About

Hi! I’m Irena, a student at Stanford University studying Computer Science (B.S. with honors) and Statistics (M.S.), graduating in June 2023.

In my research, I’m driven to understand out-of-distribution generalization: how can we train and adapt machine learning models such that they generalize beyond their training distributions? In particular, my interests center on:

  • Robustness. Models can fail unexpectedly and inequitably when distributions shift within a task, e.g., over time or data sources. How do we ensure that ML systems perform reliably in the real world?
  • Transfer across tasks. Models that generalize beyond a pretraining distribution to new task distributions are especially useful. How do we design pretraining datasets that enable broad transfer?

I’m especially excited about the intersection of these topics with interactive learning.

Outside of research, I love Art History! Here’s a random sample of a favorite work.

Selected Publications

  1. Adaptive Testing of Computer Vision Models
    Irena Gao, Gabriel Ilharco, Scott Lundberg, and Marco Tulio Ribeiro
    In submission, 2023.
    arXiv
  2. CREPE: Can Vision-Language Foundation Models Reason Compositionally?
    Zixian Ma*, Jerry Hong*, Mustafa Omer Gul*, Mona Gandhi, Irena Gao, and Ranjay Krishna
    In submission, 2023.
    arXiv
  3. Out-of-distribution Robustness via Targeted Augmentations
    Irena Gao, Shiori Sagawa, Pang Wei Koh, Tatsunori Hashimoto, and Percy Liang
    In NeurIPS Workshop on Distribution Shifts 2022.
  4. Extending the WILDS benchmark for unsupervised adaptation
    Shiori Sagawa*, Pang Wei Koh*, Tony Lee*, Irena Gao*, Sang Michael Xie, Kendrick Shen, Ananya Kumar, Weihua Hu, Michihiro Yasunaga, Henrik Marklund, Sara Beery, Etienne David, Ian Stavness, Wei Guo, Jure Leskovec, Kate Saenko, Tatsunori Hashimoto, Sergey Levine, Chelsea Finn, and Percy Liang
    In International Conference on Learning Representations 2022. (Oral Presentation)
    arXiv Website
  5. WILDS: A benchmark of in-the-wild distribution shifts
    Pang Wei Koh*, Shiori Sagawa*, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, Tony Lee, and others
    In International Conference on Machine Learning 2021. (Long Talk)
    arXiv Website

* denotes equal contribution.