What is a "hallucination" in the context of an LLM?

Question

Accepted Answer

A confident answer that is false — fluent text the model simply made up — A hallucination is fluent, confident output that's simply false — the model fabricates plausible-sounding details. It's not a refusal, a pause, or an explicit warning.

Answer

A request the model refuses — output it declines to produce on safety grounds

Answer

A pause during generation — a short delay while the model fetches more data

Answer

A warning the model prints — a notice it flags whenever unsure of the facts

What is a "hallucination" in the context of an LLM?

Why this is the answer

More AI & LLM Fundamentals flashcards