Token-level Response-visual Attention Guidance for Multimodal LLMs Knowledge Distillation figure
AlphaXiv 中文概览(可滚动查看)