Yuhan Cao
caoyh1@shanghaitech.edu.cn / yuhanteafrog@gmail.com

CONNECT THE WORLD.
Currently, I am:
- Research Intern @ Shanghai Qi Zhi Institute, advised by Prof. Tianxing He
- Undergraduate Researcher @ The ShanghaiTech Multi-Agent systems Research Team (SMART), ShanghaiTech University, advised by Prof. Dengji Zhao
Previously, I was:
- Research Intern @ Microsoft Research Asia, advised by Dr. Dongsheng Li
- Research Intern @ Prof. Xiaotie Deng’s daGAME Lab, Peking University, advised by Dr. Zhaohua Chen
- Visiting Student @ University of Wisconsin - Madison
- Teaching Associate @ New York University Shanghai
I am actively looking for a Ph.D./RA position in Fall 2025.
Hi there. My research interest is about CONNECTION. I am studying Large Language Models (LLMs) and Multi-Agent System (MAS).
My current research interest:
- Are LLM-based multi-agent systems sufficiently truthful, secure and efficient? How can we design valuable mechanisms to ensure this (both theoretically and empirically)?
- How can we examine or reflect on LLMs (and multi-agent systems) from an anthropological perspective? Can we construct a culture for multi-agent systems? Furthermore, how can we evaluate the real value of highly metric-driven multi-agent systems for humans?
- How can we attack the reasoning process of LLMs? How can we prove the inherent flaws (if they exist) in reasoning models? How can we distinguish different error modes in various LLMs? Furthermore, can competitive-level problems (including Olympiad Informatics problems) contribute to these studies?
Topics that may not be easy to get started in the short term, but I am still interested:
- How can LLMs truly engage in meaningful role-playing or writing? These two tasks are similar in that they both reflect the linguistic state of a real, specific human.
- How do language models model language? Can we understand the representation of (even multimodal) linguistic concepts in language models through probing models? Can modern and contemporary philosophy of language (continental or analytic) help reshape our understanding of these questions?
If you are also interested in these directions and would like to collaborate with me, please contact me immediately!
Here are my research experience and details.
I works on how multi-agent systems connect with human society. I am exploring the evolution of competition and cooperation in LLM-based multi-agent systems (with or without human-computer interactions). At the same time, I am deeply interested in aligning large-scale multi-agent systems with human society. This includes, but is not limited to, the intersection of sociology, anthropology, economics, psychology, philosophy, and LLMs. I believe it is crucial to study whether language models can genuinely take on roles within a real social context.
I am also intrigued about how to design (incentive-compatible) mechanisms on graphs or multi-agent systems, which mainly falls within the field of Algorithmic Game Theory (AGT). My current work revolves around exploring the intersection of LLMs and AGT, with an emphasis on mechanism design for multi-agent systems.
Previously I worked on how to capture the relationships between different variables using deep learning methods. I was mainly focusing on Electroencephalogram (EEG) and Graph Neural Networks (GNN). I am also interested in the use of graph neural networks for the analysis of multi-variable time series data, chemical molecules, social networks and so on.
I love teaching and have been involved in the design of several courses as a TA. I am doing my best to reduce inequalities in course teaching and help everyone gain the same knowledge as much as possible.
I am very passionate about communicating with people. If you would like to discuss our shared research interests, or if you are my student (in SI 100B, SI 100+, or other courses) and want to chat for advice (though I am just getting started on many things myself), feel free to reach out anytime. You can also book a time with me through Calendly.
I like (non-fiction) writing and composing/mixing music.
news
May 15, 2024 | Happy to see you in WWW 2024 @ Singapore! |
---|---|
Feb 23, 2024 | Happy to see you in AAAI 2024 @ Vancouver! |
Dec 9, 2023 | One paper was accepted in AAAI 2024 (Oral Presentation Received). |
Jul 23, 2023 | We are now establishing SI 100+ Saving Project. |