Bradly C. Stadie

University of California, Berkeley, Science North

Papers

Total Citations

272

H-Index

About

Bradly C. Stadie is a leading researcher in robotics and artificial intelligence, whose work is redefining how machines learn from humans and plan in complex environments. His most influential contribution, "One-Shot Imitation Learning" (2017, 229 citations), pioneered a paradigm where robots can master new tasks from a single demonstration, dramatically reducing the data burden of traditional imitation learning. This breakthrough has become a cornerstone for sample-efficient robotic training. Stadie has since advanced the field of planning and shared autonomy. His work on "World Model as a Graph" introduces latent landmarks for structured, long-horizon planning, while "To the Noise and Back" leverages diffusion models for collaborative human-robot control. More recently, his "Wonderful Team" framework (2024) achieves zero-shot physical task planning by orchestrating multiple Vision-Language Models, enabling robots to reason about novel environments without prior training. Across these projects, Stadie consistently tackles the core challenge of enabling robots to generalize from minimal data—whether through imitation, planning, or human-robot interaction. His research sits at the intersection of imitation learning, model-based planning, and human-robot collaboration, making him a key figure in the next generation of intelligent, adaptable robotic systems.