Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models

Christian Jaumann; Annemarie Friedrich; Rainer Lienhart

doi:10.18653/v1/2025.sdp-1.21

Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models

Christian Jaumann, Annemarie Friedrich, Rainer Lienhart

Correct Metadata for

Important: The Anthology treat PDFs as authoritative. Please use this form only to correct data that is out of line with the PDF. See our corrections guidelines if you need to change the PDF.

Title Adjust the title. Retain tags such as <fixed-case>.

Authors Adjust author names and order to match the PDF.

Abstract Correct abstract if needed. Retain XML formatting tags such as <tex-math>.

Verification against PDF Ensure that the new title/authors match the snapshot below. (If there is no snapshot or it is too small, consult the PDF.)

Authors concatenated from the text boxes above:

ALL author names match the snapshot above—including middle initials, hyphens, and accents.

Abstract

This paper describes our system for the SciVQA 2025 Shared Task on Scientific Visual Question Answering. Our system employs an ensemble of two Multimodal Large Language Models and various few-shot example retrieval strategies. The model and few-shot setting are selected based on the figure and question type. We also select answers based on the models’ confidence levels. On the blind test data, our system ranks third out of seven with an average F1 score of 85.12 across ROUGE-1, ROUGE-L, and BERTS. Our code is publicly available.

Anthology ID:: 2025.sdp-1.21
Volume:: Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025)
Month:: July
Year:: 2025
Address:: Vienna, Austria
Editors:: Tirthankar Ghosal, Philipp Mayr, Amanpreet Singh, Aakanksha Naik, Georg Rehm, Dayne Freitag, Dan Li, Sonja Schimmler, Anita De Waard
Venues:: sdp | WS
SIG:
Publisher:: Association for Computational Linguistics
Note:
Pages:: 230–239
Language:
URL:: https://aclanthology.org/2025.sdp-1.21/
DOI:: 10.18653/v1/2025.sdp-1.21
Bibkey:
Cite (ACL):: Christian Jaumann, Annemarie Friedrich, and Rainer Lienhart. 2025. Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models. In Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025), pages 230–239, Vienna, Austria. Association for Computational Linguistics.
Cite (Informal):: Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models (Jaumann et al., sdp 2025)
Copy Citation:
PDF:: https://aclanthology.org/2025.sdp-1.21.pdf

PDF Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{jaumann-etal-2025-coling,
    title = "Coling-{U}ni{A} at {S}ci{VQA} 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models",
    author = "Jaumann, Christian  and
      Friedrich, Annemarie  and
      Lienhart, Rainer",
    editor = "Ghosal, Tirthankar  and
      Mayr, Philipp  and
      Singh, Amanpreet  and
      Naik, Aakanksha  and
      Rehm, Georg  and
      Freitag, Dayne  and
      Li, Dan  and
      Schimmler, Sonja  and
      De Waard, Anita",
    booktitle = "Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.sdp-1.21/",
    doi = "10.18653/v1/2025.sdp-1.21",
    pages = "230--239",
    ISBN = "979-8-89176-265-7",
    abstract = "This paper describes our system for the SciVQA 2025 Shared Task on Scientific Visual Question Answering. Our system employs an ensemble of two Multimodal Large Language Models and various few-shot example retrieval strategies. The model and few-shot setting are selected based on the figure and question type. We also select answers based on the models' confidence levels. On the blind test data, our system ranks third out of seven with an average F1 score of 85.12 across ROUGE-1, ROUGE-L, and BERTS. Our code is publicly available."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jaumann-etal-2025-coling">
    <titleInfo>
        <title>Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Christian</namePart>
        <namePart type="family">Jaumann</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Annemarie</namePart>
        <namePart type="family">Friedrich</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Rainer</namePart>
        <namePart type="family">Lienhart</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Tirthankar</namePart>
            <namePart type="family">Ghosal</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Philipp</namePart>
            <namePart type="family">Mayr</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Amanpreet</namePart>
            <namePart type="family">Singh</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Aakanksha</namePart>
            <namePart type="family">Naik</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Georg</namePart>
            <namePart type="family">Rehm</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Dayne</namePart>
            <namePart type="family">Freitag</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Dan</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sonja</namePart>
            <namePart type="family">Schimmler</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Anita</namePart>
            <namePart type="family">De Waard</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Vienna, Austria</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-265-7</identifier>
    </relatedItem>
    <abstract>This paper describes our system for the SciVQA 2025 Shared Task on Scientific Visual Question Answering. Our system employs an ensemble of two Multimodal Large Language Models and various few-shot example retrieval strategies. The model and few-shot setting are selected based on the figure and question type. We also select answers based on the models’ confidence levels. On the blind test data, our system ranks third out of seven with an average F1 score of 85.12 across ROUGE-1, ROUGE-L, and BERTS. Our code is publicly available.</abstract>
    <identifier type="citekey">jaumann-etal-2025-coling</identifier>
    <identifier type="doi">10.18653/v1/2025.sdp-1.21</identifier>
    <location>
        <url>https://aclanthology.org/2025.sdp-1.21/</url>
    </location>
    <part>
        <date>2025-07</date>
        <extent unit="page">
            <start>230</start>
            <end>239</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models
%A Jaumann, Christian
%A Friedrich, Annemarie
%A Lienhart, Rainer
%Y Ghosal, Tirthankar
%Y Mayr, Philipp
%Y Singh, Amanpreet
%Y Naik, Aakanksha
%Y Rehm, Georg
%Y Freitag, Dayne
%Y Li, Dan
%Y Schimmler, Sonja
%Y De Waard, Anita
%S Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-265-7
%F jaumann-etal-2025-coling
%X This paper describes our system for the SciVQA 2025 Shared Task on Scientific Visual Question Answering. Our system employs an ensemble of two Multimodal Large Language Models and various few-shot example retrieval strategies. The model and few-shot setting are selected based on the figure and question type. We also select answers based on the models’ confidence levels. On the blind test data, our system ranks third out of seven with an average F1 score of 85.12 across ROUGE-1, ROUGE-L, and BERTS. Our code is publicly available.
%R 10.18653/v1/2025.sdp-1.21
%U https://aclanthology.org/2025.sdp-1.21/
%U https://doi.org/10.18653/v1/2025.sdp-1.21
%P 230-239

Download as File

Markdown (Informal)

[Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models](https://aclanthology.org/2025.sdp-1.21/) (Jaumann et al., sdp 2025)

Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models (Jaumann et al., sdp 2025)

ACL

Christian Jaumann, Annemarie Friedrich, and Rainer Lienhart. 2025. Coling-UniA at SciVQA 2025: Few-Shot Example Retrieval and Confidence-Informed Ensembling for Multimodal Large Language Models. In Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025), pages 230–239, Vienna, Austria. Association for Computational Linguistics.