See What You Are Told: Visual Attention Sink in Large Multimodal Models.

Seil Kang, Jinyeong Kim, Junhyeok Kim, Seong Jae Hwang

Browse the full ICLR paper archive.