ORDER：面向检索增强生成的任务条件路由
ORDER: Task-Conditioned Routing for Retrieval-Augmented Generation
- EPITA
- EPITECH
- Bpifrance（法国国家投资银行）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
ORDER提出查询条件路由的RAG框架，通过语义聚类学习分块与过滤配置，结合监督路由器和多源采样器，在异构档案中优于现有系统。
AI中文摘要：
检索增强生成（RAG）流水线通常依赖于在预处理阶段确定的固定索引和检索配置。这种一刀切的设计不适合领域专家场景，因为异构查询需要不同的分块粒度、元数据约束和源选择策略。因此，对某一类查询有效的配置往往对其他查询表现不佳。在本文中，我们提出了ORDER（动态证据检索的最优路由），一个查询条件的RAG框架，它联合调整索引和检索以适应传入的查询。我们的方法首先在给定语料库相关的问题集上发现语义簇，并为每个簇学习一种分块策略以及合适的元数据过滤和重排序配置。在推理时，查询通过最近质心分配被路由到适当的预构建索引。为了进一步提高检索效果，我们提出了一个监督查询路由器（QRe），它预测哪些集合最可能包含相关证据，并配有一个统一多源采样器（UMS），该采样器在选定的源之间均匀分配检索预算。我们在大规模、异构的历史档案上评估了我们的框架，并表明在复杂的专家领域环境中，对索引和检索都进行条件化处理始终优于朴素基线和强状态-of-the-art RAG系统。
英文摘要：
Retrieval-Augmented Generation (RAG) pipelines typically rely on a fixed indexing and retrieval configuration determined at preprocessing time. This one-size-fits-all design is ill-suited to domain-expert settings, where heterogeneous queries require different chunking granularities, metadata constraints, and source-selection strategies. As a result, configurations that are effective for one family of queries often perform poorly for others. In this paper, we introduce ORDER (Optimal Routing for Dynamic Evidence Retrieval), a query-conditioned RAG framework that jointly adapts indexing and retrieval to the incoming query. Our approach first discovers semantic clusters over a given set of questions associated to a corpus and learns, for each cluster, a chunking strategy together with a suited metadata filtering and reranking configuration. At inference time, queries are routed to the appropriate pre-built index through nearest-centroid assignment. To further improve retrieval, we propose a supervised query router (QRe) that predicts which collections are most likely to contain relevant evidence, coupled with a Uniform Multi-source Sampler (UMS) that allocates the retrieval budget evenly across the selected sources. We evaluate our framework on large-scale, heterogeneous historical archives and show that conditioning both indexing and retrieval on the query consistently outperforms both naive baselines and strong state-of-the-art RAG systems in complex expert-domain environments.
