Tianjian Li

Center for Language and Speech Processing, Johns Hopkins University

prof_pic.jpg

Hi👋, I’m Tianjian! I’m a 1st/3rd year PhD student in Computer Science at Johns Hopkins University, proudly advised by Prof. Daniel Khashabi. Previously, I completed my Master’s degree in Computer Science at JHU, where I had the privilege of working closely with my wonderful advisors, Kenton Murray and Philipp Koehn, focusing on multilingual NLP.

My research interest lies at the intersection of machine learning and natural language processing, with a particular focus on addressing the question: how can we better leverage our vast amount of data beyond simply feeding it into our models during training? To this end, I am currently working on measuring various properties of data (e.g. quality) and curating data for training and aligning our language models.

I prefer solutions that are simple, generalizable, and theoretically sound.

If you have anything to share with me, please feel free to contact me through my email: tli104 at jhu.edu

news

Oct 4, 2024 New preprint on how to train on heavily imbalanced datasets!!
Apr 7, 2024 I will be staying at Johns Hopkins University for my PhD, working with Prof. Daniel Khashabi!
Jan 15, 2024 Error Norm Truncation has been accepted to ICLR 2024 (spotlight) !!
Nov 8, 2023 New blog post on latest advances on balanced training for Multilingual Machine Translation!
Oct 2, 2023 New preprint on truncating noisy data for training text generation models!!

selected publications

  1. preprint
    Upsample or Upweight? Balanced Training on Heavily Imbalanced Datasets
    Tianjian Li, Haoran Xu, Weiting Tan, and 2 more authors
    2024
    arXiv
  2. preprint
    Verifiable by Design: Aligning Language Models to Quote from Pre-Training Data
    Jingyu Zhang, Marc Marone, Tianjian Li, and 2 more authors
    2024
    arXiv
  3. ICLR
    Error Norm Truncation: Robust Training in the Presence of Data Noise for Text Generation Models
    Tianjian Li, Haoran Xu, Philipp Koehn, and 2 more authors
    In The Twelfth International Conference on Learning Representations
    (Spotlight - Top 5%)    , 2024
    arXiv
  4. ACL
    Why Does Zero-shot Cross-lingual Generation Fail? An Explaination and A Solution
    Tianjian Li, and Kenton Murray
    In Proceedings of the 2023 Annual Meeting of the Association for Computational Linguistics (Findings), Jul 2023
    arXiv