See What You Are Told: Visual Attention Sink in Large Multimodal Models
Seil Kang, Jin-Yeong Kim, Junhyeok Kim, Seong Jae Hwang
De quem é
- Yonsei Universityafiliação declarada pelos autores, segundo o OpenAlex
Outros papers das mesmas organizações
- Foundation Models Defining a New Era in Vision: A Survey and OutlookGeorgia Institute of Technology · 733 citações
- UNETR++: Delving Into Efficient and Accurate 3D Medical Image SegmentationGoogle · 682 citações
- MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon AgentsSingapore-MIT Alliance for Research and Technology · 239 citações
- HumanoidBench: Simulated Humanoid Benchmark for Whole-Body Locomotion and ManipulationYonsei University · 209 citações
- Per-Gaussian Embedding-Based Deformation for Deformable 3D Gaussian SplattingElectronics and Telecommunications Research Institute · 150 citações