RAG
Domain: LLM Concepts
🎯 Motivation
Retrieval-Augmented Generation: instead of relying on the model’s memory, retrieve relevant documents from a database and include them in the context.
📋 Concrete Example
User asks about company policy. RAG searches the company document database, finds the relevant policy, includes it in the prompt. The model reads it and answers accurately.
🔗 Analogy
An open-book exam - instead of memorizing everything (pure LLM), you bring reference books (retrieval). When asked a question, you look up the relevant page and base your answer on it. Much more accurate.