Managing AI Hallucinations

2

Preventing Hallucinations

In Chapter 1, Understanding AI Hallucinations, you learned what hallucinations are and why they occur. This chapter focuses on reducing them.

It approaches the problem at two levels. The first is the prompt: how structure, boundaries, and examples can give the model less room to guess. The second is the data the model can access: retrieval-augmented generation (RAG) gives the model access to relevant, up-to-date sources before it generates a response, reducing its reliance on training data alone.

In this chapter, we will cover the following topics:

• Controlling AI output through prompt design
• Preventing hallucinations through RAG

Technical Requirements

To follow the local RAG example in this chapter, you will need ...