Junhyeok Kim

2

Papers

83

Total Citations

Papers (2)

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

Your Large Vision-Language Model Only Needs A Few Attention Heads For Visual Grounding