Robust Detection of LLM-Generated Text: A Comparative Analysis