Michael Yuan | 袁承博

Michael (Chengbo) Yuan | 袁承博

🦊🦁🐮🐹🐰🐸🐶🐻🐨🐯

I am a Master student in Institute for Interdisciplinary Information Science (IIIS) at Tsinghua University, advised by Prof. Yang Gao🥳. Currently, my research interests mainly focus on Embodied AI and Robot Learning (Tactile Manipulation, Egocentric Human Data, RL). I am also interested in and exploring venture capital and quantitative investment.

Perviously, I received my bachelor degree from the HongYi Honor class at the School of CS, Wuhan University😘, where I win the highest-level scholarship (10/60k+), Lei Jun Outstanding Scholarship with 100k RMB reward. I am also minoring in Law at School of Law. I have also interned at Sharpa, MoonshotAI and Huawei.

I welcome exploration, creation and excitement! 🤗🤩🤪

Google Scholar / Github / X(Twitter) / RedBook / WeBlog / Zhihu

contact me: ycb24@mails.tsinghua.edu.cn michaelyuancb@163.com

Blog

Notes on research, engineering, and ideas I want to share. Full Blog List.

Robot learning research notes illustration

2026.08.14 The Birth of Prometheus A conceptual roadmap for building a new type of intelligent machine. Now I become Prometheus, the creator of Life.

2024.01.23 A Path towards Robotic Foundation Model A framework for building generalist robotic foundation model in early stage.

Research Here I only show representative research. For the complete list, please refer to my Google Scholar.
	FTP-1: A Generalist Foundation Tactile Policy Across Tactile Sensors for Contact-Rich Manipulation Chengbo Yuan, Zicheng Zhang, Mingjie Zhou, Wendi Chen, Yi Wang, Zhuoyang Liu, Dantong Niu, Shuo Wang, Hui Zhang, Wenkang Zhang, Yingdong Hu, Yuanqing Gong, Wanli Xing, Chuan Wen, Cewu Lu, Kaifeng Zhang, Yang Gao. project page / arXiv / code / model / dataset In Submission, 2026* FTP-1 is the first generalist foundation tactile policy across different tactile sensors for contact-rich manipulation. Pretrained on 3000 hours of tactile manipulation across 21 tactile sensors and 26 data sources, FTP-1 learns general tactile knowledge that can even transfer to unknown sensors.
	MotionTrans: Human VR Data Enable Motion-Level Learning for Robotic Manipulation Policies Chengbo Yuan, Rui Zhou, Mengzhen Liu, Yingdong Hu, Shengjie Wang, Li Yi, Chuan Wen, Shanghang Zhang, Yang Gao. project page / arXiv / code / dataset International Conference on Robotics & Automation (ICRA), 2026* MotionTrans is our first framework directly transfers motions of 13 human tasks to end-to-end policies (RGB-to-Action). The framework is designed at a systematic level, including data collection toolkit, human-to-robot data processing, and a weighted multi-task cotraining strategy.
	RoboEngine: Plug-and-Play Robot Data Augmentation with Semantic Robot Segmentation and Background Generation Chengbo Yuan, Suraj Joshi, Shaoting Zhu, Hang Su, Hang Zhao, Yang Gao. project page / arXiv / code / dataset International Conference on Intelligent Robots and Systems (IROS), 2025* RoboEngine is the first plug-and-play visual robot data augmentation toolkit. Users can effortlessly generate physics-aware robot scenes with few lines code. This enable training only in one scenes and visual generalizing to almost arbitrary scenes.
	Self-Supervised Monocular 4D Scene Reconstruction for Egocentric Videos Chengbo Yuan, Geng Chen, Li Yi, Yang Gao. project page / arXiv / code / thread International Conference on Computer Vision (ICCV), 2025 We train EgoMono4D, a fast, dense and generalizable 4D reconstruction model for egocentric videos with label-free self-supervised methods. The training is conducted sole on unlabeled videos, potentially to apply to more label-scarce fields in the future.
	General Flow as Foundation Affordance for Scalable Robot Learning Chengbo Yuan, Chuan Wen, Tong Zhang, Yang Gao. project page / arXiv / code / openreview / thread Conference on Robot Learning (CoRL), 2024 We build a 3D flow prediction model directly from large-scale RGBD human video datasets. Based on this model, we achieve stable zero-shot human-to-robot skill transfer in the real world.

Project

	MindQuantFinance: High-Performance Quantitative Asset Pricing Library A library for asset pricing (Q-quant) based on MindSpore gitee, 2023 This is an AI+Finance project I participated in, where we built a high-performance library for derivative pricing based on MindSpore. It offers features such as Black-Scholes calculations, Monte Carlo simulations, and solutions to Backward SDE.
	QFR: Tabular Prediction for Quantative Finance Research A tabular prediction library designed for financial stock prediction. github, 2023 This is one of the outcomes of an interesting experience, during which I worked with Prof. Jian Li on stock prediction problems. It provide rich tools for factors extraction and results prediction.