Tracy Zhu

Hello! I am Tracy Zhu. I’m currently a research assistant working with Prof. Chen Wei at Rice University, where I study spatially aware vision encoders in vision language action models. My research interests lie broadly in Generative Models, Multimodal Learning, and Representation Learning.

I recently earned an M.S. in Statistics from University of Chicago, where I worked with Prof. Anand Bhattad during his time at TTIC. I received my B.A. in Data Science and Mathematics from NYU, where I was fortunate to work with Prof. Tim G.J. Rudner at University of Toronto and Prof. Marco Morucci at Michigan State University on uncertainty quantification for large language models.

I am applying for a PhD in CS/ML starting in Fall 2026. Please feel free to reach out if our research interests align!

Contact: yz5880 [at] uchicago [dot] edu

news

Oct 01, 2024	Our work Weak-to-Strong Confidence Prediction is accepted to four NeurIPS workshops!

selected publications

* denotes equal contributions

A Systematic Assessment of Weak-to-Strong Confidence Prediction in Large Language Models

Tracy Zhu^*, Yukai Yang^*, Marco Morucci, and 1 more author

Under review, 2025

Bib

@article{zhu2025W2S,
  title = {A Systematic Assessment of Weak-to-Strong Confidence Prediction in Large Language Models},
  author = {Zhu, Tracy and Yang, Yukai and Morucci, Marco and Rudner, Tim G. J.},
  journal = {Under review},
  year = {2025},
  url = {https://openreview.net/forum?id=xYSzkg5qPD&noteId=BiK8opYWHF}
}

Weak-to-Strong Confidence Prediction

Yukai Yang^*, Tracy Zhu^*, Marco Morucci, and 1 more author

Workshop on Statistical Foundations of LLMs and Foundation Models, Attributing Model Behavior at Scale, Safe Generative AI, and Regulatable ML (NeurIPS 2024 Workshop) , 2024

Bib

@article{zhu2024weaktostrong,
  title = {Weak-to-Strong Confidence Prediction},
  author = {Yang, Yukai and Zhu, Tracy and Morucci, Marco and Rudner, Tim G. J.},
  journal = {
      Workshop on Statistical Foundations of LLMs and Foundation Models, 
      Attributing Model Behavior at Scale, 
      Safe Generative AI, 
      and Regulatable ML <strong>(NeurIPS 2024 Workshop)</strong>
      },
  year = {2024},
  url = {https://openreview.net/forum?id=VvA4BImacV}
}

Making the Most Out of the Limited Context Length: Predictive Power Varies with Clinical Note Type and Note Section

Hongyi Zheng, Tracy Zhu, Lavender Yao Jiang, and 2 more authors

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics - Student Research Workshop , 2023