E-commerce improves search suggestions using clicked items and AI learning
EAGER: Enrich-and-Align Generative Query Recommendation from Clicked Items in E-commerce Search
Information Retrieval
Summary
People shopping online often get helpful search hints based on what they click on, but current methods either only rely on past searches or use general language AI that misses shopping details. The authors created a two-step approach called EAGER that first teaches AI to understand clicked items more deeply and in context, and then tunes it to better match actual user preferences and business goals. This method uses detailed training stages and rewards based on clicks and rules to make suggestions that feel more personal and relevant. Their tests show it works better, and it is already being used by a large online store.
E-commerce platformsquery suggestionsuser intentsupervised fine-tuningchain-of-thought reasoningself-distillationreward optimizationA/B testingclick behaviorgenerative AI
Authors
Shuwei Yuan, Mingqian Ding, Luxin Liu, Rong Xiao, Xiaoyi Zeng
Abstract
E-commerce platforms increasingly display clickable query suggestions alongside items in the user feed, enabling users to refine or expand their intent without manually reformulating queries. Existing approaches either mine suggestions from historical logs -- limited to past behavior and blind to long-tail, personalized intents -- or rely on off-the-shelf LLMs whose lack of platform-specific knowledge yields fluent but generic queries disconnected from real click behavior. We propose EAGER (Enrich-and-AliGn gEnerative Query Recommendation), a two-stage framework for generating query suggestions from clicked items. In the enrichment stage, supervised fine-tuning (SFT) follows a four-stage curriculum that scales information richness (from item-only to user-conditioned) and reasoning depth (from direct to chain-of-thought). Each stage incorporates rationale augmentation, diversity regularization, and self-distillation. In the alignment stage, we post-train via GRPO with a hybrid reward of multiple rule-based business signals and a preference-aware click reward. Extensive offline experiments and online A/B test demonstrate the effectiveness of EAGER, which has been deployed in production at a major e-commerce platform.