Yu, Z.; Fu, Y.; Li, S.; Li, C.; Lin, Y. MIA-Former: Efficient and Robust Vision Transformers via Multi-Grained Input-Adaptation. AAAI 2022, 36, 8962-8970.