Ruiqian Nai

Ruiqian Nai 佴瑞乾

I am a fourth-year PhD student at the Institute for Interdisciplinary Information Sciences, Tsinghua University, advised by Prof. Yang Gao. Previously, I earned my B.S. degree from the Department of Automation, Tsinghua University. I am currently an Applied Scientist Intern at Amazon FAR (Frontier AI & Robotics), working on humanoid whole-body manipulation, advised by Guanya Shi, C. Karen Liu, Carmelo Sferrazza, and Rocky Duan.

Research Interests

My research focuses on robot learning, which I approach as a deeply systematic challenge spanning hardware understanding, data acquisition, and scalable training. Currently, I focus on humanoid whole-body manipulation, spanning both robotic foundation models and humanoid control. Pursuing general intelligence on today's humanoids is like building autonomous driving on a unicycle — my goal is to make that as tractable as driving a car, by tackling both ends of the stack together.

Email / Scholar / CV / Twitter / Github

Selected Publications (* indicates equal contribution)

	Humanoid Manipulation Interface: Humanoid Whole-Body Manipulation from Robot-Free Demonstrations Ruiqian Nai, Boyuan Zheng, Junming Zhao, Haodong Zhu, Sicong Dai, Zunhao Chen, Yihang Hu, Yingdong Hu, Tong Zhang, Chuan Wen, Yang Gao Preprint, 2026 project page / arXiv / X summary We present HuMI* 🤖 (Humanoid Manipulation Interface), the first robot-free 🚫🤖 framework for learning diverse humanoid whole-body manipulation tasks 👐 across various environments 🏠.
	OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoning Fanqi Lin, Ruiqian Nai, Yingdong Hu*, Jiacheng You, Junming Zhao, Yang Gao ICLR, 2026 project page / arXiv / code / X summary We introduce OneTwoVLA, a single unified vision-language-action model capable of both acting (System One)⚡ and reasoning (System Two)🤔. Importantly, it adaptively determines when to engage each mode.
	HuB: Learning Extreme Humanoid Balance Tong Zhang, Boyuan Zheng, Ruiqian Nai, Yingdong Hu, Yen-Jen Wang, Geng Chen, Fanqi Lin, Jiongye Li, Chuye Hong, Koushil Sreenath, Yang Gao CoRL 2025 (oral) project page / arXiv / X summary We propose HuB (Humanoid Balance) 🤖, a framework that enables humanoids to perform challenging quasi-static balance tasks ⚖️, including extreme single-legged poses 🦵 such as the Swallow Balance 🕊️ and Bruce Lee's Kick 🦶🥋.
	Fine-Tuning Hard-to-Simulate Objectives for Quadruped Locomotion: A Case Study on Total Power Saving Ruiqian Nai, Jiacheng You, Liu Cao, Hanchen Cui, Shiyuan Zhang, Huazhe Xu, Yang Gao ICRA 2025 project page / arXiv / X summary Better locomotion using real-world data. Our approach achieves a 24-28% net reduction in power consumption 🔋.
	Revisiting Disentanglement in Downstream Tasks: A Study on Its Necessity for Abstract Visual Reasoning Ruiqian Nai, Zixin Wen, Ji Li, Yuanzhi Li, Yang Gao AAAI 2024 arxiv / code We show that informativeness 🧠 is a more crucial factor than disentanglement 🌀 in downstream tasks.

Modified from Jon Barron's website.