1. Introduction
Relation extraction (RE), as a core subtask of information extraction (IE), has been widely used in knowledge engineering, information retrieval, knowledge graph and other fields. RE aims to identify semantic relationships between two entities from unstructured text. Conventional RE methods typically handle pre-defined relations. However, in the real world, relations are subject to dynamic evolution, with new relations continuously emerging. The dynamic nature of relations potentially results in catastrophic forgetting, which refers to the irreversible loss of previously learned relation information when adapting to new data and relations. To address this challenge, continual relation extraction (CRE) [1] has come into existence.