Why Large Language Models Hallucinate?