The recent advance in vision-language models is largely attributed to the abundance of
image-text data. We aim to replicate this success for video-language models but there …