Text this: Benchmarking publicly accessible large language models for high-myopia multiple-choice question generation in digital ophthalmic education and public health training.