Haoran (Jensen) Yuan

PhD Student · Computer Engineering · University of Illinois Urbana-Champaign

My research focuses on efficient AI, especially on how to make embodied AI generalizable, scalable, and efficient. I build visuo-tactile world-action models that adapt pretrained video models to touch without separate tactile pretraining, letting robots reason about contact that cameras cannot see. I also develop lossless parallel drafting methods that speed up large-model inference several times without changing its outputs. I am also interested in representation learning and recursive self-improvement (RSI), and I am currently exploring agentic robotics and reinforcement learning for world-action models.

I am a PhD student in PLAN Lab at UIUC, advised by Prof. Ismini Lourentzou. I have also worked with BAIR at UC Berkeley and Hao AI Lab at UC San Diego. Before my PhD, I received an M.S. in Computer Engineering from UIUC, a dual B.S. from UIUC and Zhejiang University, and interned as a software engineer at Snowflake.