What it does This repository is a living bibliography that sorts GNN-based recommender papers by where they fit in the pipeline—matching, ranking, re-ranking—and by what flavor of problem they tackle: social, sequential, cross-domain, fairness, explainability, and more. Each entry links to the paper, venue, year, and code when available. Think of it as a literature review you can grep.

The interesting bit The taxonomy itself is the contribution. The authors (from Tsinghua’s FIB Lab) published a full survey in ACM TORS, and this repo is the companion index that keeps the field organized as it balloons. It turns out “GNN for recs” is not one thing—it’s a dozen sub-communities that barely talk to each other.

Key highlights

• Covers 50+ papers from 2017–2022, from PinSage-scale industrial work to niche fairness-aware methods
• Cross-references code links (PapersWithCode, GitHub, or “NA” when none exists)
• Organized by recommendation stage (matching → ranking → re-ranking), not just by model name
• Includes emerging objectives: multi-behavior, diversity, explainability, fairness
• Backed by a peer-reviewed survey (TORS 2022) with a proper citation if you use it

Caveats

• Not all entries have code links; some point to PapersWithCode pages that may or may not have working implementations
• The README is a single markdown file—no search, no tags, no automated updates beyond manual curation
• Truncated in the source; the full list likely extends further than what’s visible

Verdict Grab this if you’re writing a related-work section, choosing a baseline, or trying to figure out if someone already solved your exact GNN-rec problem. Skip it if you want runnable code or a framework—this is a map, not a toolbox.