Text this: Evaluating the performance of large language & visual-language models in cervical cytology screening