Reproduce “Gaussian Relational Graph Transformer” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zezhong Ding, Jin Li, Xugang Wang, Xike Xie

GelGT predicts relational tasks by sampling time-valid graph neighborhoods, retaining direct links, and ranking distant records by feature similarity. It combines edge-based aggregation with attention shaped by a Gaussian temporal bias. Across 21 tasks on seven RelBench datasets, it reaches 0.8357 AUC on rel-event user-repeat versus 0.7609 for RelGT; this benchmark result is not a guarantee for every task.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.