Shopify 提出 SID-LLM：用语义 ID 生成统一搜索与推荐
原标题：Generative Search and Recommendation at Shopify
AI 摘要
Shopify 在 arXiv 发布技术报告，介绍其生成式搜索与推荐模型 SID-LLM。该模型将商品表示为语义 ID（内容嵌入量化得到的短 token 序列），联合预训练目录数据与行为数据，使搜索和推荐共享同一表示与训练目标，并通过强化学习提升搜索相关性、从任务专家蒸馏提升推荐质量。在服务端，SID-LLM 结合多 token 预测与轻量级排序器，在降低延迟的同时提升检索质量。评估显示，作为独立检索器，其质量与 Shopify 现有多模型搜索栈相当，并满足生产延迟要求；集成进搜索与推荐栈后，离线评估和线上 A/B 测试均取得显著提升。
正文节选
Generative Search and Recommendation at Shopify Abstract Search and recommendation at Shopify require retrieving relevant products from a changing catalog of billions of products under strict latency constraints. We present SID-LLM, which casts both tasks as generating semantic IDs: short sequences of content-derived tokens representing products. We jointly pre-train SID-LLM on catalog and behavioral data, giving search and recommendation a shared representation and training objective. We show t