doi.bio/winnie_xu

Winnie Xu

Early Life and Education

Winnie Xu is a researcher with a background in biochemistry and genomics. Xu received her B.S. in Statistics from South China Normal University in 2015. Prior to joining Princeton, Xu worked on studying bioinformatics at BGI-Shenzhen, where she also conducted research on population genetics and maternal-fetal medicine based on next-generation sequencing technologies.

Career

Xu is currently a researcher at the Lewis-Sigler Institute for Integrative Genomics at Princeton University. Her research interests include computational biology and human population genetics.

Xu has previously worked as a Student Researcher at Facebook AI Research (Meta AI) and collaborated with researchers at Stanford StasML and Google Brain. She also studied Artificial Intelligence Computer Science, Statistics, and Math at the University of Toronto, where she began her AI research journey by scaling methods in latent variable modeling and probabilistic inference.

Publications

Xu has published several papers in the fields of machine learning, AI, and genomics. Her recent work includes:

KTO: Model Alignment as Prospect Theoretic Optimization (2024)
Revisiting Associative Compression: I Can’t Believe It’s Not Better (2023)
Multi-Game Decision Transformers (2022)
Self–Similarity Priors: Neural Collages as Differentiable Fractal Representations (2022)
Language Model Cascades (2022)
Infinitely Deep Bayesian Neural Networks with Stochastic Differential Equations (2022)
Noisy Feature Mixup (2022)
Prioritized training on points that are learnable, worth learning, and not yet learned (2022)
Exploring the Differential Effects of Sequencing Resolution on Semi-Automated Genome Annotations (2018)