Sebastian Junges
University of California, Berkeley, Radboud University Nijmegen
Papers
5
Total Citations
101
H-Index
5
About
Sebastian Junges is a leading researcher at the intersection of formal verification, probabilistic modeling, and safe artificial intelligence. His work centers on ensuring the reliability of autonomous systems operating under uncertainty, with a particular focus on safe reinforcement learning and strategy synthesis. Junges is best known for pioneering the concept of "probabilistic shields"—efficient, formally verified safety mechanisms that constrain reinforcement learning agents to prevent catastrophic failures during training and deployment. His foundational 2020 paper on the topic has garnered 52 citations, establishing a new paradigm for trustworthy AI. Beyond safe RL, Junges has made significant contributions to abstraction-refinement techniques for hierarchical probabilistic models and strategy synthesis for partially observable Markov decision processes (POMDPs), with applications ranging from robot planning to human-robot interaction. His work bridges the gap between theoretical verification and practical deployment, enabling complex cognitive systems to operate safely alongside humans. With over 100 total citations and a rapidly growing influence, Junges is shaping how next-generation autonomous systems are designed, verified, and trusted.
Research Focus
Key Achievements
Top Papers
- 1Safe Reinforcement Learning Using Probabilistic Shields52 citations · 2020
- 2Safe Reinforcement Learning via Probabilistic Shields16 citations · 2018
- 3Abstraction-Refinement for Hierarchical Probabilistic Models13 citations · 2022
- 4Strategy Synthesis for POMDPs in Robot Planning via Game-Based Abstractions13 citations · 2020
- 5