 (SSIM) anywhere from 0.4 to 0.815: comparisons against it depend on whose number you take.
What looks settled
- Deep Blending (PSNR): best up to 2024 30.42, best in 2025–26 30.45 from 9 papers, at most 0.2% better, if at all.
- Tanks&Temples (PSNR): best up to 2024 24.96, best in 2025–26 24.27 from 8 papers, at most 0.2% better, if at all.
- Deep Blending (SSIM): best up to 2024 0.907, best in 2025–26 0.911 from 7 papers, less than 1% better.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2024 · Mip-Splatting: Alias-free 3D Gaussian Splatting (CVPR 2024 · 524 citations)
- 2025 · DepthSplat: Connecting Gaussian Splatting and Depth (CVPR 2025 · 80 citations)
- 2026 · SkySplat: Generalizable 3D Gaussian Splatting from Multi-Temporal Sparse Satellite Images (AAAI 2026 · 9 citations)
Earliest papers here
- Mip-Splatting: Alias-free 3D Gaussian Splatting (CVPR 2024 · 524 citations)
- SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering (CVPR 2024 · 478 citations)
- Gaussian Splatting SLAM (CVPR 2024 · 461 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
Mip-NeRF 360 · PSNR (higher is better)
20 papers, 2024–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 28.01 | DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · 1 citation) |
| 2025 | 29.5 | Topology-Aware 3D Gaussian Splatting: Leveraging Persistent Homology for Optimized Structural Integrity (AAAI 2025 · 0 citations) |
Mip-NeRF360 · SSIM (higher is better)
15 papers, 2024–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 0.833 | DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · 1 citation) |
| 2025 | 0.874 | Topology-Aware 3D Gaussian Splatting: Leveraging Persistent Homology for Optimized Structural Integrity (AAAI 2025 · 0 citations) |
Mip-NeRF360 · LPIPS (lower is better)
14 papers, 2024–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 0.189 | DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · 1 citation) |
| 2025 | 0.172 | Perceptual-GS: Scene-adaptive Perceptual Densification for Gaussian Splatting (ICML 2025 · 0 citations) |
Deep Blending · PSNR (higher is better)
13 papers, 2024–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 30.42 | DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · 1 citation) |
| 2025 | 30.45 | Topology-Aware 3D Gaussian Splatting: Leveraging Persistent Homology for Optimized Structural Integrity (AAAI 2025 · 0 citations) |
Tanks&Temples · PSNR (higher is better)
11 papers, 2024–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 24.96 | DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · 1 citation) |
Deep Blending · SSIM (higher is better)
10 papers, 2024–2025 · small gains only
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 0.907 | DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · 1 citation) |
| 2025 | 0.911 | Topology-Aware 3D Gaussian Splatting: Leveraging Persistent Homology for Optimized Structural Integrity (AAAI 2025 · 0 citations) |
What has been tried
113 papers here report 1,110 measured changes (a component added, removed or swapped, compared with a baseline): helped 72%, no clear effect 4%, hurt 21%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| increase refinement iterations from 2,000 to 7,000 | Mip-NeRF360, average on all scenes · PSNR | 24.87 -> 26.77 | Helped | SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering (CVPR 2024 · 478 citations) |
| add isotropic regularisation | TUM RGB-D, monocular · ATE RMSE (cm) | 4.83 -> 3.96 | Helped | Gaussian Splatting SLAM (CVPR 2024 · 461 citations) |
| add adaptive Gaussian expansion, including adding and deleting Gaussians | Replica #Room0 · ATE | 0.58 -> 0.48 | Helped | GS-SLAM: Dense Visual SLAM with 3D Gaussian Splatting (CVPR 2024 · 360 citations) |
