Fetching the paper…

MIA-Former: Efficient and Robust Vision Transformers via Multi-grained Input-Adaptation · Around