VISA: Video Skeleton-Aware Efficient Frame Selection for Long Video Understanding

Published in EMNLP 2026, 2026

Download Paper