Abstract
Large language models (LLMs) have substantially advanced event extraction, a long-standing information extraction task that has evolved from rule-based and classical machine learning systems to modern deep learning approaches. This survey provides an up-to-date overview of LLM-based event extraction methods, organizing them around two main strategies: using LLMs for data augmentation and using prompting or fine-tuning to address the task directly. In contrast to prior surveys, the paper also places special emphasis on datasets, covering resources across multiple languages including English, Chinese, and Vietnamese, with attention to domain-specific settings and annotation scope. The paper concludes with a critical analysis of current challenges and outlines promising future directions for advancing event extraction with LLMs.