Yu Zheng (郑瑜)

I am currently a postdoctoral fellow at Massachusetts Institute of Technology. Previously, I completed my PhD at FIBLAB, Department of Electronic Engineering, Tsinghua University in 2024, advised by Prof. Depeng Jin and Prof. Yong Li. I received my bachelor degree from the Department of Electronic Engineering, Tsinghua University in 2019.

My research interests lie at the intersection of artificial intelligence and science, with a focus on developing intelligent decision-making methods for both offline physical systems and online information systems. I am broadly interested in AI for Science, Science of AI, RL and its applications in real-world complex systems.

AI and Science: Understanding emergent abilities of AI using science and advancing scientific discoveries with AI. I’m deeply committed to building scientific tools for understanding how intelligence arises from AI models and implicating future development of AI. I proposed neuroscience-informed graph probing that reveals the topology of neurons in LLMs and links it to their generative behavior. I’m passionate about inventing AI4Science models that drive theoretical advancements and generate scientific insights. I designed self-inductive symbolized RL that solves network dismantling problems and discovers theoretical formulas generalizable across biological, ecological, and economic systems.

RL and its applications: Optimizing decisions in complex systems. I’m interested in building intelligent decision-making frameworks to solve real-world challenges. I developed GNN-based RL methods that enable a wide range of applications in molecular generation, spatial planning, network control, and information retrieval.