A hallucination is a fluent, confident statement from a model that is false or not supported by its sources. It happens because a model generates likely text, not checked facts, and nothing in plain generation tells it to stop when it does not know.
You cannot remove hallucinations completely, but you can make them rarer and easier to catch. Grounding answers in supplied documents, asking for citations, letting the model say it does not know, verifying claims in a separate step and checking outputs in code all help. These guides start with quick checks for beginners and move to methods such as Chain-of-Verification, SelfCheckGPT and semantic entropy, which detect unsupported claims by comparing several answers.
The library lists all guides with filters for topic, level, category and reading time.