Reproduce “KuaiRecV2: Benchmarking Large-Scale Continual Learning for Diversified and Multi-task Recommendation” 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
Chenxu Li, Shuchang Liu, Hantao Shu, Wei Yuan, Zhe Xu, Guoqing Hu, Yan Wang, Biao Yang and 10 more
KuaiRecV2 is a benchmark for evaluating recommenders as video catalogs change. It uses day-by-day training and a rolling seven-day candidate pool, measuring accuracy, item and token diversity, and multiple behaviors. In the tested retention evaluation, MiniOneRec’s HR@50 falls from 39.09 on Day 1 to 0.19 on Day 7; 5% replay reaches 3.20. This shows forgetting in the tested setup; offline–online evaluation gaps remain.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.