Reproduce “Modeling Attention Flow on 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
Xiaoran Xu, Songpeng Zu, Chengliang Gao, Yuan Zhang, Wei Feng
The model makes path-like reasoning in a graph neural network visible and lets it influence message passing. It moves attention from a source across graph edges, using graph-network information to guide transitions and feed attention back into messages. On synthetic grid trajectories, GGNN-MulMlp reaches 44.01 ± 2.60% Hits@1, versus 22.59 ± 4.28% for regular GGNN, on 64×64 LOCATION trajectories; this does not establish real-world causal explanations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.