Reproduce “WeGR: Towards Unified Generative Retrieval via Behavior-aware Tokenization and Preference Alignment” 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

Jiajun Han, Guoxin Ma, Felix, Liangsheng Xu, Ruiqi Yu, Yingdi Guo, Zongbing Wang, Xinhua Ye and 4 more

WeGR tackles e-commerce search by encoding item behavior in Semantic IDs and training generative retrieval to learn from matching prefixes before complete items. In a production test removing 80% of conventional-channel candidates, V1.2 improved relative Order, CTR, and CVR by 2.539%, 1.508%, and 1.381%, while reducing pre-ranking volume 38.10% and CPU and GPU use; results apply to the full configuration.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.