Answer: It uses external knowledge sources to improve factual accuracy

## Explanation **RAG (Retrieval-Augmented Generation)** provides significant benefits compared to using an LLM alone: ### **Key Benefit of RAG:** RAG enhances LLMs by retrieving relevant information from external knowledge sources (databases, documents, APIs) and using that information to generate more accurate, factual, and up-to-date responses. ### **Why Option C is Correct:** - **External Knowledge Sources**: RAG retrieves information from external databases or document stores before generating responses - **Improved Factual Accuracy**: By grounding responses in retrieved facts, RAG reduces hallucinations and improves reliability - **Domain-Specific Knowledge**: Can access specialized knowledge not present in the LLM's training data - **Real-Time Information**: Can incorporate current information that wasn't available during the LLM's training ### **Why Other Options are Incorrect:** - **A**: RAG still requires prompt tokens and may even use more tokens due to retrieved context - **B**: While RAG reduces hallucinations, it doesn't guarantee 0% hallucinations - **D**: RAG doesn't eliminate compute costs; it may even increase costs due to retrieval operations ### **Technical Implementation:** RAG typically works by: 1. **Retrieval**: Querying a vector database or knowledge base for relevant documents 2. **Augmentation**: Combining retrieved documents with the original prompt 3. **Generation**: Using the LLM to generate a response based on the augmented context This approach makes LLMs more useful for enterprise applications where factual accuracy and domain-specific knowledge are critical.

Author: Ritesh Yadav