Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities?

Jiahe Jin; Yanheng He; Mingyan Yang

doi:10.18653/v1/2025.findings-acl.1222

Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities?

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

In this work, we identify the “2D-Cheating” problem in 3D LLM evaluation, where these tasks might be easily solved by VLMs with rendered images of point clouds, exposing ineffective evaluation of 3D LLMs’ unique 3D capabilities. We test VLM performance across multiple 3D LLM benchmarks and, using this as a reference, propose principles for better assessing genuine 3D understanding. We also advocate explicitly separating 3D abilities from 1D or 2D aspects when evaluating 3D LLMs.

Anthology ID:: 2025.findings-acl.1222
Volume:: Findings of the Association for Computational Linguistics: ACL 2025
Month:: July
Year:: 2025
Address:: Vienna, Austria
Editors:: Wanxiang Che, Joyce Nabende, Ekaterina Shutova, Mohammad Taher Pilehvar
Venue:: Findings
SIG:
Publisher:: Association for Computational Linguistics
Note:
Pages:: 23858–23869
Language:
URL:: https://aclanthology.org/2025.findings-acl.1222/
DOI:: 10.18653/v1/2025.findings-acl.1222
Bibkey:
Cite (ACL):: Jiahe Jin, Yanheng He, and Mingyan Yang. 2025. Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities?. In Findings of the Association for Computational Linguistics: ACL 2025, pages 23858–23869, Vienna, Austria. Association for Computational Linguistics.
Cite (Informal):: Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities? (Jin et al., Findings 2025)
Copy Citation:
PDF:: https://aclanthology.org/2025.findings-acl.1222.pdf

PDF Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{jin-etal-2025-revisiting,
    title = "Revisiting 3{D} {LLM} Benchmarks: Are We Really Testing 3{D} Capabilities?",
    author = "Jin, Jiahe  and
      He, Yanheng  and
      Yang, Mingyan",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.findings-acl.1222/",
    doi = "10.18653/v1/2025.findings-acl.1222",
    pages = "23858--23869",
    ISBN = "979-8-89176-256-5",
    abstract = "In this work, we identify the ``2D-Cheating'' problem in 3D LLM evaluation, where these tasks might be easily solved by VLMs with rendered images of point clouds, exposing ineffective evaluation of 3D LLMs' unique 3D capabilities. We test VLM performance across multiple 3D LLM benchmarks and, using this as a reference, propose principles for better assessing genuine 3D understanding. We also advocate explicitly separating 3D abilities from 1D or 2D aspects when evaluating 3D LLMs."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jin-etal-2025-revisiting">
    <titleInfo>
        <title>Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities?</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jiahe</namePart>
        <namePart type="family">Jin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yanheng</namePart>
        <namePart type="family">He</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Mingyan</namePart>
        <namePart type="family">Yang</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>Findings of the Association for Computational Linguistics: ACL 2025</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Wanxiang</namePart>
            <namePart type="family">Che</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Joyce</namePart>
            <namePart type="family">Nabende</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ekaterina</namePart>
            <namePart type="family">Shutova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mohammad</namePart>
            <namePart type="given">Taher</namePart>
            <namePart type="family">Pilehvar</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">979-8-89176-256-5</identifier>
    </relatedItem>
    <abstract>In this work, we identify the “2D-Cheating” problem in 3D LLM evaluation, where these tasks might be easily solved by VLMs with rendered images of point clouds, exposing ineffective evaluation of 3D LLMs’ unique 3D capabilities. We test VLM performance across multiple 3D LLM benchmarks and, using this as a reference, propose principles for better assessing genuine 3D understanding. We also advocate explicitly separating 3D abilities from 1D or 2D aspects when evaluating 3D LLMs.</abstract>
    <identifier type="citekey">jin-etal-2025-revisiting</identifier>
    <identifier type="doi">10.18653/v1/2025.findings-acl.1222</identifier>
    <location>
        <url>https://aclanthology.org/2025.findings-acl.1222/</url>
    </location>
    <part>
        <date>2025-07</date>
        <extent unit="page">
            <start>23858</start>
            <end>23869</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities?
%A Jin, Jiahe
%A He, Yanheng
%A Yang, Mingyan
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F jin-etal-2025-revisiting
%X In this work, we identify the “2D-Cheating” problem in 3D LLM evaluation, where these tasks might be easily solved by VLMs with rendered images of point clouds, exposing ineffective evaluation of 3D LLMs’ unique 3D capabilities. We test VLM performance across multiple 3D LLM benchmarks and, using this as a reference, propose principles for better assessing genuine 3D understanding. We also advocate explicitly separating 3D abilities from 1D or 2D aspects when evaluating 3D LLMs.
%R 10.18653/v1/2025.findings-acl.1222
%U https://aclanthology.org/2025.findings-acl.1222/
%U https://doi.org/10.18653/v1/2025.findings-acl.1222
%P 23858-23869

Download as File

Markdown (Informal)

[Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities?](https://aclanthology.org/2025.findings-acl.1222/) (Jin et al., Findings 2025)

Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities? (Jin et al., Findings 2025)

ACL

Jiahe Jin, Yanheng He, and Mingyan Yang. 2025. Revisiting 3D LLM Benchmarks: Are We Really Testing 3D Capabilities?. In Findings of the Association for Computational Linguistics: ACL 2025, pages 23858–23869, Vienna, Austria. Association for Computational Linguistics.