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

2025-11-24

AI Summary

Embodied AI suffers from a scarcity of high-quality, instruction-aligned, and physically plausible vision-language-action (VLA) interaction data. Method: This paper introduces GigaWorld-0, a scalable unified world model framework serving as a data engine. It features a novel dual-branch architecture synergizing video generation with 3D geometric-physical modeling, incorporating FP8 training, sparse attention, 3D Gaussian splatting reconstruction, differentiable physical system identification, and motion planning to synthesize controllable, high-fidelity, spatiotemporally coherent, texture-rich, and geometrically consistent VLA data. Contribution/Results: GigaWorld-0 is the first framework enabling instruction-driven, physically verifiable, and massively scalable autonomous generation of embodied interaction data. Experiments show that VLA models trained exclusively on synthetic data achieve significantly higher task success rates on real robots and surpass real-data baselines in generalization—demonstrating the feasibility of purely synthetic-data-driven embodied intelligence.

Technology Category

• Intelligent Robots: Embodied AI
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

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.

Problem

• Developing scalable world models to generate embodied AI training data
• Ensuring visual coherence and physical realism in synthetic environments
• Enabling robot task success without real-world training interactions

Innovation

• Unified world model framework for Vision-Language-Action learning
• Combines video generation with 3D modeling for embodied sequences
• Efficient training using FP8-precision and sparse attention