Alignment Proposal: Adversarially Robust Augmentation and Distillation — AI Alignment Forum