| remove the epipolar encoder | RealEstate10k · PSNR | 26.09 -> 19.89 | Hurt | pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction (CVPR 2024 · 273 citations) |
| Add SAM-based hierarchical semantics | ramen scene · IoU (%) | 28.20 -> 46.74 | Helped | LangSplat: 3D Language Gaussian Splatting (CVPR 2024 · 248 citations) |
| Add pruning | Mip-NeRF360 garden · PSNR | 27.179 -> 27.083 | Hurt | Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis (CVPR 2024 · 189 citations) |
| Add visibility-based camera selection | Sci-Art · SSIM | 0.694 -> 0.885 | Helped | VastGaussian: Vast 3D Gaussians for Large Scene Reconstruction (CVPR 2024 · 187 citations) |
| Add sparse loss | Shiny Blender, half resolution · PSNR | 31.79 -> 32.09 | Helped | GaussianShader: 3D Gaussian Splatting with Shading Functions for Reflective Surfaces (CVPR 2024 · 181 citations) |
| add Gaussian pruning without co-adaptation | Room · PSNR | 31.34 -> 30.67 | Hurt | LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS (NeurIPS 2024 · 174 citations) |
| replace the Splatter Image with fully connected, unstructured output | ShapeNet-SRN single-view ablation · PSNR | 22.25 -> 20.60 | Hurt | Splatter Image: Ultra-Fast Single-View 3D Reconstruction (CVPR 2024 · 172 citations) |
| use Triplane-Gaussian instead of direct 3D Gaussian decoding | GSO novel-view synthesis · PSNR | 18.56 -> 23.15 | Helped | Triplane Meets Gaussian Splatting: Fast and Generalizable Single-View 3D Reconstruction with Transformers (CVPR 2024 · 139 citations) |
| Use linear interpolation for depth and optimize Gaussian-stored normals from depth-derived normals | TensoIR Synthetic · Normal MAE | 6.218 -> 4.948 | Helped | GS-IR: 3D Gaussian Splatting for Inverse Rendering (CVPR 2024 · 123 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| 3DGS | Mip-NeRF 360 | PSNR | 16.06 – 27.72 | 7 papers, e.g. CVPR 2025 and CVPR 2025 |
| 3DGS | Mip-NeRF 360 | SSIM | 0.4 – 0.815 | 5 papers, e.g. CVPR 2025 and CVPR 2024 |
| 3DGS | Mip-NeRF360 | LPIPS | 0.214 – 0.389 | 3 papers, e.g. CVPR 2024 and CVPR 2025 |
| FSGS | LLFF | PSNR | 19.1 – 20.43 | 3 papers, e.g. CVPR 2025 and NeurIPS 2025 |
| FSGS | Mip-NeRF360 | LPIPS | 0.293 – 0.442 | 2 papers, e.g. CVPR 2025 and CVPR 2025 |
| Mini-Splatting | Mip-NeRF 360 | LPIPS | 0.217 – 0.318 | 2 papers, e.g. NeurIPS 2024 and NeurIPS 2025 |
What reviewers will expect
Datasets and baselines named in the 113 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Mip-NeRF 360 · 28 of 113 papers
- Deep Blending · 14 of 113 papers
- Tanks & Temples · 12 of 113 papers
- Tanks and Temples · 10 of 113 papers
- DTU · 9 of 113 papers
- RealEstate10K · 9 of 113 papers
- Replica · 8 of 113 papers
- ACID · 7 of 113 papers
Baselines
- 3DGS · 38 of 113 papers
- 2DGS · 11 of 113 papers
- Scaffold-GS · 10 of 113 papers
- MVSplat · 9 of 113 papers
- Mip-Splatting · 8 of 113 papers
- pixelSplat · 8 of 113 papers
- FSGS · 8 of 113 papers
- NeuS · 7 of 113 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 269 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 14 papers
- Hong Kong University of Science and Technology · 8 papers
- Huawei · 7 papers
- Peking University · 7 papers
- Shanghai Jiao Tong University · 5 papers
- Nanyang Technological University · 5 papers
- Harbin Institute of Technology · 5 papers
- University of Toronto · 5 papers
- National University of Singapore · 5 papers
- Fudan University · 4 papers
Use it
Download BibTeX of 20 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Regularization and primitive design for 3DGS (213 papers)
- Sparse-view Gaussian splatting for novel views (111 papers)
- Animatable Gaussian human avatars (42 papers)
Elsewhere in the topic
- Deformable 4D Gaussian splatting (112 papers)
- Monocular dynamic scene reconstruction (80 papers)
- Animatable Gaussian human avatars (14 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).