Data-centric AI
Improving learning through better data quality and understanding.
Hello! We are a research group at the University of Wisconsin-Madison focused on machine learning, LLMs and Foundation Models, and data-centric AI.
Research Topics
Data-centric AI
Data- and Compute-Efficient Learning
Building models that require less data and computational resources.
Data- and Compute-Efficient Learning
Foundation Models
Foundation models and their applications to scientific problems.
Foundation Models
Weak Supervision
Learning with limited or noisy supervision.
Weak Supervision
News
February 2025
New papers accepted to start the year!
- ICLR 2025: Changho and John explain how weak-to-strong generalization works—and how to do more of it!
- NAACL 2025: Jane, Dyah, and Changho introduced an ultra-efficient way to personalize language models.
December 2024
Four new papers accepted at NeurIPS 2024!
- Brian, Cathy, and Vaishnavi show how to get rid of the LLM in LLM-based annotation. How? Distill LLMs into programs (spotlight)!
- Harit gets a huge boost in auto-labeling by learning confidence functions.
- Changho, Jitian, and Sonia show how to adjust zero-shot model predictions quickly and easily
- Chris and Jack introduce a new benchmark showcasing how valuable weak supervision can be
Sponsors
We are thankful for the generous funding award and gift from the following
