Ruiyang Zhou

Hello 🙌 I am a 2nd year Ph.D. student at the HUMAIN Lab in the IROM department at UT Austin, advised by Prof. Leqi Liu. Prior to pursuing my Ph.D., I obtained my M.S. and B.S. in Information Engineering from Shanghai Jiao Tong University, China and a double M.S. degree in Data Science & Applied Math in Télécom Paris, France.

My research goal is to understand and build user-friendly / control LLMs for real-world applications. On the understanding side, our ExPO paper explains the importance of in-distribution samples and develops a method to boost RL training on hard reasoning tasks. My ongoing project is also related to LLM understanding: using sparse autoencoders to build market map directly from LLM internals. On the controlling side, the P-RLHF paper trains the model to be personalized and align with user-interests. The Levenshtein Transformer project tries to incorporate user feedback in the real time into model outputs.

Still on my way to make them concrete and I am excited of this dynamic research area!