[CVPR 2026] InfiniDepth: Arbitrary-Resolution and Fine-Grained Depth Estimation with Neural Implicit Fields

DGGT: Feedforward 4D Reconstruction of Dynamic Driving Scenes using Unposed Images

[ECCV 2026]Official implementation of the paper: "FoundationGeo: Learning Spatial Pixel-Wise Fields for Monocular Metric Geometry"

MapAnything: Universal Feed-Forward Metric 3D Reconstruction

Cosmos-Drive-Dreams: Scalable Synthetic Driving Data Generation with World Foundation Models

[CVPR 2026] SpatialVID: A Large-Scale Video Dataset with Spatial Annotations

Let us control diffusion models!

Cameras as Relative Positional Encoding

Official Implementation of OmniWeaving: Towards Unified Video Generation with Free-form Composition and Reasoning

SymphoMotion: Joint Control of Camera Motion and Object Dynamics for Coherent Video Generation [CVPR 2026]

[CVPR 2024 Highlight] Feature 3DGS: Supercharging 3D Gaussian Splatting to Enable Distilled Feature Fields

[ICLR 2026] π^3: Permutation-Equivariant Visual Geometry Learning

NVIDIA Cosmos is an open platform of world models, datasets, and tools that enables developers to build Physical AI for robots, autonomous vehicles, smart infrastructure, and more.

Cosmos-Predict2.5, the latest version of the Cosmos World Foundation Models (WFMs) family, specialized for simulating and predicting the future state of the world in the form of video.

Official Implementation of "DriveWAM: Video Generative Priors Enable Scalable World-Action Modeling for Autonomous Driving"

News: the 10k dataset is ready for download.

Project Lyra: Open Generative 3D World Models

ViPE: Video Pose Engine for Geometric 3D Perception

HorizonDrive: Self-Corrective Autoregressive World Model for Long-horizon Driving Simulation

SuperMap is a living spatial memory for embodied AI — it perceives the world, remembers its evolution, and supports reasoning and action.

Infinite Worlds with Versatile Interactions

Scaling Mixture-of-Experts Video Pretraining for Embodied Intelligence

Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.

A feed-forward 3D foundation model for reconstructing scenes from streaming data

Collect some World Models for Autonomous Driving (and Robotic, etc.) papers.

[ICLR 2026] Official Implementation of "UniSplat: Unified Spatio-Temporal Fusion via 3D Latent Scaffolds for Dynamic Driving Scene Reconstruction""