Papers · arXiv · 计算机视觉
Beyond the Sink Token: Sink-Guided Key Editing in Vision-Language Models
Summary · 摘要
arXiv:2611.23516v1 Announce Type: new Abstract: In vision-language models (VLMs), a few sink tokens …
阅读原文 arxiv.org
继续读 · Papers
Learning What to Degrade: A Recoverability-Aware Curriculum for Reliable Smart-Glasses Super-ResolutionarXiv · 计算机视觉 · 4 小时前Image-Condition-Controlled Diffusion Network with Large-Scale CT Pretraining for CT-Free Attenuation Correction in Dynamic Cardiac PETarXiv · 计算机视觉 · 4 小时前EyeDiT: Anatomy-Grounded Multi-Modal Eye Generation and SimulationarXiv · 计算机视觉 · 4 小时前LightSEAD: Lightweight SE(3)-Aware Keypoint Detector through Selective Knowledge DistillationarXiv · 计算机视觉 · 4 小时前PRISM: Profiling and Representation of Individual Stereotypical MotionarXiv · 计算机视觉 · 4 小时前