Large language models and how they work.
In LLMs, hallucination is when the model generates fluent, confident text that is factually incorrect, fabricated, or not supported by its training data.