Does Pre-training Induce Systematic Inference? How Masked Language Models Acquire Commonsense Knowledge

Re-assign community