Advancing Open-source World Models

World Models That Know When They Don't Know: Controllable Video Generation with Calibrated Uncertainty

Official implementation for WorldScore: A Unified Evaluation Benchmark for World Generation

Holistic Evaluation of Multimodal LLMs on Spatial Intelligence

A construction kit for reinforcement learning environment management.

[ICLR 2026] Trace Anything: Representing Any Video in 4D via Trajectory Fields

Code, data and weights for the paper **What drives success in physical planning with Joint-Embedding Predictive World Models?**

Code for "Evaluating Robot Policies in a World Model".

A Comprehensive Survey on World Models for Embodied AI

Ray tracing and hybrid rasterization of Gaussian particles

Official repository for "RLVR-World: Training World Models with Reinforcement Learning" (NeurIPS 2025), https://arxiv.org/abs/2505.13934

HY-World 1.5: A Systematic Framework for Interactive World Modeling with Real-Time Latency and Geometric Consistency

Joint 3D Geometry Reconstruction and Motion Generation for 4D Synthesis from a Single Image

Code implementation of the paper "World-in-World: World Models in a Closed-Loop World"

MapAnything: Universal Feed-Forward Metric 3D Reconstruction

Educational Resource for Isaac Lab

Code for "Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation"

SkyRL: A Modular Full-stack RL Library for LLMs

[NeurIPS 2025] The official repository of "Sekai: A Video Dataset towards World Exploration"

Async RL Training at Scale

SimWorld: An Open-ended Realistic Simulator for Autonomous Agents in Physical and Social Worlds

GigaWorld-0: World Models as Data Engine to Empower Embodied AI

A generative world for general-purpose robotics & embodied AI learning.

ENACT is a benchmark that evaluates embodied cognition through world modeling from egocentric interaction. It is designed to be simple and have a scalable dataset.

Inferix: A Block-Diffusion based Next-Generation Inference Engine for World Simulation

[NeurIPS 2025] Elevating Visual Perception in Multimodal LLMs with Visual Embedding Distillation

G2VLM: Geometry Grounded Vision Language Model with Unified 3D Reconstruction and Spatial Reasoning

[Arxiv 2025] In-Video Instructions: Visual Signals as Generative Control

HoloMotion: A Foundation Model for Whole-Body Humanoid Control