Shuanghao Bai

Shuanghao Bai

Ph.D. student @ Xi'an Jiaotong University

I am a fourth-year direct Ph.D. student in Artificial Intelligence at Xi’an Jiaotong University, where I am advised by Prof. Badong Chen. Prior to my doctoral studies, I received my Bachelor’s degree in Automation from Chongqing University in 2022, under the supervision of Prof. Min Zhao.

🔭 My research interests lie in generalization in computer vision and robot learning, vision-language models, and vision language action models.

✉️ Welcome to contact me for any discussion and cooperation!

🔥 News

  • [2025/08/02]: Long-VLA - the first VLA model to enable skill chaining in long-horizon tasks, accompanied by the introduction of a new benchmark, L-CALVIN; accepted to CoRL 2025! See Project page.
  • [2025/05/06]: We released OpenHelix, which provides a short survey and empirical analysis of dual-system VLA, and introduces a novel open-source dual-system VLA model.
  • [2025/05/01]: BC-IB, the first to introduce information bottleneck theory into robotic manipulation through visual imitation learning under the lens of information theory, got accepted for ICML 2025! See Project page.
  • [2025/03/24]: One paper on causal discovery that integrates Minimum Error Entropy to enable dynamic adaptation to varying levels of complexity and noise got accepted for Neural Networks 2025!
  • [2025/01/23]: VLAS, the first vision-language-action model that incorporates speech instructions for robotic manipulation, got accepted for ICLR 2025!
📜 Historical News
  • [2024/12/21]: PromptTA, a novel VLM-based source-free domain generalization method integrating a text adapter and diverse prompt inputs, got accepted by ICASSP 2025!
  • [2024/10/23]: The GitHub repository Awesome-Robotics-Manipulation is now public! Let’s work together to build a comprehensive and valuable resource for the robotics and AI community!
  • [2024/07/04]: SPG, a novel VLM-based domain generalization method that introduces generative concepts into prompt learning, got accepted by ECCV 2024.
  • [2024/05/02]: JRNGC, a unified causal discovery method that leverages the Jacobian matrix to address high-dimensional multivariate causal discovery, got accepted by ICML 2024!
  • [2024/12/14]: One paper on cross-domain few-shot classification got accepted by ICASSP 2024.
  • [2023/12/09]: PDA, a novel VLM-based prompt learning approach for unsupervised domain adaptation that integrates and thoroughly evaluates diverse prompt learning methods, got accepted by AAAI 2024!

📝 Publications

Published

BCIB

Rethinking Latent Redundancy in Behavior Cloning: An Information Bottleneck Approach for Robot Manipulation

Shuanghao Bai, Wanqi Zhou, Pengxiang Ding, Wei Zhao, Donglin Wang, Badong Chen

ICML 2025

arXiv| Project | Code

SPG

Soft Prompt Generation for Domain Generalization

Shuanghao Bai*, Yuedi Zhang*, Wanqi Zhou, Yicong He, Zhirong Luan, Badong Chen

ECCV 2024

Paper | arXiv | Code | Chinese Intro

JRNGC

Jacobian Regularizer-based Neural Granger Causality

Wanqi Zhou, Shuanghao Bai, Shujian Yu, Qibin Zhao, Badong Chen

ICML 2024

Paper | arXiv | Code

PDA

Prompt-based Distribution Alignment for Unsupervised Domain Adaptation

Shuanghao Bai, Min Zhang, Wanqi Zhou, Siteng Huang, Zhirong Luan, Donglin Wang, Badong Chen

AAAI 2024

Paper | arXiv | Code

Yiguo Fan*, Pengxiang Ding*, Shuanghao Bai*, Xinyang Tong*, Yuyang Zhu, Hongchao Lu, Fengqi Dai, Wei Zhao, Yang Liu, Siteng Huang, Zhaoxin Fan, Badong Chen, Donglin Wang. "Long-VLA: Unleashing Long-Horizon Capability of Vision Language Action Model for Robot Manipulation". [arXiv] [Project]

Wanqi Zhou, Shuanghao Bai, Qibin Zhao, Badong Chen. "An Information-Theoretic Approach for Heterogeneous Differentiable Causal Discovery". [Paper] [Code]

Wei Zhao, Pengxiang Ding, Zhang Min, Zhefei Gong, Shuanghao Bai, Han Zhao, Donglin Wang. "VLAS: Vision-Language-Action Model with Speech Instructions for Customized Robot Manipulation". [arXiv] [Code]

Haoran Zhang*, Shuanghao Bai*, Wanqi Zhou, Jingwen Fu, Badong Chen. "PromptTA: Prompt-driven Text Adapter for Source-free Domain Generalization". [Paper] [arXiv] [Code]

Shuanghao Bai, Wanqi Zhou, Zhirong Luan, Donglin Wang, Badong Chen. "Improving Cross-domain Few-shot Classification with Multilayer Perceptron". [Paper] [arXiv] [Code]

Preprints & Under Submission

Yuedi Zhang, Shuanghao Bai, Wanqi Zhou, Zhirong Luan, Badong Chen. "Dual-Path Stable Soft Prompt Generation for Domain Generalization". [arXiv] [Code]

Can Cui, Pengxiang Ding, Wenxuan Song, Shuanghao Bai, Xinyang Tong, Zirui Ge, Runze Suo, Wanqi Zhou, Yang Liu, Bofang Jia, Han Zhao, Siteng Huang, Donglin Wang. "Openhelix: A Short Survey, Empirical Analysis, and Open-source Dual-system VLA Model for Robotic Manipulation". [arXiv] [Code] [Project]

Wanqi Zhou*, Shuanghao Bai*, Danilo Mandic, Qibin Zhao, Badong Chen. "Revisiting the Adversarial Robustness of Vision Language Models: a Multimodal Perspective". [arXiv] [Code]

📖 Research Experience

🏅 Honors and Awards

  • National Scholarship, 2024
  • Outstanding Undergraduate Thesis of College of Automation, Chongqing University, 2022
  • National Scholarship, 2019
  • Outstanding Student of Chongqing University, 2019