The recent advancements in the research area of reinforcement learning (RL) and service systems have shown significant progress in addressing longstanding challenges and introducing innovative methodologies. In RL, the focus has shifted towards enhancing exploration strategies, leveraging intrinsic rewards, and integrating prior knowledge from demonstrations to accelerate learning. Novel approaches such as Temporally Correlated Latent Exploration and Bounded Exploration with World Model Uncertainty have demonstrated robustness against stochastic environments and improved performance in various benchmarks. Additionally, methods like Skill-Enhanced Reinforcement Learning Acceleration from Demonstrations have shown promise in overcoming data scarcity issues by utilizing skill-level adversarial learning and data enhancement techniques. On the service systems front, there has been a notable emphasis on estimating customer behavior and market dynamics in complex systems, with innovative solutions for handling unobserved variables and optimizing operational decisions. Predictive models for subway passenger flows under incident situations have also advanced, incorporating causality to enhance interpretability and accuracy. Overall, these developments indicate a trend towards more sophisticated and adaptive models that integrate multiple learning paradigms and leverage causal insights for better decision-making in dynamic environments.