Reproduce “Neural Structural Reasoner: A Brain-inspired Architecture for Reasoning over Structured Knowledge” 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
Zixing Jia, Yuhang Pan, Ni Ji
NSR represents knowledge-graph relations and compositions in network connections and changing activity, making inference paths inspectable while ranking candidate answers. It encodes triples and follows relation paths to rank candidate tails; scores are not probabilities. On Nations, filtered MRR is 0.8142 versus ConvE’s 0.8029, while disabling composition lowers MRR from 0.95 to 0.46 on a constructed kinship dataset, so results are dataset-dependent.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.