Reproduce “E-GenRM: Generative Reward Modeling for Open-Ended E-Commerce Shopping Agents” 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
Ying Sun, Qiuyu Zhao, Xiao Wang, Xiyu Zhao, Gege Shi, Zenghui Sun, Jinsong Lan, Haihong Tang and 1 more
E-GenRM turns shopping behavior into preference examples and trains a model to compare responses using task-specific criteria and product evidence. It evaluates candidates criterion by criterion, with optional retrieval when context lacks decisive facts. On E-GenRM-Bench, retrieval-enabled E-GenRM-27B reaches 87.40% accuracy, 6.97 points above the strongest baseline. A 14-day Taobao A/B test reports assistance and satisfaction gains of 1.39 and 2.33 percentage points, respectively, within that deployment setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.