Reproduce “Discrete Semantic Tokenization for Deep CTR Prediction” 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
Qijiong Liu, Hengchang Hu, Jiahao Wu, Jieming Zhu, Min-Yen Kan, Xiao-Ming Wu
The authors introduce UIST, a recommender that predicts clicks using short semantic codes for items and users instead of cached dense representations. Residual quantization creates the codes, and a hierarchy-aware method combines their levels. In offline tests on the MIND news dataset, UIST used 3 ms per sample and projected roughly 200-fold lower storage, while ranking scores trailed the embedding-based comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.