Alec Xu

Contact: alecx [at] umich [dot] edu

Hello! I'm a third-year PhD student in the Electrical and Computer Engineering Department at the University of Michigan, where I'm fortunate to be advised by Qing Qu as part of the DeepThink Lab. I also collaborate closely with Laura Balzano. Previously, I earned my MSE and BSE degrees also at the University of Michigan, where I worked with Jeff Fessler.

My current research focuses on leveraging low-dimensional structure in deep learning to advance their theoretical foundations. I'm also broadly interested in efficiency of large models via model compression, as well as interpretability of large models through their internal representations.

CV (Updated Jan. 2026) | Google Scholar | LinkedIn

Select Publications

• Soo Min Kwon*, Alec S. Xu*, Can Yaras, Laura Balzano, Qing Qu. Out-of-Distribution Generalization of In-Context Learning: A Low-Dimensional Subspace Perspective. AISTATS 2026.
• Alec S. Xu, Can Yaras, Peng Wang, Qing Qu. Understanding How Nonlinear Layers Create Linearly Separable Features for Low-Dimensional Data. AISTATS 2026.
• Can Yaras, Alec S. Xu, Pierre Abillima, Changwoo Lee, Laura Balzano. MonarchAttention: Zero-Shot Conversion to Fast, Hardware-Aware Structured Attention. NeurIPS 2025 (Spotlight).
• Alec S. Xu, Nina I. Shamsi, Lars A. Gjesteby, Laura J. Brattain. Self-Supervised Edge Detection Reconstruction for Topology-Oriented 3D Axon Segmentation and Centerline Detection. WACV 2024.
• Alec S. Xu, Laura Balzano, Jeffrey A. Fessler. HeMPPCAT: Mixtures of Probabilistic Principal Component Analysers for Data with Heteroscedastic Noise. ICASSP 2023.