“In the field of cross modal retrieval of video text, experts have proposed a data iterative optimization method guided by large language models, which effectively alleviates the one to many problem in datasets and significantly improves model performance.”