Theory
- Study how task structure, data recipes, and inference algorithms shape generalization beyond training experience.
- Characterize the abstractions that make learning, reasoning, and planning efficient.
Research
Our research goal is to understand the principles that enable machines to continually acquire concepts from experience and use them to reason and plan in open-ended physical worlds.
// Themes
System
- Build systems that capture useful structure for inference: objects, relations, programs, geometry, physics, and action abstractions.
- Use search, optimization, and probabilistic reasoning to manipulate these abstractions.
Applications
- Apply structured learning and reasoning to embodied agents, robot manipulation, and physical scene understanding.
- Study applications in language and cognitive science, including human language emergence, evolution, and acquisition, and human problem solving in physical domains.