WikiDO: A New Benchmark Evaluating Cross-Modal Retrieval for Vision-Language Models

Cross-modal (image-to-text and text-to-image) retrieval is an established task used in evaluation benchmarks to test the performance of vision-language models (VLMs).