Optimization Dynamics for ML Behaviors
Understand and design training, post-training, and test-time optimization algorithms to select, stabilize, and steer intelligent behaviors.
Our research pillars
Foundations of RL & Multi-Agent AI
Develop theory and algorithms for AI systems that learn and reason from feedback under uncertainty, as well as coordinate, compete, and act under strategic interactions.
Generative Models for Planning and Control
Use diffusion, flow, and foundation models as structured priors for decision-making, scientific design, and controllable generation.