R$^3$AG: Retriever Routing for Retrieval-Augmented Generation
This work addresses the limitation of one-size-fits-all retrieval in RAG by dynamically selecting retrievers based on query-specific preferences, improving downstream task performance.
R^3AG introduces a routing framework for retrieval-augmented generation that models dynamic alignment between queries and retriever capabilities by decomposing retriever capability into retrieval quality and generation utility, outperforming both individual retrievers and static routing methods on knowledge-intensive tasks.
Retrieval-augmented generation (RAG) has become a cornerstone for knowledge-intensive tasks. However, the efficacy of RAG is often bottlenecked by the ``one-size-fits-all'' retrieval paradigm, as different queries exhibit distinct preferences for different retrievers. While recent routing techniques attempt to select the optimal retriever dynamically, they typically operate under a ``single and static capability'' assumption, selecting retrievers solely based on semantic relevance. This overlooks a critical distinction in RAG: a retrieved document must not only be relevant but also effectively support the generator in producing correct answers. To address this limitation, we propose R$^3$AG, a novel routing framework that explicitly models the dynamic alignment between queries and retriever capabilities. Unlike previous approaches, R$^3$AG decomposes retriever capability into two learnable dimensions: retrieval quality and generation utility. We employ a contrastive learning objective that leverages complementary supervision signals, \textit{i.e.}, document assessments and downstream answer correctness, to capture query-specific preference shifts. Extensive experiments on several knowledge-intensive tasks show that R$^3$AG consistently outperforms both the best individual retrievers and state-of-the-art static routing methods.