[ECCV 2026]

[CVPR 2026]

Compute FID scores with PyTorch.

Our inference and training framework to run on the Cosmos Models

[CVPR'25 Oral] Official implementation for "DiffusionRenderer: Neural Inverse and Forward Rendering with Video Diffusion Models"

[SIGGRAPH Asia 2026] Lift4D: Harmonizing Single-View 3D Estimation for 4D Reconstruction In-the-Wild

Mixture-of-Transformers: A Sparse and Scalable Architecture for Multi-Modal Foundation Models. TMLR 2025.

NVIDIA Alpamayo 2 Super is an open 34B multi-task foundation model designed to supercharge autonomous vehicle development.

The first multiplayer video world model in Minecraft

Code for "Scaling Language-Free Visual Representation Learning" paper (Web-SSL).

Our insights of Openpilot, a deepdive project on it

InstantNuRec: Feed-Forward 3D Gaussian Reconstruction from Driving Logs

10k crowdsourced images for training segnets

A curated list of materials on AI efficiency

Open Frontier Intelligence

N₀-TWAM: A Tactile-Native World Action Model for Contact-Rich Manipulation

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.

[SIGGRAPH Asia 2026] UniMate: One Unified Model to Animate Diverse Skeletons

[CVPR 2026] PropFly: Learning to Propagate via On-the-Fly Supervision from Pre-trained Video Diffusion Models

Apple-π: Benchmarking Thinking with Video Towards Law-Grounded Physical Intelligence

[CVPR 2026] Gen3R: 3D Scene Generation Meets Feed-Forward Reconstruction

Code for Xiaomi-Robotics-1

[CVPR 2026] The official PyTorch implementation of the "Vision Transformer Needs More Than Registers".

A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training

Official implementation of ARDY: Autoregressive Diffusion with Hybrid Representation for Interactive Human Motion Generation (SIGGRAPH 2026).