Generative retrieval system improves relevance and value in e-commerce search
VARG: Value-Aware and Ranking-Aligned Generative Retrieval for Dynamic E-commerce Search
Information Retrieval
Summary
E-commerce search needs to quickly find products that are not only relevant to what users want, but also good for the business. The authors created VARG, a system that generates product candidates with careful attention to both customer preferences and business priorities. It uses a special way to encode products and queries, learns step-by-step to improve matching, and updates daily to keep results fresh. Tests show it improves sales and user engagement on a major shopping app while keeping search results relevant with fewer options shown.
What this means in practice
- •For e-commerce platform engineers: Integrate VARG to generate product candidates that improve sales and user engagement by aligning retrieval with business goals and relevance.$Commercial implications: This enables a new product search system that directly boosts revenue and user clicks on e-commerce platforms like Tmall.
- •For online advertising teams: Use VARG-generated candidates to feed ranking systems which optimize both relevance and commercial value of advertised products.
Authors
Xiaopeng Chu, Jianbo Zhu, Mingmin Jin, Jing Wang, Xing Fang, Wenyi Zhang
Abstract
Integrating recall and pre-ranking in e-commerce search requires candidate generation to account for relevance, personalization, and business value before final ranking. To this end, we present VARG, a generative retrieval system for Tmall App search that directly admits generated item candidates to the existing final ranker. VARG-ID constructs semantic prefixes using RQ-VAE, enhances search relevance through bidirectional query-item contrastive learning, and combines these prefixes with a value-ordered third token to provide fine-grained item addresses and a business-value prior. Three-stage supervised fine-tuning progressively learns item-to-identifier mappings, query-semantic retrieval, and personalized retrieval. Personalized model training combines value-aware and hierarchy-aligned supervision with expanded user context, and uses local ordinal supervision (LO-SFT) to learn the local within-cluster ordering encoded by the third token. Prefix-GRPO combines gated rewards based on output legality, user behavior, ranker advantage, and search relevance with prefix-aware token weighting to align candidate generation with business value and ranking objectives. Coordinated daily product and model updates preserve existing item addresses while incorporating new products and behavioral feedback. Offline experiments on tens of millions of products validate identifier stability and demonstrate gains in retrieval quality and head-level value recall from SFT strategies and Prefix-GRPO over their respective baselines. In a 14-day online A/B test covering 20% of search traffic, VARG directly admits generated candidates to the final ranker and improves GMV by 1.45%, per-user IPV by 0.22%, and PCTR by 0.31%. Online shopping-guide query evaluations further show that VARG maintains competitive relevance with a smaller candidate quota.