Reproduce “A Common Language for Dynamic Graphs: Interaction Tokenization Enables Foundation Models” 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
Fuyuan Ma, Yili Wang, Xin He, Xin Zhang, Ying Wang, Xin Wang
CL-DGFM predicts temporal links across graphs by describing candidate interaction histories with shared features and conditioning a fixed-parameter predictor on labeled examples from each target graph. Trained on synthetic interaction streams, it represents interaction memory, historical connectivity, and ordered destination transitions without node identities. It achieves the highest reported mean reciprocal rank on 10 of 13 TGB/TGB-Seq datasets, not all.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.