StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developing
Year: 2026
Type: software
Abstract: StarVLA is a modular and flexible codebase for developing Vision-Language Model (VLM) to Vision-Language-Action (VLA) models. In StarVLA (also a pun on “start VLA” ), each functional component (model, data, trainer, config, evaluation, etc.) follows a top-down, intuitive separation and high cohesion and low coupling principle, which enabling plug-and-play design, rapid prototyping, and independent debugging.
Authors: Jinhui Ye, Axi404, Yilun Chen, MichaelYu, Fangjing Wang, Zeng Shuang, Qixiu Li, Microsoft Open Source, HENG, Shijie Lian, Bin Yu, GUO Weiyu, Zixuan WANG, Jiachen Shen, Danyal Ahmad, Cheng Yin, zhihelu-zero, yuxin chen, rakybond007, jrryzh(SII), Zhijie-Song, Yi Yang (SII), Yang Tian, Wendi Chen (SII), Travor King, Sukai Huang, Senqiao Yang (杨森乔), LiuRicky
Institutions: Hong Kong University of Science and Technology, Xi'an Jiaotong University, Chinese University of Hong Kong, Fudan University, Tsinghua University, Microsoft (United States), Huazhong University of Science and Technology, Harbin Institute of Technology, Shanghai Institute of Immunology, Shanghai Jiao Tong University, Shanghai Innovation Institute
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7124425754