Research Overview
My current research focuses on how we can characterize and control the behavior of astonishingly complex AI systems, particularly those driven by Large Language Models. This is especially pertinent in areas beyond supervised learning, such as employing Reinforcement Learning and Game Theory to help align LLMs in context of Trustworthy AI or enhance their capabilities in long-horizon reasoning for LLM agents.
I am also interested in Multimodal Models and embodied AI, particularly in how these systems perceive and model the world.
Publications
|
Red Teaming Game: A Game-Theoretic Framework for Red Teaming Language Models
Chengdong Ma*,
Ziran Yang*,
Minquan Gao,
Hai Ci,
Jun Gao,
Xuehai Pan,
Yaodong Yang
Under Review
|
|
Panacea: Pareto Alignment via Preference Adaptation for LLMs
Yifan Zhong*,
Chengdong Ma*,
Xiaoyuan Zhang*,
Ziran Yang,
Qingfu Zhang*,
Siyuan Qi*,
Yaodong Yang*,
Under Review
|
Before 2022
|
Ministry of Education Talent Program Thesis
I was selected for the Ministry of Education Talent Program and conducted research in the math track guided by Falai Chen
at University of Science and Technology of China (USTC) during my high school years.
I was also selected as the Outstanding Thesis of the National Math Forum 2019 of the Ministry of Education Talent Program.
|
Selected Awards
2024: Fifth Yuanpei Young Scholar Award
2023: Peking University Institute for Artificial Intelligence Annual Technology Day: Best Innovation Award
2023: Peking University Learning Excellence Award
2023: Peking University Shu Qi Scholarship
2022: Peking University Learning Excellence Award
2022: Peking University Lee Wai Wing Scholarship
2021: Peking University Freshman Scholarship
2020: Chinese High School Mathematics League: First Prize
2019: Ministry of Education Talent Program: annual Outstanding Thesis
|
|