A Survey on the Verification of Reinforcement Learning Policies
For researchers and practitioners in safety-critical RL, this survey organizes a fragmented field, making it easier to understand and compare verification approaches.
This survey provides a unifying taxonomy and theoretical foundation for reinforcement learning policy verification, categorizing methods along three axes and identifying emerging directions.
Reinforcement learning (RL) is increasingly applied in complex, safety-critical domains, yet the lack of rigorous behavioral guarantees for neural network-based policies remains a major barrier to deployment. Recent advances in policy expressiveness and scale have intensified this challenge, leading to a rapidly growing but conceptually fragmented body of work on RL policy verification. This survey provides a unifying perspective on RL verification methods. We introduce a taxonomy that clarifies relationships among existing approaches along three axes: verification paradigm (formal versus probabilistic), temporal scope (step-wise versus multi-step), and guarantees strength. Beyond taxonomy, we unify underlying theoretical foundations, make implicit assumptions and limitations explicit, and identify emerging directions.