Language Models Can Improve Event Prediction by Few-Shot Abductive Reasoning

Large language models have shown astonishing performance on a wide range of reasoning tasks. In this paper, we investigate whether they could reason about real-world events and help improve the prediction performance of event sequence models.