Yuyang Jiang
MS student
Department of Statistics, University of Chicago
Email: [email protected]
My journey into human-centered AI development has taken many turns, and I am especially grateful to all of the mentors and collaborators who have kindly guided and inspired me along this way. Specifically, I have had the honor of engaging in research with Prof. Hua Shen (Department of Computer Science, NYU Shanghai & NYU; BiAlign Lab) on bidirectional human-LLM alignment, co-advised by Prof. Tanu Mitra (Information School, UW); with Prof. Chenhao Tan (Department of Computer Science, UChicago; Chicago Human+AI Lab) on human-centered NLP; with Prof. Samuel G. Armato (Department of Radiology, UChicago) on clinical NLP; and with Prof. Dacheng Xiu (Booth School of Business, UChicago) on financial NLP.
Before confirming my passion, I first earned my B.Econ. in Economics and Mathematics from China Economics and Management Academy at CUFE (Beijing, China, 2023), where I focused on modeling human behavior at both micro and macro levels. To deepen my expertise in data-driven methods, I later joined Department of Statistics for an M.S. in Statistics at UChicago (Chicago, IL, 2025), where I have been strengthening my theory foundations in machine learning and deep learning.
In my spare time, I enjoy yoga 🧘♀️, tennis 🎾, arts 🖼️ (painting, visual symbolism, exhibitions, etc.), and traveling 🗺️.
Research Focus: Human-Centered AI Development
- From AI: Aligning LLM/VLLM for Responsible Human Use
- Evaluate risks and biases in state-of-the-art LLM/VLLM systems to support responsible deployment.
- Construct and curate high-quality, well-structured datasets (embedded with human reasoning) for auditing LLM/VLLM performance.
- Post-train LLM/VLLM behavior to ensure alignment, particularly in high-stakes domains such as healthcare.
- From Human: Empowering Humans for Better Interaction with LLMs
- Study the bi-directional and temporal impact between humans and LLM-powered systems.
- Build reliable human-AI interactive applications that safely serve target users and enhance human performance.