Investigating LLM Hallucination in Search

Language models create plausible text by predicting patterns, not verifying facts. When the data they learn from is incomplete or ambiguous, they can.