Research map / 3D vision / Gaussian splatting
Sparse-view Gaussian splatting for novel views
Uses depth, epipolar and view priors for generalizable novel view synthesis from few images.
111 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 2 of 2 tracked benchmarks.
13 papers in 2023–24 and 98 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 10.7% to 13.6%.
Open questions the data points to
- DTU, 3 views (PSNR): best went from 19.99 (up to 2024) to 20.51 in 2025 — Novel View Synthesis from A Few Glimpses via Test-Time Natural Video Completion (NeurIPS 2025 · 0 citations)
- LLFF, 3 views (PSNR): best went from 20.44 (up to 2024) to 21.07 in 2025 — NexusGS: Sparse View Synthesis with Epipolar Depth Priors in 3D Gaussian Splatting (CVPR 2025 · 24 citations)
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 · Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 2024 · 463 citations)
- 2025 · DIFIX3D+: Improving 3D Reconstructions with Single-Step Diffusion Models (CVPR 2025 · 48 citations)
- 2026 · WeatherEdit: Controllable Weather Editing with 4D Gaussian Field (AAAI 2026 · 4 citations)
Earliest papers here
- Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 2024 · 463 citations)
- DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth Normalization (CVPR 2024 · 202 citations)
- GSDF: 3DGS Meets SDF for Improved Neural Rendering and Reconstruction (NeurIPS 2024 · 24 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.
DTU, 3 views · PSNR (higher is better)
4 papers, 2024–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 19.99 | How to Use Diffusion Priors under Sparse Views? (NeurIPS 2024 · 3 citations) |
| 2025 | 20.51 | Novel View Synthesis from A Few Glimpses via Test-Time Natural Video Completion (NeurIPS 2025 · 0 citations) |
LLFF, 3 views · PSNR (higher is better)
4 papers, 2024–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 20.44 | How to Use Diffusion Priors under Sparse Views? (NeurIPS 2024 · 3 citations) |
| 2025 | 21.07 | NexusGS: Sparse View Synthesis with Epipolar Depth Priors in 3D Gaussian Splatting (CVPR 2025 · 24 citations) |
What has been tried
52 papers here report 509 measured changes (a component added, removed or swapped, compared with a baseline): helped 81%, no clear effect 3%, hurt 15%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| use anchor growing | DB-PLAYROOM · PSNR | 29.12 -> 31.07 | Helped | Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 2024 · 463 citations) |
| add shape freezing during depth regularization | LLFF, 3 views · PSNR | 17.96 -> 19.12 | Helped | DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth Normalization (CVPR 2024 · 202 citations) |
| Apply DIFIX directly to rendered views, without 3D updates | Nerfbusters · FID | 134.65 -> 49.87 | Helped | DIFIX3D+: Improving 3D Reconstructions with Single-Step Diffusion Models (CVPR 2025 · 48 citations) |
| replace average depth blending with Flow-Resilient Depth Blending | LLFF (3 Views) · PSNR | 20.39 -> 20.90 | Helped | NexusGS: Sparse View Synthesis with Epipolar Depth Priors in 3D Gaussian Splatting (CVPR 2025 · 24 citations) |
| Add depth-guided ray sampling to the SDF branch | Mip-NeRF360, Tanks&Temples, Deep Blending · PSNR | 29.26 -> 29.38; 27.30 -> 27.37; 30.17 -> 30.38 | Helped | GSDF: 3DGS Meets SDF for Improved Neural Rendering and Reconstruction (NeurIPS 2024 · 24 citations) |
| add the progressive denoising module | bike low-light · PSNR | 22.36 -> 22.75 | Helped | LITA-GS: Illumination-Agnostic Novel View Synthesis via Reference-Free 3D Gaussian Splatting and Physical Priors (CVPR 2025 · 19 citations) |
| add SVD refinement conditioned on rendered images | DL3DV-10K · SSIM | 0.433 -> 0.399 | Hurt | MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 · 17 citations) |
| Add joint end-to-end training | nuScenes SF, 352 × 640 · PSNR | 21.50 -> 21.67 | Helped | DrivingForward: Feed-forward 3D Gaussian Splatting for Driving Scene Reconstruction from Flexible Surround-view Input (AAAI 2025 · 14 citations) |
| remove LiDAR condition (ablation implies adding it) | WOD unrecorded views @3m · FID | 21.3 -> 18.6 | Helped | FreeSim: Toward Free-viewpoint Camera Simulation in Driving Scenes (CVPR 2025 · 14 citations) |
| replace depth uncertainty modeling with directly predicting spatial extent | nuScenes vehicle segmentation · IoU | 1.0% lower performance | Hurt | Toward Real-world BEV Perception: Depth Uncertainty Estimation via Gaussian Splatting (CVPR 2025 · 12 citations) |
| Add triplane cropping | Phototourism · PSNR | 29.15 -> 29.65; 24.93 -> 24.99; 23.61 -> 24.45 | Helped | Wild-GS: Real-Time Novel View Synthesis from Unconstrained Photo Collections (NeurIPS 2024 · 12 citations) |
| add chart encodings | DTU · Chamfer Distance | 2.693 -> 1.04 | Helped | MAtCha Gaussians: Atlas of Charts for High-Quality Geometry and Photorealism From Sparse Views (CVPR 2025 · 9 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 |
|---|---|---|---|---|
| Scaffold-GS | Mip-NeRF360 | PSNR | 27.71 – 29.34 | 2 papers, e.g. NeurIPS 2025 and NeurIPS 2024 |
What reviewers will expect
Datasets and baselines named in the 49 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 · 11 of 49 papers
- DTU · 8 of 49 papers
- Tanks and Temples · 6 of 49 papers
- LLFF · 5 of 49 papers
- nuScenes · 4 of 49 papers
- DL3DV · 4 of 49 papers
- Replica · 4 of 49 papers
- Waymo Open Dataset · 3 of 49 papers
Baselines
- 3DGS · 16 of 49 papers
- MVSplat · 6 of 49 papers
- FSGS · 6 of 49 papers
- 2DGS · 5 of 49 papers
- DNGaussian · 4 of 49 papers
- pixelSplat · 3 of 49 papers
- EmerNeRF · 3 of 49 papers
- LangSplat · 3 of 49 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 111 papers; from OpenAlex and the papers' affiliations).
- Zhejiang University · 6 papers
- Shanghai Jiao Tong University · 3 papers
- Beihang University · 3 papers
- Huawei · 3 papers
- Meta · 3 papers
- Harbin Institute of Technology · 2 papers
- East China Normal University · 2 papers
- Chinese Academy of Sciences · 2 papers
- Tsinghua University · 2 papers
- Google · 2 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
- Compact and efficient Gaussian splatting (269 papers)
- Regularization and primitive design for 3DGS (213 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).