Reproduce “CITE: A Comprehensive Benchmark for Heterogeneous Text-Attributed Graphs on Catalytic Materials” 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
Chenghao Zhang, Qingqing Long, Ludi Wang, Wenjuan Cui, Jianjun Yu, Yi Du
CITE provides a benchmark for evaluating graph-learning methods on catalytic-materials literature. It combines Crossref text and typed links among papers, authors, journals, and keywords, and predicts 85 journal-derived paper labels. On a shared split, TAPE reports Micro-F1 0.9936 and Macro-F1 0.5655, versus GraphSAGE’s 0.5117 and 0.0474, showing that high aggregate performance does not establish balanced results across labels.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.