Suggest a paper

AI papers, preferably vision-language models and their ancestors. Curators review every suggestion; nothing is added automatically.