Reproduce “LSNR: Learned Sparse News Recommendations with Attentive User Modeling” 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
David Vos, Andrew Yates, Maarten de Rijke
LSNR represents each clicked news article separately and learns attention weights to combine sparse lexical and semantic signals into a user profile. Candidate vectors can be precomputed for indexing. On MIND and EBNeRD, it improves recommendation quality for short histories; at one click on EBNeRD, it scores 0.5022 NDCG@10 versus STAR at 0.4656. The result is limited to tested impression-based benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.