Generative world models used as closed-loop test oracles require a five-level admissibility ladder (L0-L4) because visual fidelity does not predict action-robustness.
hub Canonical reference
Nerf: Representing scenes as neural radiance fields for view synthesis
Canonical reference. 83% of citing Pith papers cite this work as background.
28
Pith papers citing it
6,027
external citations · Crossref
Background
83% of classified citations
hub tools
citation-role summary
background 4
method 2
citation-polarity summary
representative citing papers
ScaLe-INR is a multi-branch INR architecture that applies directional scaling per the Fourier inverse theorem and a directional edge guidance loss to disentangle scales and improve reconstruction fidelity.
A semantics-driven optimization of URBS curves via score distillation sampling produces single continuous line drawings from text prompts or images.
AdpSplit adaptively splits Gaussians using pixel-error statistics to reduce 3DGS training time by 9-22% without quality loss.
S2C-3D reconstructs complete high-fidelity 3D scenes from as few as 6-8 images by finetuning a diffusion model on scene data, applying consistency-conditioned sampling, and planning trajectories for full coverage.
Text Encoded Extrusions (TEE) lets LLMs generate and edit manifold 3D meshes by learning sequences of face extrusions from decomposed quadrilateral meshes.
Volumetric ergodic control optimizes spatial coverage for robots with physical volume using sample-based models, more than doubling coverage efficiency over point-based methods while preserving asymptotic guarantees and achieving 100% task completion.
BEVCALIB performs LiDAR-camera calibration from raw data by fusing camera and LiDAR bird's-eye view features with a novel feature selector and reports state-of-the-art accuracy on KITTI and NuScenes.
By injecting per-Gaussian motion variance and short-window temporal attention into the deformation network, MVFusion-GS improves dynamic-static decomposition and achieves top average metrics on Neu3D and NeRF On-the-go.
PointSplat infers compact Gaussian splats directly in 3D space from input point sets via ray casting and Point-Image Transformer to reduce inter-view redundancy and improve novel-view quality for humans.
FalconTrack automates photorealistic dataset creation via Gaussian Splatting and achieves high zero-shot sim-to-real performance in vision-based aerial tracking using multi-head perception and class-conditioned EKF.
3D Gaussian transient rendering enables NLOS imaging from arbitrary relay geometries in both confocal and non-confocal setups, achieving SOTA on real measurements.
LEGS improves 3D Gaussian Splatting by replacing first-order edge guidance with second-order Laplacian structural guidance and nonlinear pixel-wise weighting, yielding up to 1.68 dB PSNR gain over baseline 3DGS on Tanks&Temples and Mip-NeRF360.
PG-3DGS couples 3D Gaussian Splatting with differentiable physics so that optimized shapes satisfy both visual fidelity and physical objectives such as pouring and aerodynamic lift, with real-world 3D-printed validation.
PropSplat uses optimized 3D Gaussians initialized on transmitter-receiver paths to achieve lower RMSE than NeRF2, GSRF, and WRF-GS+ on outdoor drive-test and indoor BLE datasets while enabling map-free RF reconstruction.
FreeOcc enables training-free open-vocabulary 3D occupancy prediction from RGB-D sequences by combining SLAM, dense Gaussian maps, off-the-shelf vision-language models, and probabilistic projection, achieving over 2x gains on benchmarks and zero-shot transfer to novel scenes.
RealLiFe optimizes multi-plane images with HSGD to deliver real-time light field reconstruction from sparse views, claiming 100x speedup over offline methods and 2 dB PSNR gain over online ones.
Benchmark finds location encoders recover primary spatial coefficients consistently but secondary ones vary by scale, with raw-coordinate baseline competitive throughout.
GRAR refines glass masks from multi-modal vision models and uses a reflection-aware local-global geometric similarity descriptor to remove artifacts in LiDAR point clouds.
FreeTimeGS++ improves dynamic scene reconstruction by identifying emergent temporal partitioning and photometric-motion decoupling in 4DGS, then applying targeted techniques for reduced run-to-run variance.
Continuous trajectory representations of lithium-ion battery aging enable consistent knee-point detection and early remaining useful life predictions that remain robust across heterogeneous datasets.
