EVP: Event-Aware Prompt Learning for Dynamic Graphs

A research paper on arXiv (2510.11339) introduces EVP, an event-aware dynamic graph prompt learning framework designed to enhance existing dynamic graph neural networks (DGNNs). Unlike current methods that focus on node-time relationships, EVP leverages historical event knowledge. It extracts events for each node and uses an event adaptation mechanism to align event characteristics with downstream tasks, plus an event aggregation mechanism to integrate historical knowledge into node representations. EVP serves as a plug-in to improve DGNN performance.