Glance, then Select: Evidence-Tempered Visual Token Selection for Extremely Compressed Multimodal LLMs
1 version
Author records are still filling in while the Hub is in beta. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.