CP v2
Johns Hopkins researchers found many AI models still miss social context in short videos, a key limitation for safer self-driving cars and assistive robots.