Abstract: Retrieval-Augmented Generation (RAG) has emerged as an effective approach for question answering over domain-specific documents by grounding large language model outputs in external ...