Reproduce “TS-GEAR: Taobao Shangou GEnerative Ad Retrieval” 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
Guofeng Tong, Xu Zhang, Guangliang Yu, Haiqiang Wang, Guannan Cong, Xing Guo, Yang Zhou, Jingjing Duan and 3 more
TS-GEAR adapts generative retrieval to on-demand delivery search by generating merchant IDs with constrained geographic prefixes and conditioning user history on the current query and location. It also trains candidate groups to reward coverage rather than duplicates. On TS-Delivery, hit rate at 50 within 20 km is 0.8080, versus 0.5929 for OneLoc; the gain is only 1.2% at 5 km.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.