AI RESEARCH
Debiasing Message Passing to Mitigate Popularity Bias in GNN-based Collaborative Filtering
arXiv CS.LG
•
ArXi:2605.11145v1 Announce Type: cross Collaborative filtering (CF) models based on graph neural networks (GNNs) achieve strong performance in recommender systems by propagating user-item signals over interaction graphs. However, they are highly susceptible to popularity bias, since skewed interaction distributions and repeated message passing across high-order neighborhoods amplify the influence of popular items while suppressing long-tail ones.