Reproduce “Utility-Aware Generative Recommendation: Utility-Enriched Semantic Codes and Utility-Guided Decoding” 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
Yang Hu, Duo Wang, Ning Li, Kang Li, Junda She, Chengcheng Guo, Kuo Cai, Bin Liu and 2 more
UAGR addresses semantic-ID recommendation by adding bucketed item utility to each item's final code token and using predicted post-exposure outcomes to guide beam pruning. In a one-week Kuaishou test, watch time rose 0.206%, app usage time 0.074%, and video views 0.164%, while QPS decreased 0.03%. These package-level results do not isolate either component’s contribution or measure latency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.