HyperEvent: A Strong Baseline for Dynamic Link Prediction via Relative Structural Encoding

TL;DR

HyperEvent introduces a simple yet effective method for dynamic link prediction in continuous-time graphs by using relative structural encoding, serving as a strong baseline that rivals complex models.

Contribution

It presents HyperEvent, a straightforward approach that captures relative structural patterns with an intuitive encoding, providing a reliable baseline for dynamic link prediction.

Findings

HyperEvent achieves competitive results across multiple benchmarks.

It often matches the performance of more complex models.

The method demonstrates the effectiveness of simple structural encoding.

Abstract

Learning representations for continuous-time dynamic graphs is critical for dynamic link prediction. While recent methods have become increasingly complex, the field lacks a strong and informative baseline to reliably gauge progress. This paper proposes HyperEvent, a simple approach that captures relative structural patterns in event sequences through an intuitive encoding mechanism. As a straightforward baseline, HyperEvent leverages relative structural encoding to identify meaningful event sequences without complex parameterization. By combining these interpretable features with a lightweight transformer classifier, HyperEvent reframes link prediction as event structure recognition. Despite its simplicity, HyperEvent achieves competitive results across multiple benchmarks, often matching the performance of more complex models. This work demonstrates that effective modeling can be achieved through simple structural encoding, providing a clear reference point for evaluating future advancements.