About
We develop algorithms and mathematical foundations for machine learning, optimization, and control. Our research spans online learning and decision-making, dynamical systems and sequence modeling, and the evaluation and alignment of increasingly capable AI systems.
Our work on AI evaluation and alignment is part of Princeton AI Alignment and Safety (AIAS). We also collaborate with Google DeepMind Princeton.
See our research, meet the team, or read our blog.
Research
- AI evaluation and alignment: measuring capabilities beyond human expertise and designing incentives for reliable behavior.
- Online learning and optimization: regret minimization, adaptive optimization, and decision-making with bandit feedback.
- Control and sequence modeling: online nonstochastic control, spectral methods, and efficient prediction over long sequences.
From the blog
-
Two decades of online convex optimization: a personal recollection — September 16, 2026
-
The Statistical Orthogonality Thesis — June 27, 2026
-
AI Alignment via Incentives and Correction — May 7, 2025