@inproceedings{liu-etal-2026-unigui,
title = "{U}ni{GUI}-Bench: A Real-World Diagnostic Benchmark {---} What Capabilities Do {GUI} Agents Most Need?",
author = "Liu, Jiaxing and
Sun, Ruoqi and
Ma, Liya and
Chen, FengJiao and
Wang, Ziwen and
Cao, Xuezhi",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.2288/",
pages = "49040--49053",
ISBN = "979-8-89176-462-0",
abstract = "GUI agents powered by Multimodal Large Language Models (MLLMs) are advancing rapidly, yet most existing benchmarks evaluate only task-level success or failure{---}leaving a fundamental question unanswered: \textit{what capabilities do GUI agents most need to succeed in real-world settings?} We introduce \textbf{UniGUI-Bench}, a benchmark that jointly verifies task-level outcomes and process-level capability diagnosis. It spans 6 device platforms and 23 application scenarios with bilingual support, comprising over 10,000 evaluation items. Its core design links the two evaluation axes: each task failure can be traced to specific capability deficiencies, enabling fine-grained attribution beyond binary pass/fail. Experiments on 14 frontier models reveal three findings: (1) model rankings diverge substantially from single-platform leaderboards under broader evaluation; (2) different scenarios impose distinct capability bottlenecks identifiable through our attribution framework; (3) targeted training on diagnosed bottleneck dimensions yields consistent improvements on both UniGUI-Bench and OSWorld, providing preliminary evidence toward closing the evaluation{--}improvement loop. The UniGUI-Bench dataset and evaluation code are available at \url{https://huggingface.co/datasets/AGI-Eval/UniGUI-Bench}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-unigui">
<titleInfo>
<title>UniGUI-Bench: A Real-World Diagnostic Benchmark — What Capabilities Do GUI Agents Most Need?</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jiaxing</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruoqi</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liya</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">FengJiao</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ziwen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuezhi</namePart>
<namePart type="family">Cao</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>GUI agents powered by Multimodal Large Language Models (MLLMs) are advancing rapidly, yet most existing benchmarks evaluate only task-level success or failure—leaving a fundamental question unanswered: what capabilities do GUI agents most need to succeed in real-world settings? We introduce UniGUI-Bench, a benchmark that jointly verifies task-level outcomes and process-level capability diagnosis. It spans 6 device platforms and 23 application scenarios with bilingual support, comprising over 10,000 evaluation items. Its core design links the two evaluation axes: each task failure can be traced to specific capability deficiencies, enabling fine-grained attribution beyond binary pass/fail. Experiments on 14 frontier models reveal three findings: (1) model rankings diverge substantially from single-platform leaderboards under broader evaluation; (2) different scenarios impose distinct capability bottlenecks identifiable through our attribution framework; (3) targeted training on diagnosed bottleneck dimensions yields consistent improvements on both UniGUI-Bench and OSWorld, providing preliminary evidence toward closing the evaluation–improvement loop. The UniGUI-Bench dataset and evaluation code are available at https://huggingface.co/datasets/AGI-Eval/UniGUI-Bench.</abstract>
<identifier type="citekey">liu-etal-2026-unigui</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2288/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>49040</start>
<end>49053</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T UniGUI-Bench: A Real-World Diagnostic Benchmark — What Capabilities Do GUI Agents Most Need?
%A Liu, Jiaxing
%A Sun, Ruoqi
%A Ma, Liya
%A Chen, FengJiao
%A Wang, Ziwen
%A Cao, Xuezhi
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F liu-etal-2026-unigui
%X GUI agents powered by Multimodal Large Language Models (MLLMs) are advancing rapidly, yet most existing benchmarks evaluate only task-level success or failure—leaving a fundamental question unanswered: what capabilities do GUI agents most need to succeed in real-world settings? We introduce UniGUI-Bench, a benchmark that jointly verifies task-level outcomes and process-level capability diagnosis. It spans 6 device platforms and 23 application scenarios with bilingual support, comprising over 10,000 evaluation items. Its core design links the two evaluation axes: each task failure can be traced to specific capability deficiencies, enabling fine-grained attribution beyond binary pass/fail. Experiments on 14 frontier models reveal three findings: (1) model rankings diverge substantially from single-platform leaderboards under broader evaluation; (2) different scenarios impose distinct capability bottlenecks identifiable through our attribution framework; (3) targeted training on diagnosed bottleneck dimensions yields consistent improvements on both UniGUI-Bench and OSWorld, providing preliminary evidence toward closing the evaluation–improvement loop. The UniGUI-Bench dataset and evaluation code are available at https://huggingface.co/datasets/AGI-Eval/UniGUI-Bench.
%U https://aclanthology.org/2026.findings-emnlp.2288/
%P 49040-49053
Markdown (Informal)
[UniGUI-Bench: A Real-World Diagnostic Benchmark — What Capabilities Do GUI Agents Most Need?](https://aclanthology.org/2026.findings-emnlp.2288/) (Liu et al., Findings 2026)
ACL