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

TL;DR

GigaWorld-0 introduces a comprehensive world model framework that generates diverse, realistic, and controllable embodied interaction data, significantly enhancing embodied AI performance without real-world training data.

Contribution

The paper presents GigaWorld-0, a novel unified world model framework combining large-scale video generation and 3D modeling to produce high-quality data for embodied AI training.

Findings

Generated data is diverse, realistic, and controllable.

VLA models trained on GigaWorld-0 data outperform previous methods.

Significant improvements in real-world robot tasks without real-world data.

Abstract

World models are emerging as a foundational paradigm for scalable, data-efficient embodied AI. In this work, we present GigaWorld-0, a unified world model framework designed explicitly as a data engine for Vision-Language-Action (VLA) learning. GigaWorld-0 integrates two synergistic components: GigaWorld-0-Video, which leverages large-scale video generation to produce diverse, texture-rich, and temporally coherent embodied sequences under fine-grained control of appearance, camera viewpoint, and action semantics; and GigaWorld-0-3D, which combines 3D generative modeling, 3D Gaussian Splatting reconstruction, physically differentiable system identification, and executable motion planning to ensure geometric consistency and physical realism. Their joint optimization enables the scalable synthesis of embodied interaction data that is visually compelling, spatially coherent, physically plausible, and instruction-aligned. Training at scale is made feasible through our efficient GigaTrain framework, which exploits FP8-precision and sparse attention to drastically reduce memory and compute requirements. We conduct comprehensive evaluations showing that GigaWorld-0 generates high-quality, diverse, and controllable data across multiple dimensions. Critically, VLA model (e.g., GigaBrain-0) trained on GigaWorld-0-generated data achieve strong real-world performance, significantly improving generalization and task success on physical robots without any real-world interaction during training.