ShotBench: Expert-Level Cinematic Understanding in Vision-Language Models Paper • 2506.21356 • Published 4 days ago • 19