09fb05dd477d4ae6479985ca56c5a12d-AuthorFeedback.pdf

Learning from entailment ispretty different from what we seek here: itasks ustoproduce aset offormulasH that38 replicates a desired set of entailment judgments, e.g., thatφ1 is entailed butφ2 is not, etc. Our task formulation is39 much closer to learning from interpretations in ILP,where our partial models are partial intepretations.