task-adaptive-learning

FashionLens proposes a unified fashion image retrieval framework using multimodal large language models with adaptive calibration and sampling, achieving state-of-the-art performance across diverse retrieval scenarios.