Zexuan Zhong

zzhong@cs.princeton.edu

     

About

I completed my Ph.D. at Princeton University, advised by Prof. Danqi Chen. Before that, I received an M.S. from University of Illinois at Urbana-Champaign and a B.S. from Peking University.

I am working at xAI now.

I am deeply interested in natural language processing (NLP) and deep learning. My recent research has focused on:

  • Semi-parametric / Sparse Language Models: We study approaches to build highly efficient, effective, and updatable language models by leveraging retrieval augmentation [1,2,3,4], conditional computations [5], or the model sparsity [6].
  • Language Models and World Knowledge: We show how to extract knowledge from text using LMs [7], recall knowledge from LMs [8], enhance LMs with external knowledge [1], and update knowledge in LMs [3].
  • Robustness and Generalization: We investigate the vulnerability and resilience of ML models against adversarial attacks [9,10], risks of privacy leakage [11,12], and distribution shifts [13].

Papers     (show full / show selected)

(* indicates equal contribution)

Experience

  • 2023.6 - 2023.12, Research Intern, Meta
    Mentor: Mike Lewis
  • 2018.5 - 2018.8, Visiting Research Assistant, MIT Media Lab
    Advisor: Fadel Adib
  • 2017.9 - 2019.5, Research Assistant, University of Illinois at Urbana-Champaign
    Advisor: Tao Xie
  • 2016.7 - 2017.5, Research Intern, Microsoft Research Asia
    Mentor: Zaiqing Nie

Selected Honors