Fetching the paper…

Layer-wise Alignment: Examining Safety Alignment Across Image Encoder Layers in Vision Language Models · Around