MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents

Multi-modal document retrieval is designed to identify and retrieve various forms of multi-modal content, such as figures, tables, charts, and layout information from extensive documents. Despite its significance, there is a notable lack of a robust benchmark to effectively evaluate the performance...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Dong, Kuicai, Chang, Yujing, Goh, Xin Deik, Li, Dexun, Tang, Ruiming, Liu, Yong
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Dong, Kuicai
Chang, Yujing
Goh, Xin Deik
Li, Dexun
Tang, Ruiming
Liu, Yong
description Multi-modal document retrieval is designed to identify and retrieve various forms of multi-modal content, such as figures, tables, charts, and layout information from extensive documents. Despite its significance, there is a notable lack of a robust benchmark to effectively evaluate the performance of systems in multi-modal document retrieval. To address this gap, this work introduces a new benchmark, named as MMDocIR, encompassing two distinct tasks: page-level and layout-level retrieval. The former focuses on localizing the most relevant pages within a long document, while the latter targets the detection of specific layouts, offering a more fine-grained granularity than whole-page analysis. A layout can refer to a variety of elements such as textual paragraphs, equations, figures, tables, or charts. The MMDocIR benchmark comprises a rich dataset featuring expertly annotated labels for 1,685 questions and bootstrapped labels for 173,843 questions, making it a pivotal resource for advancing multi-modal document retrieval for both training and evaluation. Through rigorous experiments, we reveal that (i) visual retrievers significantly outperform their text counterparts, (ii) MMDocIR train set can effectively benefit the training process of multi-modal document retrieval and (iii) text retrievers leveraging on VLM-text perform much better than those using OCR-text. These findings underscores the potential advantages of integrating visual elements for multi-modal document retrieval.
doi_str_mv 10.48550/arxiv.2501.08828
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2501_08828</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2501_08828</sourcerecordid><originalsourceid>FETCH-arxiv_primary_2501_088283</originalsourceid><addsrcrecordid>eNpjYJA0NNAzsTA1NdBPLKrILNMzMjUw1DOwsDCy4GSw8_V1yU_2DLJScErNS87ITSzKzsxLV_AtzSnJ1PXNT0nMUQhKLSnKTC0DstLyixR88oHSQC2lual5JcU8DKxpiTnFqbxQmptB3s01xNlDF2xTfEFRJtDEyniQjfFgG40JqwAAFK81Qg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents</title><source>arXiv.org</source><creator>Dong, Kuicai ; Chang, Yujing ; Goh, Xin Deik ; Li, Dexun ; Tang, Ruiming ; Liu, Yong</creator><creatorcontrib>Dong, Kuicai ; Chang, Yujing ; Goh, Xin Deik ; Li, Dexun ; Tang, Ruiming ; Liu, Yong</creatorcontrib><description>Multi-modal document retrieval is designed to identify and retrieve various forms of multi-modal content, such as figures, tables, charts, and layout information from extensive documents. Despite its significance, there is a notable lack of a robust benchmark to effectively evaluate the performance of systems in multi-modal document retrieval. To address this gap, this work introduces a new benchmark, named as MMDocIR, encompassing two distinct tasks: page-level and layout-level retrieval. The former focuses on localizing the most relevant pages within a long document, while the latter targets the detection of specific layouts, offering a more fine-grained granularity than whole-page analysis. A layout can refer to a variety of elements such as textual paragraphs, equations, figures, tables, or charts. The MMDocIR benchmark comprises a rich dataset featuring expertly annotated labels for 1,685 questions and bootstrapped labels for 173,843 questions, making it a pivotal resource for advancing multi-modal document retrieval for both training and evaluation. Through rigorous experiments, we reveal that (i) visual retrievers significantly outperform their text counterparts, (ii) MMDocIR train set can effectively benefit the training process of multi-modal document retrieval and (iii) text retrievers leveraging on VLM-text perform much better than those using OCR-text. These findings underscores the potential advantages of integrating visual elements for multi-modal document retrieval.</description><identifier>DOI: 10.48550/arxiv.2501.08828</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Computation and Language ; Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Information Retrieval</subject><creationdate>2025-01</creationdate><rights>http://creativecommons.org/licenses/by-nc-sa/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2501.08828$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2501.08828$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Dong, Kuicai</creatorcontrib><creatorcontrib>Chang, Yujing</creatorcontrib><creatorcontrib>Goh, Xin Deik</creatorcontrib><creatorcontrib>Li, Dexun</creatorcontrib><creatorcontrib>Tang, Ruiming</creatorcontrib><creatorcontrib>Liu, Yong</creatorcontrib><title>MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents</title><description>Multi-modal document retrieval is designed to identify and retrieve various forms of multi-modal content, such as figures, tables, charts, and layout information from extensive