Hallucination is Inevitable for LLMs with the Open World Assumption

Bowen Xu

Large Language Models (LLMs) exhibit impressive linguistic competence but also produce inaccurate or fabricated outputs, often called ``hallucinations''. Engineering approaches usu…