2 papers across 2 sessions
Our approach improves Vision Transformer Dense Representations via Self-Distillation