ResearchBenchmarksImage-to-Text Retrieval on Flickr30K (R@5, R@10)Follow80.8Recall@5UniCon14.55231.75148.9566.149Apr 17, 2026Evaluation ResultsMethodMethodLinksRecall@5Recall@10UniConBackbone=CLIP ViT-B/32...Backbone=CLIP ViT-B/32, Train time (s)=11.15, Zero-shot transfer=true2026.0480.887.9UniConBackbone=RN-50 + SBERT...Backbone=RN-50 + SBERT, Train time (s)=11.11, Zero-shot transfer=true2026.0417.126.1