A arxiv.org LensVLM: Selective Context Expansion for Compressed Visual Representation OfText 1 points KitN 1 month ago 0 comments