Young Kyun Jang, Donghyun Kim. "Visual delta generation with large multi-modal models enhances composed image retrieval using unlabeled data." Scientific Reports, 2025. doi:10.1038/s41598-025-07798-6