The field of social media graph analytics is moving towards the development of more robust and versatile models that can handle diverse tasks and datasets. Researchers are focusing on integrating propagation-aware representation learning, kinetic-guided propagation modules, and Gaussian mixtures to model evolving multi-modal beliefs and opinion uncertainty. Additionally, there is a growing interest in using large language models (LLMs) and dynamic simulation frameworks to model rumor propagation and identify echo chambers. Noteworthy papers in this area include: RumorSphere, which presents a novel dynamic and hierarchical social network simulation framework that supports simulations with millions of agents, demonstrating a strong alignment between simulated and real-world rumor dynamics. CleanNews, which proposes a comprehensive architecture to identify fake news in real-time accurately, using advanced deep learning architectures and a novel embedding technique that fuses textual information with user network structure.