The field of retrieval-augmented generation is moving towards more efficient and effective methods of incorporating external knowledge into large language models. Researchers are exploring new ways to optimize the retrieval process, such as using graph-based methods and reinforcement learning, to improve the accuracy and diversity of generated text. A key focus area is the development of methods that can handle complex, real-world queries and temporal knowledge graphs. Noteworthy papers in this area include: Prior Makes It Possible, which characterizes the necessary and sufficient number of augmentation steps for a model to generate an accurate answer given partial prior knowledge. GraphFlow, which efficiently retrieves accurate and diverse knowledge required for real-world queries from text-rich knowledge graphs. STAR-RAG, which proposes a temporal GraphRAG framework that relies on building a time-aligned rule graph and conducting propagation on this graph to narrow the search space and prioritize semantically relevant, time-consistent evidence. LoongRL, which introduces a data-driven reinforcement learning method for advanced long-context reasoning that induces an emergent plan-retrieve-reason-recheck reasoning pattern.