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:
- Retrieval: Querying a vector database or knowledge base for relevant documents
- Augmentation: Combining retrieved documents with the original prompt
- 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.
