PailitaoMMSearch2.0 技术报告：将生成式检索扩展至统一端到端多模态搜索
PailitaoMMSearch2.0 Technical Report: Scaling Generative Retrieval to Unified End-to-End Multimodal Search
Xiaohan Ye · Zihan Gong · Jingqian Zhao · Lianyu Du · Yunmeng Shu · Xu Chen · et al.
中文摘要
生成式检索模型在搜索推荐系统中应用广泛，但工业界模型通常局限于特定在线场景，仅支持有限的查询形式。本研究旨在构建统一的端到端多模态搜索系统，以处理开放式图文查询及异构购物意图。PailitaoMMSearch 直接从原始查询图像和自由指令生成商品语义 ID，并在三个维度进行扩展。数据扩展利用可复用工具集，支持冷启动合成、质量控制及真实流量优化。意图扩展通过持续预训练（CPT）、混合预算监督微调（SFT）及统一强化学习（RL），赋予模型电商知识及可调推理预算。场景扩展通过模型容量和推理预算展示质量与成本的权衡点。该系统已部署于淘宝拍立淘，在线 A/B 测试显示商品浏览量提升 17.90%，交易量提升 27.04%，商品交易总额提升 8.94%。
关键要点
- 01解决工业生成式检索模型局限于固定在线场景、查询形式单一的问题，构建统一端到端多模态搜索系统。
- 02提出 PailitaoMMSearch，直接从图像和指令生成商品语义 ID，通过数据、意图和场景三个维度进行扩展。
- 03采用持续预训练、混合预算 SFT 及结合意图平衡优化的统一 RL，实现模型对多检索关系和推理预算的适应。
- 04在淘宝拍立淘部署后，A/B 测试显示商品浏览量、交易量及商品交易总额分别显著提升 17.90%、27.04% 和 8.94%。
解读
尚无解读。
原始英文摘要
arXiv:2610.36958v1 Announce Type: cross Abstract: Generative retrieval models have been increasingly adopted in search and recommendation systems. However, industrial generative retrieval models are commonly specialized to a fixed online scenario: they learn from scenario-specific interactions, support a narrow set of query forms and retrieval objectives. We study how to remove these couplings and build a unified end-to-end multimodal search system for open-ended image--text queries, heterogeneous shopping intents, and online scenarios with different resource constraints. We present \textbf{PailitaoMMSearch}, which directly generates product Semantic IDs from a raw query image and free-form instruction and scales along three dimensions. \textbf{Data scaling} uses a reusable toolset that can be composed into intent- and lifecycle-specific workflows for cold-start synthesis, quality control, real-traffic refinement, and continuous data production. \textbf{Intent scaling} progressively equips one model with e-commerce product knowledge, multiple retrieval relations, and adjustable reasoning budgets through continual pre-training (CPT), mixed-budget supervised fine-tuning (SFT), and unified reinforcement learning (RL). The RL stage combines intent-balanced optimization and Sample--Beam rollouts aligned with multi-candidate serving. \textbf{Scenario scaling} exposes quality--cost operating points across model capacity and reasoning budget. The system is deployed in multimodal search applications on Taobao Pailitao. A controlled online A/B test reports statistically significant improvements, including item page views $\textbf{+17.90\%}$, transactions $\textbf{+27.04\%}$, and gross merchandise value $\textbf{+8.94\%}$.