LongRAG: A Dual-Perspective Retrieval-Augmented Generation Paradigm for Long-Context Question Answering

TL;DR

LongRAG introduces a dual-perspective retrieval-augmented generation system that significantly improves long-context question answering by better integrating global information and factual details, outperforming existing models.

Contribution

The paper presents LongRAG, a novel dual-perspective RAG paradigm that enhances long-context understanding and is adaptable across domains and large language models.

Findings

LongRAG outperforms long-context LLMs by 6.94%.

LongRAG surpasses advanced RAG by 6.16%.

LongRAG exceeds Vanilla RAG by 17.25%.

Abstract

Long-Context Question Answering (LCQA), a challenging task, aims to reason over long-context documents to yield accurate answers to questions. Existing long-context Large Language Models (LLMs) for LCQA often struggle with the "lost in the middle" issue. Retrieval-Augmented Generation (RAG) mitigates this issue by providing external factual evidence. However, its chunking strategy disrupts the global long-context information, and its low-quality retrieval in long contexts hinders LLMs from identifying effective factual details due to substantial noise. To this end, we propose LongRAG, a general, dual-perspective, and robust LLM-based RAG system paradigm for LCQA to enhance RAG's understanding of complex long-context knowledge (i.e., global information and factual details). We design LongRAG as a plug-and-play paradigm, facilitating adaptation to various domains and LLMs. Extensive experiments on three multi-hop datasets demonstrate that LongRAG significantly outperforms long-context LLMs (up by 6.94%), advanced RAG (up by 6.16%), and Vanilla RAG (up by 17.25%). Furthermore, we conduct quantitative ablation studies and multi-dimensional analyses, highlighting the effectiveness of the system's components and fine-tuning strategies. Data and code are available at https://github.com/QingFei1/LongRAG.