Reproduce “T-GRAB: A Synthetic Diagnostic Benchmark for Learning on Temporal Graphs” 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
Alireza Dizaji, Benedict Aaron Tjandra, Mehrab Hamidi, Shenyang Huang, Guillaume Rabusseau
T-GRAB builds synthetic temporal-graph link-prediction tasks that isolate counting repeated patterns, delayed memory, and information transfer across time and graph distance. It varies temporal rules and evaluates models with F1, including at change points where persistence fails. On deterministic two-pattern tasks with n = 32, EdgeBank outperforms every tested TGNN in all-time-step F1; results are diagnostic, not generally transferable.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.