Integrating contrastive learning and conservative Q-learning objectives can enhance the stability and performance of reinforcement learning-based sequential recommender systems.
