Evaluating Multi-Hop Reasoning in RAG Systems: A Comparison of LLM-Based Retriever Evaluation Strategies

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Brehme, Lorenz, Ströhle, Thomas, Breu, Ruth
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866918456892325888
author Brehme, Lorenz
Ströhle, Thomas
Breu, Ruth
author_facet Brehme, Lorenz
Ströhle, Thomas
Breu, Ruth
contents Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge to answer questions more accurately. However, research on evaluating RAG systems-particularly the retriever component-remains limited, as most existing work focuses on single-context retrieval rather than multi-hop queries, where individual contexts may appear irrelevant in isolation but are essential when combined. In this research, we use the HotPotQA, MuSiQue, and SQuAD datasets to simulate a RAG system and compare three LLM-as-judge evaluation strategies, including our proposed Context-Aware Retriever Evaluation (CARE). Our goal is to better understand how multi-hop reasoning can be most effectively evaluated in RAG systems. Experiments with LLMs from OpenAI, Meta, and Google demonstrate that CARE consistently outperforms existing methods for evaluating multi-hop reasoning in RAG systems. The performance gains are most pronounced in models with larger parameter counts and longer context windows, while single-hop queries show minimal sensitivity to context-aware evaluation. Overall, the results highlight the critical role of context-aware evaluation in improving the reliability and accuracy of retrieval-augmented generation systems, particularly in complex query scenarios. To ensure reproducibility, we provide the complete data of our experiments at https://github.com/lorenzbrehme/CARE.
format Preprint
id arxiv_https___arxiv_org_abs_2604_18234
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Evaluating Multi-Hop Reasoning in RAG Systems: A Comparison of LLM-Based Retriever Evaluation Strategies
Brehme, Lorenz
Ströhle, Thomas
Breu, Ruth
Information Retrieval
Artificial Intelligence
Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge to answer questions more accurately. However, research on evaluating RAG systems-particularly the retriever component-remains limited, as most existing work focuses on single-context retrieval rather than multi-hop queries, where individual contexts may appear irrelevant in isolation but are essential when combined. In this research, we use the HotPotQA, MuSiQue, and SQuAD datasets to simulate a RAG system and compare three LLM-as-judge evaluation strategies, including our proposed Context-Aware Retriever Evaluation (CARE). Our goal is to better understand how multi-hop reasoning can be most effectively evaluated in RAG systems. Experiments with LLMs from OpenAI, Meta, and Google demonstrate that CARE consistently outperforms existing methods for evaluating multi-hop reasoning in RAG systems. The performance gains are most pronounced in models with larger parameter counts and longer context windows, while single-hop queries show minimal sensitivity to context-aware evaluation. Overall, the results highlight the critical role of context-aware evaluation in improving the reliability and accuracy of retrieval-augmented generation systems, particularly in complex query scenarios. To ensure reproducibility, we provide the complete data of our experiments at https://github.com/lorenzbrehme/CARE.
title Evaluating Multi-Hop Reasoning in RAG Systems: A Comparison of LLM-Based Retriever Evaluation Strategies
topic Information Retrieval
Artificial Intelligence
url https://arxiv.org/abs/2604.18234