Robotics: VLAs & World Models

​Foundation models transformed vision, language, and coding — and robotics is now entering a similar era. Two major paradigms are emerging at the center of modern robot learning: Vision-Language-Action (VLA) models and world models.

​In this session, two UROP students will give short ~10-minute presentations exploring recent advances in robotic foundation models, scaling laws, data composition, and the rise of world models for planning and generalization. The talks will synthesize insights from recent papers across academia and industry and discuss the key trade-offs, open questions, and future directions shaping the frontier of robotics research.