Enhancing Distractor Generation for Multiple-Choice Questions with Retrieval Augmented Pretraining and Knowledge Graph Integration

In this paper, we tackle the task of distractor generation (DG) formultiple-choice questions. Our study introduces two key designs. First, wepropose retrieval augmented pretraining, which involves refining thelanguage model pretraining to align it more closely with the downstream task ofDG. Second, we explore the integration of knowledge graphs to enhance theperformance of DG. Through experiments with benchmarking datasets, we show thatour models significantly outperform the state-of-the-art results. Ourbest-performing model advances the F1@3 score from 14.80 to 16.47 in MCQdataset and from 15.92 to 16.50 in Sciq dataset.