Reproduce “NodeGround: A Node Classification Benchmark in the Graph Foundation Model Era” 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
Jinmo Lee, Dooho Lee, Minho Jeong, Jaemin Yoo
NodeGround benchmarks six pretrained graph models against 15 supervised methods across 51 node-classification datasets, testing whether reuse preserves prediction quality and lowers workflow cost. It evaluates limited-label and label-rich splits, and compares rankings and workflow costs. GraphPFN ranks third overall in the label-rich split, but tuned GNNs lead overall; no GFM extends the estimated frontier under limited-label conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.