A reading list for when your GNN acts like a black box

What it does This repository is a curated bibliography of academic papers focused on making Graph Neural Networks interpretable. It organizes surveys, benchmark platforms, and research papers by publication year, linking directly to PDFs and code where available. Think of it as a living literature review for anyone trying to understand why a graph model made a particular prediction.

The interesting bit The list separates foundational surveys and implementation platforms from the raw firehose of annual research, which keeps it from becoming an unreadable link dump. The maintainer is already indexing 2026 conference papers, suggesting active curation of a literature base that is expanding rapidly.

Key highlights

• Surveys from major venues like ACM Computing Surveys and TPAMI provide structured entry points
• Platforms section links to usable libraries such as PyTorch Geometric, DIG, and GraphXAI
• Coverage spans sub-fields: counterfactual explanations, temporal graphs, heterogeneous networks, and anomaly detection
• Active maintenance indicated by inclusion of 2026 conference and preprint papers
• 808 stars suggest the research community treats this as a working reference

Caveats

• The README offers minimal annotation beyond titles and links, so expect a reference index rather than a guided review
• This is strictly a reading list: the repository contains no original code or reproducibility artifacts
• The source is truncated after the 2026 section, leaving the depth of historical coverage only partially visible

Verdict Bookmark this if you are doing a literature review or choosing an explainability method for a graph project. Look elsewhere if you need runnable code or synthesized takeaways—this is a map, not a tour guide.