The Year in AI Papers
The year/Independent research

Paper 2510.19430

GigaBrain-0: A World Model-Powered Vision-Language-Action Model

GigaBrain Team, Angen Ye, Boyuan Wang, Chaojun Ni, Guan Huang, Guosheng Zhao, Haoyun Li, Jie Li, Jiagang Zhu, Lv Feng, Peng Li, Qiuping Deng, et al.

Published
Oct 2025
Research lab
Independent
Citations
42
GitHub
Not linked

01 In brief

Summary

GigaBrain-0 is a Vision-Language-Action (VLA) foundation model that uses world model-generated data to reduce reliance on costly real-world robot data and improve generalization.

It employs a mixture-of-transformers architecture with a pretrained PaliGemma2 VLM and a Diffusion Transformer for action generation, enhanced by RGBD input modeling and embodied Chain-of-Thought (CoT) supervision.

Training data includes real-world data (public datasets and 1,182 hours of proprietary data) and world-model-generated data via Real2Real, View, Sim2Real, Human Video Transfer, and video generation with inverse dynamics.

Experiments on dexterous, long-horizon, and mobile manipulation tasks show GigaBrain-0 outperforms π0 by 10-30% in success rate.

Generalization experiments demonstrate that increasing the sampling probability of generated data (α) from 0 to 50-90% boosts success rates to over 80% for appearance, placement, and viewpoint variations.

GigaBrain-0-Small, a lightweight variant with 402M parameters, achieves 80% success rate on table bussing with 0.13s latency on NVIDIA Jetson AGX Orin, compared to π0's 1.28s and 3.2B parameters.

02 From the paper

Abstract

Training Vision-Language-Action (VLA) models for generalist robots typically requires large-scale real-world robot data, which is expensive and time-consuming to collect. The inefficiency of physical data collection severely limits the scalability, and generalization capacity of current VLA systems. To address this challenge, we introduce GigaBrain-0, a novel VLA foundation model empowered by world model-generated data (e.g., video generation, real2real transfer, human transfer, view transfer, sim2real transfer data). By leveraging world models to generate diverse data at scale, GigaBrain-0 significantly reduces reliance on real robot data while improving cross-task generalization. Our approach further improves policy robustness through RGBD input modeling and embodied Chain-of-Thought (CoT) supervision, enabling the model to reason about spatial geometry, object states, and long-horizon dependencies during task execution. This leads to substantial gains in real-world performance on dexterous, long-horizon, and mobile manipulation tasks. Extensive experiments demonstrate that GigaBrain-0 achieves superior generalization across variations in appearances (e.g., textures, colors), object placements, and camera viewpoints. Additionally, we present GigaBrain-0-Small, an optimized lightweight variant designed to run efficiently on devices such as the NVIDIA Jetson AGX Orin.