Understanding LLM Hallucinations

What is an AI Hallucination?

An AI hallucination occurs when a Large Language Model generates incorrect or nonsensical information but presents it confidently as fact.

Common Causes

• Training Data Bias: Models learn from vast internet data, which includes inaccuracies.
• Lack of Real-time Knowledge: Most base models have a knowledge cutoff date.

Mitigation Strategies

Using RAG (Retrieval-Augmented Generation) helps ground the AI's responses in factual documents provided by the user.