Empowerment under Uncertainty: Bayesian Exploration in Model-Based RL | Oliver Obst | IMOL 2025

Oliver Obst presents “Empowerment under Uncertainty: Approximate Bayesian Exploration in Model-Based RL” at IMOL 2025. The title used in the presentation is “Bayesian Approximate Empowerment for RL Exploration (BAERLE).” The talk considers approximate Bayesian empowerment as an exploration mechanism for model-based reinforcement learning. Speaker: Oliver Obst, Western Sydney University Co-authors: Fabian C. Weigend, Harvard University; Lisa Graf, University of Freiburg Recorded: 8 September 2025 Event: The 7th International Workshop on Intrinsically Motivated Open-ended Learning (IMOL 2025), University of Hertfordshire, Hatfield, UK Workshop: https://imol2025.github.io/ Programme: https://imol2025.github.io/pages/sche... Captions: AI-generated, checked by the organisers, and still being improved. #IMOL2025 #ReinforcementLearning #Empowerment