BLINK : Multimodal Large Language Models  Can See but Not Perceive (ECCV 2024)

Hidden in plain sight: VLMs overlook their visual representations

A life hack for autonomous vehicle researchers