SOMBRL: Scalable and Optimistic Model-Based RL

We address the challenge of efficient exploration in model-based reinforcement learning (MBRL), where the system dynamics are unknown and the RL agent must learn directly from online interactions.