发表机构
Google; University of Illinois Urbana-Champaign（谷歌; 伊利诺伊大学厄巴纳-香槟分校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
RPTune提出端到端框架，通过学习型目录策展与LLM后训练相结合，在小型商户目录搜索中显著提升准确率，增益最高达31.4个百分点。
AI 中文摘要
对于目录规模适配于长上下文LLM的小型商户企业（SMBs）而言，全目录提示（full-catalog prompting）为多阶段检索提供了一种有吸引力的替代方案，后者主要针对拥有数百万商品的大型市场设计。然而，将完整目录放入上下文窗口并不能确保模型能够有效利用它，因为LLM并非均匀地利用长上下文。因此，我们通过两个互补的问题研究上下文内目录搜索：（1）如何策展并向LLM呈现目录，（2）如何使LLM适应于在策展上下文上进行商品选择。我们提出RPTune，一个端到端框架，将学习型目录策展与基于自动生成、目录接地监督的LLM后训练相结合。一个编码器-重组器策展器（encoder-reorganizer curator）根据下游LLM反馈对商品进行排序和剪枝，而生成的策展目录反过来通过上下文相对奖励（context-relative reward）提升LLM后训练的效果。我们在7个覆盖不同零售垂直领域的真实商户上评估RPTune，每个商户使用100个复杂对话查询。RPTune在专有和开源权重LLM上均持续提升搜索准确率，其中上下文策展带来高达31.4个百分点的增益，后训练平均额外增加10.3个百分点。
英文摘要
For small merchant businesses (SMBs) whose catalogs fit within a long-context LLM, full-catalog prompting offers a compelling alternative to multi-stage retrieval designed primarily for large marketplaces with millions of items. However, fitting the full catalog into the context window does not ensure that the model can use it effectively, since LLMs do not exploit long contexts uniformly. We therefore study in-context catalog search through two complementary questions: (1) how to curate and present catalogs to the LLM, and (2) how to adapt the LLM for product selection on curated contexts. We propose RPTune, an end-to-end framework that couples learned catalog curation with LLM post-training using automatically generated, catalog-grounded supervision. An encoder-reorganizer curator orders and prunes products guided by downstream LLM feedback, while the resulting curated catalogs in turn improve the effectiveness of LLM post-training with a context-relative reward. We evaluate RPTune on 7 real merchants spanning distinct retail verticals, using 100 complex conversational queries per merchant. RPTune consistently improves search accuracy across both proprietary and open-weight LLMs, with context curation yielding gains of up to 31.4 percentage points and post-training adding a further 10.3 points on average.
Comments23 pages, 9 figures, 4 tables
