DA-VPT: Semantic-Guided Visual Prompt Tuning for Vision Transformers | Read Paper on Bytez