<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lv-etal-2026-hear">
    <titleInfo>
        <title>HEAR-RAG: A Hypergraph-Based Evidence-Aware Retrieval-Augmented Generation Framework for Multi-Hop Question Answering</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Changbin</namePart>
        <namePart type="family">Lv</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Weihua</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Lianlong</namePart>
        <namePart type="family">Tao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jiale</namePart>
        <namePart type="family">Xing</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Feilong</namePart>
        <namePart type="family">Bao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Hung-Yi</namePart>
            <namePart type="family">Lee</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sunipa</namePart>
            <namePart type="family">Dev</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jessy</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Desmond</namePart>
            <namePart type="family">Elliott</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>Retrieval-Augmented Generation (RAG) improves model factuality, but multi-hop question answering still suffers from incomplete and disconnected evidence retrieval across documents or chunks. Although graph-enhanced RAG methods organize evidence with graphs or hypergraphs, existing structures still struggle to balance the completeness of higher-order factual structures with fine-grained evidence connectivity. Graph-based methods mainly rely on pairwise edges and may fail to preserve complex multi-entity facts, while hypergraph-based methods can capture higher-order contexts but may underuse low-order relations for step-by-step evidence tracing. Both types of methods may further suffer from evidence fragmentation when semantically equivalent entity mentions are not effectively organized across chunks. To address these issues, we propose HEAR-RAG, a hypergraph-based evidence-aware retrieval framework for multi-hop question answering. HEAR-RAG organizes distributed evidence into a unified factual evidence hypergraph that jointly preserves higher-order factual contexts, low-order relational clues, and source chunks. It further introduces relation-aware entity mention clustering to reduce evidence fragmentation, and uses hybrid structural retrieval to construct connected evidence paths for answer generation. Experiments on multiple multi-hop QA benchmarks demonstrate its effectiveness.</abstract>
    <identifier type="citekey">lv-etal-2026-hear</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2210/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>50274</start>
            <end>50289</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T HEAR-RAG: A Hypergraph-Based Evidence-Aware Retrieval-Augmented Generation Framework for Multi-Hop Question Answering
%A Lv, Changbin
%A Wang, Weihua
%A Tao, Lianlong
%A Xing, Jiale
%A Bao, Feilong
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F lv-etal-2026-hear
%X Retrieval-Augmented Generation (RAG) improves model factuality, but multi-hop question answering still suffers from incomplete and disconnected evidence retrieval across documents or chunks. Although graph-enhanced RAG methods organize evidence with graphs or hypergraphs, existing structures still struggle to balance the completeness of higher-order factual structures with fine-grained evidence connectivity. Graph-based methods mainly rely on pairwise edges and may fail to preserve complex multi-entity facts, while hypergraph-based methods can capture higher-order contexts but may underuse low-order relations for step-by-step evidence tracing. Both types of methods may further suffer from evidence fragmentation when semantically equivalent entity mentions are not effectively organized across chunks. To address these issues, we propose HEAR-RAG, a hypergraph-based evidence-aware retrieval framework for multi-hop question answering. HEAR-RAG organizes distributed evidence into a unified factual evidence hypergraph that jointly preserves higher-order factual contexts, low-order relational clues, and source chunks. It further introduces relation-aware entity mention clustering to reduce evidence fragmentation, and uses hybrid structural retrieval to construct connected evidence paths for answer generation. Experiments on multiple multi-hop QA benchmarks demonstrate its effectiveness.
%U https://aclanthology.org/2026.emnlp-main.2210/
%P 50274-50289

Markdown (Informal)

[HEAR-RAG: A Hypergraph-Based Evidence-Aware Retrieval-Augmented Generation Framework for Multi-Hop Question Answering](https://aclanthology.org/2026.emnlp-main.2210/) (Lv et al., EMNLP 2026)

• HEAR-RAG: A Hypergraph-Based Evidence-Aware Retrieval-Augmented Generation Framework for Multi-Hop Question Answering (Lv et al., EMNLP 2026)

ACL

• Changbin Lv, Weihua Wang, Lianlong Tao, Jiale Xing, and Feilong Bao. 2026. HEAR-RAG: A Hypergraph-Based Evidence-Aware Retrieval-Augmented Generation Framework for Multi-Hop Question Answering. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 50274–50289, Budapest, Hungary. Association for Computational Linguistics.