A collaborative VR workflow with GenAI lets users merge prompts and creatively repurpose outputs to co-create 3D artifacts that narrate shared cultural heritage experiences.
Anisotropic Gaussian primitives compress 3D Taylor–Green turbulence at 1e3–1e4× while recovering more intermediate- and high-wavenumber content than isotropic kernels.
GaNI combines NeuS geometry reconstruction with a light-position-aware inverse neural radiosity stage that adds implicit near-field modeling, surface angle loss, and roughness smoothness priors to recover reflectance parameters from co-located light-camera captures.
citing papers explorer
-
Validate the Dream Before You Trust Its Verdict: Admissibility for World-Model Simulators
Generative world models used as closed-loop test oracles require a five-level admissibility ladder (L0-L4) because visual fidelity does not predict action-robustness.
-
ScaLe-INR: Scale and Learn Implicit Neural Representations
ScaLe-INR is a multi-branch INR architecture that applies directional scaling per the Fourier inverse theorem and a directional edge guidance loss to disentangle scales and improve reconstruction fidelity.
-
Single-Line Drawing Generation via Semantics-Driven Optimization
A semantics-driven optimization of URBS curves via score distillation sampling produces single continuous line drawings from text prompts or images.
-
AdpSplit: Error-Driven Adaptive Splitting for Faster Geometry Discovery in 3D Gaussian Splatting
AdpSplit adaptively splits Gaussians using pixel-error statistics to reduce 3DGS training time by 9-22% without quality loss.
-
Sparse-to-Complete: From Sparse Image Captures to Complete 3D Scenes
S2C-3D reconstructs complete high-fidelity 3D scenes from as few as 6-8 images by finetuning a diffusion model on scene data, applying consistency-conditioned sampling, and planning trajectories for full coverage.
-
Learning to Build Shapes by Extrusion
Text Encoded Extrusions (TEE) lets LLMs generate and edit manifold 3D meshes by learning sequences of face extrusions from decomposed quadrilateral meshes.
-
Volumetric Ergodic Control
Volumetric ergodic control optimizes spatial coverage for robots with physical volume using sample-based models, more than doubling coverage efficiency over point-based methods while preserving asymptotic guarantees and achieving 100% task completion.
-
BEVCALIB: LiDAR-Camera Calibration via Geometry-Guided Bird's-Eye View Representations
BEVCALIB performs LiDAR-camera calibration from raw data by fusing camera and LiDAR bird's-eye view features with a novel feature selector and reports state-of-the-art accuracy on KITTI and NuScenes.
-
MVFusion-GS: Motion-Variance Guided Temporal Attention for High-Quality Dynamic Gaussian Splatting
By injecting per-Gaussian motion variance and short-window temporal attention into the deformation network, MVFusion-GS improves dynamic-static decomposition and achieves top average metrics on Neu3D and NeRF On-the-go.
-
PointSplat: Compact Gaussian Splatting via Human-Centric Prediction
PointSplat infers compact Gaussian splats directly in 3D space from input point sets via ray casting and Point-Image Transformer to reduce inter-view redundancy and improve novel-view quality for humans.
-
FalconTrack: Photorealistic Auto-Labeled Perception and Physics-Aware Vision-Based Aerial Tracking
FalconTrack automates photorealistic dataset creation via Gaussian Splatting and achieves high zero-shot sim-to-real performance in vision-based aerial tracking using multi-head perception and class-conditioned EKF.
-
Non-line-of-sight imaging with arbitrary relay surface geometries via 3D Gaussian Transient Rendering
3D Gaussian transient rendering enables NLOS imaging from arbitrary relay geometries in both confocal and non-confocal setups, achieving SOTA on real measurements.
-
LEGS: Laplacian-Enhanced Gaussian Splatting with a Nonlinear Weighted Loss
LEGS improves 3D Gaussian Splatting by replacing first-order edge guidance with second-order Laplacian structural guidance and nonlinear pixel-wise weighting, yielding up to 1.68 dB PSNR gain over baseline 3DGS on Tanks&Temples and Mip-NeRF360.
