Study shows the best visual learning models fail at very basic visual identification tests [TechSpot]

Researchers from Auburn University and the University of Alberta recently published a paper titled “Vision language models are blind.” The study used eight straightforward visual acuity tests to highlight deficiencies in visual learning models (VLM). The tasks included counting intersecting lines, identifying circled letters, counting nested shapes and others. These…Read Entire Article