GLIMPSE: Do Large Vision-Language Models Truly Think With Videos or Just Glimpse at Them?

Yiyang Zhou, Linjie Li, Shi Qiu +10

Existing video benchmarks often resemble image-based benchmarks, with question types like "What actions does the person perform throughout the video?" or "What color is the woman's…