I’m a second year PhD Student at UC San Diego interested in generative modeling. I find inspiration for new ideas in simplicity and human perception, and I would like to understand the fundamental reason why our generative modeling paradigms work and how they can improve. My recent work has focused on removing unnecessary priors for denoising video generation models, and on dynamic memory in world models.
I’m fortunate to be advised by Profs. Rose Yu and Taylor Berg-Kirkpatrick at UCSD. I also am actively collaborating with Prof. Yilun Du. Read more about my Publications and Research Interests if you are curious! ~
Recent Highlights
Link to original
Equilibrium Forcing: Adaptive Video Generation Without Noise Conditioning
Hansen Jin Lillemark*, Alex Rojas*, Zachary Novack, Runqian Wang, Yilun Du, Yian Ma, Taylor Berg-Kirkpatrick, Rose Yu
Arxiv 2026
project page / paper / code (available on request) / bibtex
Flow Equivariant World Models: Structured Memory for Dynamic Environments
Hansen Jin Lillemark*, Benhao Huang*, Fangneng Zhan, Yilun Du, T. Anderson Keller
International Conference on Machine Learning (ICML) 2026, 4th NeurReps Workshop @ NeurIPS 2025
project page / paper / code / bibtex
I enjoyed Summer 2025 as a visiting PhD student at Harvard advised by Yilun and Andy Keller. Previously I was advised by Prof. Zhiting Hu at UCSD, by Prof. Bruno Olshausen at UC Berkeley’s Redwood Center for Theoretical Neuroscience, and by Prof. Kurt Keutzer at BAIR. I have also been a founding research scientist at New Theory AI, and in a past life, I was the co-founder and CTO of a financial technology startup. I received my bachelor’s degree in Computer Science from UC Berkeley.
Apart from research, I like to try new recipes, go on long runs, and read books on various topics. Check out my Library and Misc obsessions!
