The growing prevalence of large language models (LLMs) has exposed a critical limitation: hallucination, where models generate plausible-sounding but factually incorrect content.