Why RAG, and its limits
On This Lesson 4 topics
This lesson is planned. The outline below lists the topics it will cover. The full lesson has not been published yet.
In this lesson
explain why RAG grounds answers in your data and where it still fails.
Planned topics
- A model only knows its training data: it does not know Shop's return policy and can invent one.
- RAG (retrieval-augmented generation): find relevant passages in your data, put them in the prompt, ask the model to answer from them.
- Benefits: current data without retraining, answers traceable to sources, access control at retrieval time.
- Failures: retrieval misses, stale chunks, the model ignoring or misreading context, too much irrelevant context.
- RAG reduces hallucination; it does not remove it.
Draft outline details
Demonstration
flow: question → embed → search index → top chunks → prompt → model → answer with sources.
Check your understanding
Name two ways a RAG answer can still be wrong.
Show answer
Retrieval misses the right chunk; the chunk is stale; the model ignores the context.
Practice
Apply this lesson with the module's lab, exercises and quiz.
Marking a lesson complete is optional and you can undo it any time. Your progress stays in this browser.