Addressing Stereotypes in Large Language Models: A Critical Examination and Mitigation