Reproduce “Beyond Trained Models: Compiling GNNs for a Sound Explainer Benchmark” 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
Steve Azzolin, Francesco Paolo Nerini, Stefano Teso, Francesco Bonchi, Bruno Lepri, André Panisson, Andrea Passerini
The authors introduce GracrBench, a controlled benchmark that tests graph neural network explainers against model behavior. Gracr compiles graded modal logic rules into GNNs and derives ground truth from the union of minimal sufficient edge sets. On CompetingRedundancy, SALIENCY’s precision and F1 shift from 0.04 and 0.09 to 1.00 across equivalent compilations. These results concern controlled models, not production GNNs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.