Fetching the paper…

Scaling Video-Language Models to 10K Frames via Hierarchical Differential Distillation · Around