Google DeepMind has launched FACTS Grounding, a new benchmark designed to evaluate the factual accuracy of large language models (LLMs) across over 1,700 tasks. This initiative aims to enhance the performance of Retrieval-Augmented Generation (RAG) systems by providing a systematic way to assess how well LLMs can generate accurate, document-based responses. The benchmark is part of a broader effort to improve the reliability of AI-generated information, addressing challenges such as imperfect retrieval that can lead to the inclusion of irrelevant or misleading data. Alongside this, various enhancements and frameworks for RAG systems have been introduced, including RAGServe, which optimizes query scheduling and configurations to reduce generation latency by up to 2.54 times, and RemoteRAG, a privacy-preserving cloud service that maintains retrieval quality while safeguarding user queries. Other advancements include the C-FedRAG system developed by NVIDIA and Deloitte for secure data connections and the OmniEval framework for evaluating RAG models in the financial sector.