Enhancing Temporal Understanding in Video-LLMs through Stacked Temporal Attention in Vision Encoders Paper • 2510.26027 • Published Oct 29, 2025