Zhaoyang Yu

Zhaoyang Yu
于兆洋

Email: yuzhaoyang0713 [AT] ruc.edu.cn

GitHubZhihuTwitter

I received my undergraduate degree from Gaoling School of AI, Renmin University of China in June 2024. I'm actively seeking a PhD position at present. My research interest include efficiently building robust agent system and enhancing reasoning ability of agent.

Publications

MobileExperts: A Dynamic Tool-Enabled Agent Team in Mobile Devices
Jiayi Zhang, Chuang Zhao, Yihan Zhao, Zhaoyang Yu, Ming He, Jianping Fan*.
arXiv
2024
AFlow: Automating Agentic Workflow Generation
Jiayi Zhang, Jinyu Xiang, Zhaoyang Yu, Fengwei Teng, Xionghui Chen, Jiaqi Chen, Mingchen Zhuge, Xin Cheng, Sirui Hong, Jinlin Wang, Bingnan Zheng, Bang Liu, Yuyu Luo*, Chenglin Wu*.
arXivGithub
2024

Projects

Audio-Visual Navigation on CyberDog
- Integrated the capabilities of large-scale pre-trained multimodal models into robots, enabling them to perform diverse actions in different visual situations.
- Based on the SoundSpaces simulation environment data, a GRU-based sequential decision model was trained using PPO, with audio and images as input. The model is finally deployed on CyberDog.
Advised by Prof. Di Hu and Prof. Ruihua Song
GeWu-LabBilibili[1]Bilibili[2]
2022-2023
Sensory Substitution for the Blind (Public Welfare Projects)
- To make the visually disabled 'SEE' the world again (with the help of Beijing Disabled Persons' Foundation).
ZhihuGeWu-Lab
2022-suspended (I hope to restart this project)

Education

Renmin University of China
- B.E., Artificial Intelligence
2020-2024

Experience

GeWu-Lab
Research Intern
- Audio-Visual Navigation on CyberDog. Displayed as an excellent project on 2022 BAAI Conference.
2021.07-2022.08
Xiaomi - MiRoboticsLab
Algorithm Engineer
- Developed a versatile AI agent for CyberDog by creating reusable APIs with Python and ROS2; Based on ReACT to build an embodied agent on CyberDog2.0 with multi-modal input; Construct a dataset for CyberDog application scenarios and fine-tune MiLM.
2024.01-2024.04
AI Engine Lab
Research Intern
- Undergraduate Thesis: Multi-head Reflexion. Discussed three weaknesses of reflexion: randomness, stubbornness and misunderstanding. Desigend specialized reflexion head with personas to overcome these shortcomings.
2024.02-2024.06

Service

Microsoft Student Learning Ambassador
- Promoting MS techniques and products, Organizing and participating in MS events
2021-2023

Award

- Sa Shixuan Scholarship (Second Prize) in RUC, School of Information 2024
- Alibaba Global Mathematics Competition AI Challenge Top 3 (500+ teams) 2024
- The 15th National College Student Mathematics Competition Second Prize 2023
- Tencent Youth Game Designer Challenge Merit Award 2022
- MCM/ICM Contest Meritorious Winner (top 7%) 2022
- 2020-2021 Learning Excellence Scholarship in RUC 2020-2021

Misc

- Xiaomi CyberDog Emotional Development (Chinese) [Link] 2022
GitHub last commit