Qizhen Zhang (Irene)

I am a second year machine learning PhD student at the University of Oxford, where I work on large language models and reinforcement learning. I'm also spending half of my time doing research at Cohere.

Prior to my PhD, I was a member of technical staff at Cohere building frameworks and training LLMs. I wrote my Master's thesis on cooperative multi-agent reinforcement learning at the University of Toronto and the Vector Institute.

Email  /  Tweets  /  Google Scholar

profile photo

Publications

project image

Analysing the Sample Complexty of Opponent Shaping


Kitty Fung*, Qizhen Zhang*, Chris Lu, Jia Wan, Timon Willi, Jakob Foerster
International Conference on Autonomous Agents and Multiagent Systems (AAMAS). Full paper, Oral Presentation, 2024
paper /

We present R-FOS, a tabular opponent shaping algorithm. We derive a sample complexity bound for R-FOS that is exponential in the cardinality of the inner state and action space and the number of agents. We also empiracally investigate how R-FOS’s sample complexity scales in the size of state-action space.

project image

Centralized Model and Exploration Policy for Multi-Agent RL


Qizhen Zhang, Chris Lu, Animesh Garg, Jakob Foerster
International Conference on Autonomous Agents and Multiagent Systems (AAMAS). Full paper, Oral Presentation, 2022
paper / talk

We propose a model-based method for fully cooperative multi-agent settings (Dec-POMDPs). Our method learns a centralized model, and is up to 20x more sample efficient in three commuication tasks. We also show theoretical sample complexity bounds for model-based methods learning in tabular Dec-POMDPs.




Teaching

I was a teaching assistant for the following courses.

cs

CSC311: Introduction to Machine Learning

CSC413/2516: Neural Networks and Deep Learning

CSC384: Introduction to Artificial Intelligence





Template from Jon Barron