Xavier Hickman, Yang Lu, Daniel Prince: Hybrid safe reinforcement learning: Tackling distribution shift and outliers with the Student-t's process. Neurocomputing 634: 129912 (2025)