ResearchBenchmarksWeakly Supervised Referring Expression Comprehension on RefCOCOg (val)Follow60.4AccuracySVCR‡53.161655.040856.9258.7992Jul 6, 2026Evaluation ResultsMethodMethodLinksAccuracyInference Speed (FPS)SVCR‡Backbone=YOLOv5Backbone=YOLOv52026.0760.420.2SVCRBackbone=YOLOv3Backbone=YOLOv32026.0756.9918.28DViNBackbone=YOLOv3Backbone=YOLOv32026.0755.0414.49WeakMCN†Backbone=YOLOv3, Found...Backbone=YOLOv3, Foundation Model Variant=ViT-small-SAM2026.0754.6216.92WeakMCN*Backbone=YOLOv3, Found...Backbone=YOLOv3, Foundation Model Variant=ViT-tiny-SAM2026.0753.4417.27