documents. Despite its significance, there is a notable lack of a robust benchmark to effectively evaluate the performance of systems in multi-modal document retrieval. To address this gap, this work introduces a new benchmark, named as MMDocIR, encompassing two distinct tasks: page-level and layout-level retrieval. The former focuses on localizing the most relevant pages within a long document, while the latter targets the detection of specific layouts, offering a more fine-grained granularity than whole-page analysis. A layout can refer to a variety of elements such as textual paragraphs, equations, figures, tables, or charts. The MMDocIR benchmark comprises a rich dataset featuring expertly annotated labels for 1,685 questions and bootstrapped labels for 173,843 questions, making it a pivotal resource for advancing multi-modal document retrieval for both training and evaluation. Through rigorous experiments, we reveal that (i) visual retrievers significantly outperform their text counterparts, (ii) MMDocIR train set can effectively benefit the training process of multi-modal document retrieval and (iii) text retrievers leveraging on VLM-text perform much better than those using OCR-text. These findings underscores the potential advantages of integrating visual elements for multi-modal document retrieval.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Computation and Language</subject><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Information Retrieval</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2025</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNpjYJA0NNAzsTA1NdBPLKrILNMzMjUw1DOwsDCy4GSw8_V1yU_2DLJScErNS87ITSzKzsxLV_AtzSnJ1PXNT0nMUQhKLSnKTC0DstLyixR88oHSQC2lual5JcU8DKxpiTnFqbxQmptB3s01xNlDF2xTfEFRJtDEyniQjfFgG40JqwAAFK81Qg</recordid><startdate>20250115</startdate><enddate>20250115</enddate><creator>Dong, Kuicai</creator><creator>Chang, Yujing</creator><creator>Goh, Xin Deik</creator><creator>Li, Dexun</creator><creator>Tang, Ruiming</creator><creator>Liu, Yong</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20250115</creationdate><title>MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents</title><author>Dong, Kuicai ; Chang, Yujing ; Goh, Xin Deik ; Li, Dexun ; Tang, Ruiming ; Liu, Yong</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-arxiv_primary_2501_088283</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2025</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Computation and Language</topic><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Information Retrieval</topic><toplevel>online_resources</toplevel><creatorcontrib>Dong, Kuicai</creatorcontrib><creatorcontrib>Chang, Yujing</creatorcontrib><creatorcontrib>Goh, Xin Deik</creatorcontrib><creatorcontrib>Li, Dexun</creatorcontrib><creatorcontrib>Tang, Ruiming</creatorcontrib><creatorcontrib>Liu, Yong</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Dong, Kuicai</au><au>Chang, Yujing</au><au>Goh, Xin Deik</au><au>Li, Dexun</au><au>Tang, Ruiming</au><au>Liu, Yong</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents</atitle><date>2025-01-15</date><risdate>2025</risdate><abstract>Multi-modal document retrieval is designed to identify and retrieve various forms of multi-modal content, such as figures, tables, charts, and layout information from extensive documents. Despite its significance, there is a notable lack of a robust benchmark to effectively evaluate the performance of systems in multi-modal document retrieval. To address this gap, this work introduces a new benchmark, named as MMDocIR, encompassing two distinct tasks: page-level and layout-level retrieval. The former focuses on localizing the most relevant pages within a long document, while the latter targets the detection of specific layouts, offering a more fine-grained granularity than whole-page analysis. A layout can refer to a variety of elements such as textual paragraphs, equations, figures, tables, or charts. The MMDocIR benchmark comprises a rich dataset featuring expertly annotated labels for 1,685 questions and bootstrapped labels for 173,843 questions, making it a pivotal resource for advancing multi-modal document retrieval for both training and evaluation. Through rigorous experiments, we reveal that (i) visual retrievers significantly outperform their text counterparts, (ii) MMDocIR train set can effectively benefit the training process of multi-modal document retrieval and (iii) text retrievers leveraging on VLM-text perform much better than those using OCR-text. These findings underscores the potential advantages of integrating visual elements for multi-modal document retrieval.</abstract><doi>10.48550/arxiv.2501.08828</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2501.08828
ispartof
issn
language eng
recordid cdi_arxiv_primary_2501_08828
source arXiv.org
subjects Computer Science - Artificial Intelligence
Computer Science - Computation and Language
Computer Science - Computer Vision and Pattern Recognition
Computer Science - Information Retrieval
title MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-10T05%3A32%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=MMDocIR:%20Benchmarking%20Multi-Modal%20Retrieval%20for%20Long%20Documents&rft.au=Dong,%20Kuicai&rft.date=2025-01-15&rft_id=info:doi/10.48550/arxiv.2501.08828&rft_dat=%3Carxiv_GOX%3E2501_08828%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true