FashionLens: Toward Versatile Fashion Image Retrieval via Task-Adaptive Learning

Researchers introduced FashionLens, a task-adaptive multimodal retrieval framework for fashion search.

They also released U-FIRE, a benchmark that unifies fragmented fashion datasets and query types into one evaluation setting.

The model adapts query representations and training across tasks, achieving state-of-the-art results.

It shows stronger generalization on unseen retrieval tasks, addressing a major gap in real-world e-commerce search.