SAM 3D Objects

Native and Compact Structured Latents for 3D Generation

Voyager is an interactive RGBD video generation model conditioned on camera input, and supports real-time 3D reconstruction.

TurboDiffusion: 100–200× Acceleration for Video Diffusion Models

Sharp Monocular View Synthesis in Less Than a Second

Cambrian-S: Towards Spatial Supersensing in Video

Light Video Generation Inference Framework

Official Torch/CUDA Implementation of Faithful Contouring

Dataset and Benchmark code for EgoEdit

Official implementation of Continuous 3D Perception Model with Persistent State

Code for ICCV'2025 (Best student paper honorable mention) "RayZer: A Self-supervised Large View Synthesis Model"

[NeurIPS 2025] WorldMem: Long-term Consistent World Simulation with Memory

WorldGrow: Generating Infinite 3D World [AAAI 2026 Oral]

Code of π^3: Permutation-Equivariant Visual Geometry Learning

VITRA: Scalable Vision-Language-Action Model Pretraining for Robotic Manipulation with Real-Life Human Activity Videos

NEO Series: Native Vision-Language Models from First Principles

Official Repo for Rolling Forcing: Autoregressive Long Video Diffusion in Real Time

Flash Attention in ~100 lines of CUDA (forward pass only)

A general memory system for agents, powered by deep-research

Does Understanding Inform Generation in Unified Multimodal Models? From Analysis to Path Forward

iMontage: Unified, Versatile, Highly Dynamic Many-to-many Image Generation

Part-X-MLLM: Part-aware 3D Multimodal Large Language Model

DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

This is the code for Deformable Neural Radiance Fields, a.k.a. Nerfies.

Research code artifacts for Code World Model (CWM) including inference tools, reproducibility, and documentation.

Unified Vision-Language-Action Model

OpenMMLab Foundational Library for Training Deep Learning Models