HF Daily Papers
· Papers
VideoChat3: Fully Open Video MLLM for Efficient and Generalist Video Understanding
Recent advances in video understanding have spanned motion, long video, and streaming interaction, driving this field toward real-world applications. Despite this progress, current open-source models remain limited in several ways. They often struggle to generalize across diverse video types, making