Abstract
Recent works have demonstrated the effectiveness of retrieval augmentation inthe Event Argument Extraction (EAE) task. However, existing retrieval-based EAEmethods have two main limitations: (1) input length constraints and (2) the gapbetween the retriever and the inference model. These issues limit the diversityand quality of the retrieved information. In this paper, we propose aCompressive Memory-based Retrieval (CMR) mechanism for EAE, which addresses thetwo limitations mentioned above. Our compressive memory, designed as a dynamicmatrix that effectively caches retrieved information and supports continuousupdates, overcomes the limitations of the input length. Additionally, afterpre-loading all candidate demonstrations into the compressive memory, the modelfurther retrieves and filters relevant information from memory based on theinput query, bridging the gap between the retriever and the inference model.Extensive experiments show that our method achieves new state-of-the-artperformance on three public datasets (RAMS, WikiEvents, ACE05), significantlyoutperforming existing retrieval-based EAE methods.