Seunghyun Yoo

Undergraduate Affiliate

Seunghyun is a second year undergraduate student majoring in Computer Science and Artificial Intelligence with a minor in Mathematics at Purdue University. He is interested in large language model safety benchmarking and mechanistic interpretability. At GRAIL, he works on the AGORA project, where he analyzes why large language models produce hallucinations and incorrect responses in the context of AI Governance Law. His research focuses on mechanistic analysis, particularly attention layers, and he explores steering techniques to suppress such behaviors when possible.