AbductionRules: Training Transformers to Explain Unexpected Inputs - 42Papers