What am I missing here?: Evaluating Large Language Models for Masked Sentence Prediction