SECAI Core · phase 5 of 15
Retrieval-augmented generation
A technique where a generative model retrieves relevant information from an external knowledge source and then uses that to generate its answer
The Explain card
- Plain English
- Retrieval-augmented generation (RAG) has a model fetch relevant information from an external knowledge source first, then use that material to generate its answer, instead of relying only on training memory.
- Example
- A company assistant answers "What is our laptop encryption policy?" by searching the internal policy repository, pulling the matching document, and composing a reply from it. The answer reflects current policy, not whatever the model absorbed during training.
- Why it matters
- RAG reduces hallucination and keeps answers current, but every retrieved document becomes model input. A poisoned page in the knowledge base can carry instructions or falsehoods straight into the answer.
- Hook
- RAG is open-book exam mode for the model, so guard the book.
Where it sits in the deck
Phase 5: Interacting with Models: Prompting, APIs, and Retrieval
With a working model and clean data, learn the interaction layer — how users and systems communicate instructions and retrieve grounded answers.
