Alexander Robey

California University of Pennsylvania, Carnegie Mellon University

Papers

Total Citations

H-Index

About

Alexander Robey is an emerging researcher at the intersection of machine learning theory, robotics, and AI safety, with particular expertise in imitation learning, control systems, and the secure integration of large language models into physical systems. His foundational work on stability-constrained imitation learning — his most cited contribution with 17 citations — broke new theoretical ground by demonstrating how the stability properties of expert policies directly shape the sample complexity of learning tasks, offering the first rigorous analysis of this nuanced connection in continuous control settings. More recently, Robey has turned his attention to the rapidly evolving challenges posed by LLM-enabled robotics. His influential research on jailbreaking LLM-controlled robots has exposed critical vulnerabilities in systems spanning manipulation, locomotion, and autonomous vehicles, accumulating over a dozen citations across multiple iterations of the work. Complementing this, his research on safety guardrails for LLM-enabled robots proposes principled defenses against both everyday model failures and adversarial attacks, addressing one of the most pressing open problems in deploying intelligent robots responsibly. Collectively, Robey's portfolio reflects a rigorous, safety-conscious approach to modern AI systems, making his work essential reading for researchers navigating the frontier of trustworthy autonomy.