Unlocking the Potential of Large Language Models for Clinical Text Anonymization: A Comparative Study