Alignment & Safety
Hallucination
When a model states something false or made up as if it were fact, fluently and confidently.
A hallucination is output that sounds plausible but is wrong or invented, like a fabricated citation or a confident but incorrect answer. It happens because a language model generates statistically likely text rather than looking facts up, so it has no built-in sense of truth. Grounding techniques such as retrieval-augmented generation and careful prompting reduce hallucination, but no current model eliminates it, which is why verification still matters.
Reference: Wikipedia: Hallucination
Related terms
Retrieval-Augmented Generation
A technique that looks up relevant documents and feeds them to a model so its answers are grounded in real sources.
Guardrails
Controls placed around a model to keep its inputs and outputs safe, on-topic, and within policy.
Temperature
A setting that controls how random or predictable a model’s output is.
Large Language Model
A model trained on massive amounts of text that predicts and generates language, powering tools like ChatGPT and Claude.