The recent developments in the field of Retrieval Augmented Generation (RAG) systems have significantly advanced the integration of multimodal data and fine-grained attribution mechanisms. Researchers are focusing on enhancing the contextual understanding of retrieved evidence by incorporating source metadata and surrounding text, as well as developing causal explanation approaches like counterfactual attribution. These advancements aim to improve both the retrieval and answering quality in conversational Question Answering (ConvQA) systems. Additionally, there is a growing emphasis on multimodal Retrieval Augmented Generation, where systems like VisDoMRAG are combining visual and textual RAG to handle visually rich documents, leading to improved accuracy and verifiability. Fine-grained attribution mechanisms are also being refined to provide more granular evidence support for generated answers, leveraging techniques such as dependency parsing augmentation. Benchmarking efforts are evolving to include more complex and authentic chart types from scientific literature, ensuring a more realistic assessment of model capabilities. Overall, the field is moving towards more sophisticated, context-aware, and multimodal solutions that enhance the reliability and interpretability of RAG systems.