RaDIO: Real-Time Hallucination Detection with Contextual Index Optimized Query Formulation for Dynamic Retrieval Augmented Generation

Retrieval-Augmented Generation (RAG) is an emerging approach that enhances language models by integrating document retrieval into the generation process. This paper provides a comprehensive study of RAG systems, examining their architecture-comprising retrievers, fusion techniques, and generators-and their performance across knowledge-intensive tasks. We explore the historical development of RAG, compare traditional language models with RAG pipelines, and analyze use cases in healthcare, law, education, and enterprise settings. The study further discusses retrieval and generation methods, optimization strategies such as re-ranking and prompt rewriting, and evaluates model performance using metrics like Recall@k, ROUGE, and FEVER. While RAG addresses key limitations of traditional LLMs, such as hallucination and static memory, challenges remain in retrieval precision, latency, and corpus freshness. The paper concludes by reflecting on RAG's practical value and future potential as a foundation for more grounded and reliable AI systems.