<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="abro-etal-2026-learning">
    <titleInfo>
        <title>Learning Long-Document Embeddings via Chunk–Context Entailment</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Waheed</namePart>
        <namePart type="given">Ahmed</namePart>
        <namePart type="family">Abro</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Naïm</namePart>
        <namePart type="family">Es-Sebbani</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zied</namePart>
        <namePart type="family">Bouraoui</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-05</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Fifteenth Language Resources and Evaluation Conference</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Stelios</namePart>
            <namePart type="family">Piperidis</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Núria</namePart>
            <namePart type="family">Bel</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Henk</namePart>
            <namePart type="family">van den Heuvel</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Nancy</namePart>
            <namePart type="family">Ide</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Simon</namePart>
            <namePart type="family">Krek</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Antonio</namePart>
            <namePart type="family">Toral</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>ELRA Language Resource Association</publisher>
            <place>
                <placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>Learning faithful embeddings for long documents remains challenging, especially in domains like law and medicine where inputs are long, structured, and semantically heterogeneous. We introduce the Chunk Prediction Encoder (CPE), a self-supervised framework that treats chunk–context compatibility as an unsupervised NLI problem. Given a document, CPE masks a chunk and learns (i) a contrastive objective that aligns the masked document with its held-out chunk against in-batch negatives, and (ii) a binary entailment head that predicts whether a candidate chunk belongs to the document. This joint objective encourages both geometric smoothness and directional semantic consistency, yielding robust document-level embeddings. We evaluate CPE with hierarchical and sparse-attention backbones on five benchmarks spanning legal and biomedical domains under frozen-embedding and end-to-end fine-tuning protocols. CPE consistently outperforms baselines, and is more compute-efficient than prompt-only LLM baselines under matched token budgets. Ablations demonstrate the effect of chunk length, the contrastive-vs-entailment balance, and skimming strategies.</abstract>
    <identifier type="citekey">abro-etal-2026-learning</identifier>
    <identifier type="doi">10.63317/4iz34o26i4tt</identifier>
    <location>
        <url>https://aclanthology.org/2026.lrec-1.586/</url>
    </location>
    <part>
        <date>2026-05</date>
        <extent unit="page">
            <start>7405</start>
            <end>7414</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Learning Long-Document Embeddings via Chunk–Context Entailment
%A Abro, Waheed Ahmed
%A Es-Sebbani, Naïm
%A Bouraoui, Zied
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F abro-etal-2026-learning
%X Learning faithful embeddings for long documents remains challenging, especially in domains like law and medicine where inputs are long, structured, and semantically heterogeneous. We introduce the Chunk Prediction Encoder (CPE), a self-supervised framework that treats chunk–context compatibility as an unsupervised NLI problem. Given a document, CPE masks a chunk and learns (i) a contrastive objective that aligns the masked document with its held-out chunk against in-batch negatives, and (ii) a binary entailment head that predicts whether a candidate chunk belongs to the document. This joint objective encourages both geometric smoothness and directional semantic consistency, yielding robust document-level embeddings. We evaluate CPE with hierarchical and sparse-attention backbones on five benchmarks spanning legal and biomedical domains under frozen-embedding and end-to-end fine-tuning protocols. CPE consistently outperforms baselines, and is more compute-efficient than prompt-only LLM baselines under matched token budgets. Ablations demonstrate the effect of chunk length, the contrastive-vs-entailment balance, and skimming strategies.
%R 10.63317/4iz34o26i4tt
%U https://aclanthology.org/2026.lrec-1.586/
%U https://doi.org/10.63317/4iz34o26i4tt
%P 7405-7414

Markdown (Informal)

[Learning Long-Document Embeddings via Chunk–Context Entailment](https://aclanthology.org/2026.lrec-1.586/) (Abro et al., LREC 2026)

• Learning Long-Document Embeddings via Chunk–Context Entailment (Abro et al., LREC 2026)

ACL

• Waheed Ahmed Abro, Naïm Es-Sebbani, and Zied Bouraoui. 2026. Learning Long-Document Embeddings via Chunk–Context Entailment. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, pages 7405–7414, Palma de Mallorca, Spain. ELRA Language Resource Association.