Why it exists

LLMs predict likely word sequences rather than querying a factual database. Obscure queries cause models to generate plausible-sounding false statements.

How it works

Mitigated by grounding outputs with RAG context, system prompt constraints, lower sampling temperature, and verification guardrails.