Ruizhe Shi



About me

I am a first-year Ph.D. student in Paul G. Allen School of Computer Science & Engineering at the University of Washington. Previously, I received my bachelor's degree in Computer Science (Yao class), with a minor in Literature, from Tsinghua University. I'm currently working on principled algorithm design and theoretical analysis of deep learning and foundation models. I'm also interested in combinatorics, and the application of AI in medicine.


Contact



Some papers

An Improved Upper Bound for Colorings Without Symmetrically Colored k-Term Arithmetic Progressions.
Ruizhe Shi, Yiqi Dong
Preprint

Understanding the Performance Gap in Preference Learning: A Dichotomy of RLHF and DPO.
Ruizhe Shi*, Minhak Song*, Runlong Zhou, Zihan Zhang, Maryam Fazel, Simon S. Du
ICML 2026

The Crucial Role of Samplers in Online Direct Preference Optimization. [A short note for improved analysis]
Ruizhe Shi*, Runlong Zhou*, Simon S. Du
ICLR 2025

Decoding-Time Language Model Alignment with Multiple Objectives.
Ruizhe Shi, Yifang Chen, Yushi Hu, Alisa Liu, Hannaneh Hajishirzi, Noah A. Smith, Simon S. Du
NeurIPS 2024

Rethinking Transformers in Solving POMDPs.
Chenhao Lu, Ruizhe Shi*, Yuyao Liu*, Kaizhe Hu, Simon S. Du, Huazhe Xu
ICML 2024


New notes, new thoughts

Palette sparsification paper reading. [slide]

One-step flow matching paper reading. [slide]

Understanding the gaps between two-stage and direct preference-based policy learning. [slide]

The crucial role of samplers in online direct preference optimization. [slide][recording]

Logit mixing and RLHF paper reading. [slide]

Decoding-time language model alignment with multiple objectives. [slide][recording]

An incomplete list of books I like, randomly maintained. [list]