Utilizing Large Language Models in an iterative paradigm with domain feedback for zero-shot molecule optimization