13 retrieval tasks

The '13 retrieval tasks' dataset/benchmark is a collection of multimodal retrieval challenges designed to evaluate the performance of models in retrieving relevant information across three modalities: text, vision, and audio.