Reproduce “OneTrans-GSR: A Generative Slate Ranking Paradigm Unifying Reranking and Mix-Ranking with One Autoregressive Transformer” 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
Feng Cheng, Yu-Hsuan Huang, Fengyi Wang, Chaofan Zhou, Zhennan Zhu, Huan He, Bo Zhang, Ruoqi Zhang and 8 more
OneTrans-GSR replaces a Generator–Evaluator cascade with an autoregressive policy generating an ordered slate from pre-scored candidates. A reward model evaluates slates offline; constraints preserve requirements, while soft preferences guide training. In a week-long A/B test, the replacement improved GMV/u by 2.63%, advertiser value by 2.06%, and repurchase rate by 2.03%, and reduced reranking and mix-ranking latency by 13% in one system.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.