Research map / 3D vision / NeRF and radiance fields
Sparse-input NeRF regularization
Improves NeRF view synthesis with geometry, depth and ray priors, and in-the-wild robustness.
106 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2021–2026) · an approach in NeRF and radiance fields, keyword cluster “radiance fields · fields nerf · nerfs”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Few recent reports The tracked benchmarks have at most one report from 2025–26 that could beat the old record, too few to say whether they are still moving.
64 papers in 2023–24 and 20 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 24.1% to 22.2%.
No longer reported by papers here: Synthetic-NeRF (PSNR, last in 2023); LLFF (PSNR, last in 2024).
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
- 2021 · pixelNeRF: Neural Radiance Fields From One or Few Images (CVPR 2021 · 1,530 citations)
- 2022 · Direct Voxel Grid Optimization: Super-Fast Convergence for Radiance Fields Reconstruction (CVPR 2022 · 932 citations)
- 2023 · FreeNeRF: Improving Few-Shot Neural Rendering With Free Frequency Regularization (CVPR 2023 · 307 citations)
- 2024 · ReconFusion: 3D Reconstruction with Diffusion Priors (CVPR 2024 · 185 citations)
- 2025 · FrugalNeRF: Fast Convergence for Extreme Few-shot Novel View Synthesis without Learned Priors (CVPR 2025 · 8 citations)
- 2026 · HDR-NSFF: High Dynamic Range Neural Scene Flow Fields (ICLR 2026 · 0 citations)
Earliest papers here
- pixelNeRF: Neural Radiance Fields From One or Few Images (CVPR 2021 · 1,530 citations)
- NeRF in the Wild: Neural Radiance Fields for Unconstrained Photo Collections (CVPR 2021 · 1,399 citations)
- Stereo Radiance Fields (SRF): Learning View Synthesis for Sparse Views of Novel Scenes (CVPR 2021 · 215 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.
Synthetic-NeRF · PSNR (higher is better)
5 papers, 2022–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 32.8 | Direct Voxel Grid Optimization: Super-Fast Convergence for Radiance Fields Reconstruction (CVPR 2022 · 932 citations) |
| 2023 | 32.86 | Compressing Volumetric Radiance Fields to 1 MB (CVPR 2023 · 73 citations) |
LLFF · PSNR (higher is better)
4 papers, 2023–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 26.46 | Compressing Volumetric Radiance Fields to 1 MB (CVPR 2023 · 73 citations) |
| 2024 | 27.41 | AltNeRF: Learning Robust Neural Radiance Field via Alternating Depth-Pose Optimization (AAAI 2024 · 5 citations) |
LLFF (3-view) · PSNR (higher is better)
4 papers, 2023–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 19.63 | FreeNeRF: Improving Few-Shot Neural Rendering With Free Frequency Regularization (CVPR 2023 · 307 citations) |
| 2024 | 21.34 | ReconFusion: 3D Reconstruction with Diffusion Priors (CVPR 2024 · 185 citations) |
What has been tried
92 papers here report 936 measured changes (a component added, removed or swapped, compared with a baseline): helped 70%, no clear effect 6%, hurt 23%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| replace PixelNeRF conditioning with pose conditioning | In-domain datasets · PSNR | 25.15 -> 20.57 | Hurt | ReconFusion: 3D Reconstruction with Diffusion Priors (CVPR 2024 · 185 citations) |
| Use grid resolution M > 256 instead of a smaller grid | Lego, NeRF Synthetic · AUSE | No clear effect | Bayes' Rays: Uncertainty Quantification for Neural Radiance Fields (CVPR 2024 · 58 citations) | |
| construct the volume in the target view rather than the reference input view | RealEstate10K large baseline · PSNR | 20.20 -> 21.29 | Helped | MuRF: Multi-Baseline Radiance Fields (CVPR 2024 · 34 citations) |
| Remove time-dependent network components | COMPOSITE · visual fidelity | Hurt | Neural Underwater Scene Representation (CVPR 2024 · 32 citations) | |
| Add semantic filtering of dynamic classes | Pandaset dynamic scenes · rotation error (°) | 2.36° -> 0.41° | Helped | SOAC: Spatio-Temporal Overlap-Aware Multi-Sensor Calibration using Neural Radiance Fields (CVPR 2024 · 19 citations) |
| Add spatial consistency loss to VideoRF | ReRF · PSNR | Helped | VideoRF: Rendering Dynamic Radiance Fields as 2D Feature Video Streams (CVPR 2024 · 19 citations) | |
| Use depth-guided sampling and semantic rendering rather than uniform sampling with a pretrained 2D segmentor | ScanNet · PSNR | 28.48 -> 31.49 | Helped | GSNeRF: Generalizable Semantic Neural Radiance Fields with Enhanced 3D Scene Understanding (CVPR 2024 · 16 citations) |
| Add single-frame Lidar depth supervision | Pandaset lane shift · FID @ 3.7m | 151.0 -> 138.2 | Helped | LidaRF: Delving into Lidar for Neural Radiance Field on Street Scenes (CVPR 2024 · 15 citations) |
| Add SfM heuristics, color-residual heuristics, and SAM static maps | Kubric (Avg.) · PSNR | 29.76 -> 40.40 | Helped | NeRF-HuGS: Improved Neural Radiance Fields in Non-static Scenes Using Heuristics-Guided Segmentation (CVPR 2024 · 15 citations) |
| remove the feature encoder and learn the latent code directly | Chair, NeRF-Synthetic · PSNR | 35.08 -> 34.41 dB | Hurt | NeRFCodec: Neural Feature Compression Meets Neural Radiance Fields for Memory-Efficient Scene Representation (CVPR 2024 · 13 citations) |
| Add regularization terms | KITTI-360 · PSNR | 27.46 -> 27.52 | Helped | Video2Game: Real-time Interactive Realistic and Browser-Compatible Environment from a Single Video (CVPR 2024 · 13 citations) |
| Add HDR fusion and the exposure evaluator | LOM average low- and high-exposure · PSNR | 5.99 -> 19.47; 6.53 -> 18.59; 6.08 -> 20.97; 6.08 -> 16.05; 6.27 -> 20.83 | Helped | Enhancing Neural Radiance Fields with Adaptive Multi-Exposure Fusion: A Bilevel Optimization Approach for Novel View Synthesis (AAAI 2024 · 12 citations) |
What reviewers will expect
Datasets and baselines named in the 13 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- DTU · 2 of 13 papers
Baselines
- TensoRF · 2 of 13 papers
- 3D-GS · 2 of 13 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 84 papers; from OpenAlex and the papers' affiliations).
- Google · 6 papers
- Stanford University · 5 papers
- Tsinghua University · 3 papers
- Meta · 3 papers
- Shanghai Jiao Tong University · 3 papers
- Huawei · 3 papers
- University of California, San Diego · 3 papers
- Adobe · 3 papers
- Simon Fraser University · 3 papers
- Southeast University · 2 papers
Use it
Download BibTeX of 39 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
- Dynamic and editable NeRF (88 papers)
- Single-view and streaming radiance fields (49 papers)
- NeRF for scene decomposition and understanding (30 papers)
Elsewhere in the topic
- Inverse rendering with neural fields (110 papers)
- Few-shot neural rendering architectures (53 papers)
- 3D-aware generative radiance fields (20 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).