AI RESEARCH
Context Features Are Cheap: Rank-Aware Decomposition for Efficient Feature Interaction in Recommender Systems
arXiv CS.LG
•
ArXi:2605.27450v1 Announce Type: cross Modern industrial recommender systems use a deep ranking model to score N candidates against the same user and context features. Standard implementations broadcast context features early in the forward pass, redundantly computing context-only operations N times per request.