Small Vision-Language Models are Smart Compressors for Long Video Understanding Paper • 2604.08120 • Published 22 days ago • 20