-
PG-3DGS: Optimizing 3D Gaussian Splatting to Satisfy Physics Objectives
PG-3DGS couples 3D Gaussian Splatting with differentiable physics so that optimized shapes satisfy both visual fidelity and physical objectives such as pouring and aerodynamic lift, with real-world 3D-printed validation.
-
PropSplat: Map-Free RF Field Reconstruction via 3D Gaussian Propagation Splatting
PropSplat uses optimized 3D Gaussians initialized on transmitter-receiver paths to achieve lower RMSE than NeRF2, GSRF, and WRF-GS+ on outdoor drive-test and indoor BLE datasets while enabling map-free RF reconstruction.
-
FreeOcc: Training-Free Embodied Open-Vocabulary Occupancy Prediction
FreeOcc enables training-free open-vocabulary 3D occupancy prediction from RGB-D sequences by combining SLAM, dense Gaussian maps, off-the-shelf vision-language models, and probabilistic projection, achieving over 2x gains on benchmarks and zero-shot transfer to novel scenes.
-
RealLiFe: Real-Time Light Field Reconstruction via Hierarchical Sparse Gradient Descent
RealLiFe optimizes multi-plane images with HSGD to deliver real-time light field reconstruction from sparse views, claiming 100x speedup over offline methods and 2 dB PSNR gain over online ones.
-
Do Location Encoders Capture Spatial Effects? A GeoShapley Benchmark Across Scales
Benchmark finds location encoders recover primary spatial coefficients consistently but secondary ones vary by scale, with raw-coordinate baseline competitive throughout.
-
GRAR: Glass-induced Reflection Artifact Removal in LiDAR Point Clouds
GRAR refines glass masks from multi-modal vision models and uses a reflection-aware local-global geometric similarity descriptor to remove artifacts in LiDAR point clouds.
-
FreeTimeGS++: Secrets of Dynamic Gaussian Splatting and Their Principles
FreeTimeGS++ improves dynamic scene reconstruction by identifying emergent temporal partitioning and photometric-motion decoupling in 4DGS, then applying targeted techniques for reduced run-to-run variance.
-
Continuous ageing trajectory representations for knee-aware lifetime prediction of lithium-ion batteries across heterogeneous dataset
Continuous trajectory representations of lithium-ion battery aging enable consistent knee-point detection and early remaining useful life predictions that remain robust across heterogeneous datasets.
-
"From remembering to shaping": Narrating Shared Experiences by Co-Designing Cultural Heritage Artifacts in Collaborative VR
A collaborative VR workflow with GenAI lets users merge prompts and creatively repurpose outputs to co-create 3D artifacts that narrate shared cultural heritage experiences.
-
Gaussian Field Representations for Turbulent Flow: Compression, Scale Separation, and Physical Fidelity
Anisotropic Gaussian primitives compress 3D Taylor–Green turbulence at 1e3–1e4× while recovering more intermediate- and high-wavenumber content than isotropic kernels.
-
GaNI: Global and Near Field Illumination Aware Neural Inverse Rendering
GaNI combines NeuS geometry reconstruction with a light-position-aware inverse neural radiosity stage that adds implicit near-field modeling, surface angle loss, and roughness smoothness priors to recover reflectance parameters from co-located light-camera captures.
-
FalconApp: Rapid iPhone Deployment of End-to-End Perception via Automatically Labeled Synthetic Data
A short iPhone capture of a rigid object is reconstructed as a GSplat, auto-labeled into synthetic training data, and deployed as a ~30 ms on-device mask+pose model that beats PnP on 4/5 tested objects.
-
Implicit neural representations as a coordinate-based framework for continuous environmental field reconstruction from sparse ecological observations
Implicit neural representations enable stable, resolution-independent reconstruction of continuous environmental fields from sparse and irregular ecological data.
-
3D Reconstruction Techniques in the Manufacturing Domain: Applications, Research Opportunities and Use Cases
A survey of 106 papers finds quality inspection dominates 3D reconstruction use in manufacturing at 40 percent of applications, with a shift toward hybrid sensor systems and a noted gap in unified frameworks.
-
A Survey on Deep Learning Architectures for Point Cloud Classification and Segmentation
A survey that categorizes deep learning models for point cloud tasks by backbone architecture, evaluates benchmark performance, and outlines challenges and future research directions.