@inproceedings{jiang-etal-2026-vista,
title = "Vista: An Efficient Late Interaction Engine for Multi-Vector Visual Document Retrieval",
author = "Jiang, Haolin and
Pan, Jinwen and
Saini, Abhishek and
Li, Weihang and
Liu, Hang",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2135/",
pages = "48529--48543",
ISBN = "979-8-89176-459-0",
abstract = "Visual document retrieval (VDR) bridges language-based systems with visually grounded knowledge. Recent VDR models adopt multi-vector late interaction, representing each document with multiple visual token embeddings and scoring relevance with MaxSim. However, existing late interaction engines struggle to support multi-vector VDR at scale due to the higher dimensionality and larger number of visual tokens per document.\par We propose VISTA, a VDR-specific late-interaction engine that exploits the correlation among visual tokens within each document to optimize both compression and search. To the best of our knowledge, VISTA is the first VDR-specific late-interaction engine for scalable multi-vector visual document retrieval. Experiments on ViDoRe V2 and ViDoRe V3 show that VISTA achieves comparable or even better retrieval quality than compressed baselines, while consistently reducing latency by 1.2{--}6.7$\times$ and index size by 1.4{--}6.1$\times$."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jiang-etal-2026-vista">
<titleInfo>
<title>Vista: An Efficient Late Interaction Engine for Multi-Vector Visual Document Retrieval</title>
</titleInfo>
<name type="personal">
<namePart type="given">Haolin</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinwen</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Abhishek</namePart>
<namePart type="family">Saini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weihang</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hang</namePart>
<namePart type="family">Liu</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>Visual document retrieval (VDR) bridges language-based systems with visually grounded knowledge. Recent VDR models adopt multi-vector late interaction, representing each document with multiple visual token embeddings and scoring relevance with MaxSim. However, existing late interaction engines struggle to support multi-vector VDR at scale due to the higher dimensionality and larger number of visual tokens per document.\par We propose VISTA, a VDR-specific late-interaction engine that exploits the correlation among visual tokens within each document to optimize both compression and search. To the best of our knowledge, VISTA is the first VDR-specific late-interaction engine for scalable multi-vector visual document retrieval. Experiments on ViDoRe V2 and ViDoRe V3 show that VISTA achieves comparable or even better retrieval quality than compressed baselines, while consistently reducing latency by 1.2–6.7\times and index size by 1.4–6.1\times.</abstract>
<identifier type="citekey">jiang-etal-2026-vista</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2135/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>48529</start>
<end>48543</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Vista: An Efficient Late Interaction Engine for Multi-Vector Visual Document Retrieval
%A Jiang, Haolin
%A Pan, Jinwen
%A Saini, Abhishek
%A Li, Weihang
%A Liu, Hang
%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 jiang-etal-2026-vista
%X Visual document retrieval (VDR) bridges language-based systems with visually grounded knowledge. Recent VDR models adopt multi-vector late interaction, representing each document with multiple visual token embeddings and scoring relevance with MaxSim. However, existing late interaction engines struggle to support multi-vector VDR at scale due to the higher dimensionality and larger number of visual tokens per document.\par We propose VISTA, a VDR-specific late-interaction engine that exploits the correlation among visual tokens within each document to optimize both compression and search. To the best of our knowledge, VISTA is the first VDR-specific late-interaction engine for scalable multi-vector visual document retrieval. Experiments on ViDoRe V2 and ViDoRe V3 show that VISTA achieves comparable or even better retrieval quality than compressed baselines, while consistently reducing latency by 1.2–6.7\times and index size by 1.4–6.1\times.
%U https://aclanthology.org/2026.emnlp-main.2135/
%P 48529-48543
Markdown (Informal)
[Vista: An Efficient Late Interaction Engine for Multi-Vector Visual Document Retrieval](https://aclanthology.org/2026.emnlp-main.2135/) (Jiang et al., EMNLP 2026)
ACL