Retriever Portfolios: A Principled Approach to Adaptive RAG

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Stouras, Miltiadis, Cohen-Addad, Vincent, Lattanzi, Silvio, Svensson, Ola
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913172609302528
author Stouras, Miltiadis
Cohen-Addad, Vincent
Lattanzi, Silvio
Svensson, Ola
author_facet Stouras, Miltiadis
Cohen-Addad, Vincent
Lattanzi, Silvio
Svensson, Ola
contents Retrieval-augmented generation (RAG) systems typically rely on a single retriever and a single set of hyperparameters, despite facing highly heterogeneous queries that range from simple factoid questions to complex multi-hop reasoning. We propose a method that automatically selects a small, diverse subset of retrievers (a portfolio) from a large pool of candidates, to cover different regions of the target query distribution. We formalize this setting via an expected best-of-$k$ objective over the query distribution and show that it admits an efficient portfolio construction algorithm with near-optimal guarantees. Across multiple QA benchmarks, our learned portfolios and router pipeline consistently outperform single-retriever and naive multi-retriever baselines on both retrieval metrics and answer quality. In addition, compared to inference-time hyperparameter tuning approaches, fixed portfolios enable parallel retrieval and LLM calls, achieving comparable (and sometimes better) accuracy with substantially lower latency and token cost.
format Preprint
id arxiv_https___arxiv_org_abs_2605_31176
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Retriever Portfolios: A Principled Approach to Adaptive RAG
Stouras, Miltiadis
Cohen-Addad, Vincent
Lattanzi, Silvio
Svensson, Ola
Machine Learning
Data Structures and Algorithms
Retrieval-augmented generation (RAG) systems typically rely on a single retriever and a single set of hyperparameters, despite facing highly heterogeneous queries that range from simple factoid questions to complex multi-hop reasoning. We propose a method that automatically selects a small, diverse subset of retrievers (a portfolio) from a large pool of candidates, to cover different regions of the target query distribution. We formalize this setting via an expected best-of-$k$ objective over the query distribution and show that it admits an efficient portfolio construction algorithm with near-optimal guarantees. Across multiple QA benchmarks, our learned portfolios and router pipeline consistently outperform single-retriever and naive multi-retriever baselines on both retrieval metrics and answer quality. In addition, compared to inference-time hyperparameter tuning approaches, fixed portfolios enable parallel retrieval and LLM calls, achieving comparable (and sometimes better) accuracy with substantially lower latency and token cost.
title Retriever Portfolios: A Principled Approach to Adaptive RAG
topic Machine Learning
Data Structures and Algorithms
url https://arxiv.org/abs/2605.31176