Qinyuan Wu

I am Qinyuan Wu (吴沁媛), a final year PhD student at the CS@Max Planck and the Max Planck Institute for Software Systems (MPI-SWS), advised by Krishna Gummadi and Muhammad Bilal Zafar (Ruhr University Bochum). I am also fortunate to closely collaborate with and receive guidance from Abhilasha Ravichander (MPI-SWS), Evimaria Terzi (Boston University), Mariya Toneva (MPI-SWS). Before I joined MPI-SWS, I got my bachelor’s degree in mathematics-physics from University of Electronic Science and Technology of China (UESTC).

My research takes AI agents as objects of study, with a particular emphasis on LLM-based agents, examining how they acquire knowledge, form memories, and use what they know to act. Through this lens, I seek to understand the principles underlying intelligent behavior and how these systems can become more reliable, interpretable, and aligned with human needs.

My research investigates how AI models internalize, structure, and utilize knowledge to execute reliable actions. Specifically, I study how internal parametric knowledge interacts with dynamic runtime environments—such as context windows, memory structures, and external tools—to bridge the gap between descriptive model behavior and normative frameworks. More broadly, I aim to understand and enhance the end-to-end loop of how intelligent systems learn, remember, retrieve, and act, advancing toward more trustworthy, interpretable, and human-centered AI.

For more detail on my research and future directions, see my research statement.

I also actively collaborate on interdisciplinary projects covering:

• Privacy & Security in AI: Balancing data protection with system utility and operational efficiency.
• Neuroscience-Inspired Modeling: Exploring parallels between human memory architectures and artificial cognition.