Sparse Attention Vectors: Generative Multimodal Model Features Are Discriminative Vision-Language Classifiers

VLM
Re-assign community