Vision Transformers, just like other neural networks for image classification, have long been interpretable thanks to guided backpropagation to visualise attention. Likewise, simply by visualising them, it is clear that individual neurons i...