LaVi: Efficient Large Vision-Language Models via Internal Feature Modulation

MLLMVLMRe-assign community