Nathan Herr

Papers

Total Citations

H-Index

About

Nathan Herr is a leading researcher at the intersection of robotics, computer vision, and large-scale machine learning. His work focuses on overcoming the fundamental data scarcity bottleneck in robot learning by leveraging diverse, internet-sourced video data. In his highly influential survey, “Towards Generalist Robot Learning from Internet Video” (2025), Herr systematically explores how scaling deep learning with massive, unstructured video—the same paradigm that revolutionized natural language processing and video generation—can be adapted to train generalist robots. This work has already garnered significant attention (5 citations in its first year), establishing him as a key voice in the push toward foundation models for robotics. Herr’s contributions are critical for enabling robots to learn flexible, reusable skills without costly, task-specific human demonstrations. His research promises to democratize robot learning, moving the field beyond narrow, lab-trained behaviors toward truly adaptive, real-world agents. For students and researchers, Herr’s work represents a vital roadmap for bridging the gap between internet-scale data and embodied intelligence.