<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="li-etal-2026-timeblind">
    <titleInfo>
        <title>TimeBlind: A Spatio-Temporal Compositionality Benchmark for Video LLMs</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Baiqi</namePart>
        <namePart type="family">Li</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kangyi</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ce</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Chancharik</namePart>
        <namePart type="family">Mitra</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jean</namePart>
        <namePart type="given">De</namePart>
        <namePart type="given">Dieu</namePart>
        <namePart type="family">Nyandwi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Gedas</namePart>
        <namePart type="family">Bertasius</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>Fine-grained spatio-temporal understanding is essential for video reasoning and embodied AI. Yet, while Multimodal Large Language Models (MLLMs) master static semantics, their grasp of temporal dynamics remains brittle. We present TimeBlind, a diagnostic benchmark for compositional spatio-temporal understanding. Inspired by cognitive science, TimeBlind categorizes fine-grained temporal understanding into three levels: recognizing atomic events, characterizing event properties, and reasoning about event interdependencies. Unlike benchmarks that conflate recognition with temporal reasoning, TimeBlind leverages a minimal-pairs paradigm: video pairs share identical static visual content but differ solely in temporal structure, utilizing complementary questions to neutralize language priors. Evaluating over 20 state-of-the-art MLLMs (e.g., GPT-5, Gemini 3 Pro) on 600 curated instances (2400 video-question pairs), reveals that the Instance Accuracy (correctly distinguishing both videos in a pair) of the best performing MLLM is only 48.2%, far below the human performance (98.2%). These results demonstrate that even frontier models exhibit significant deficiencies in temporal reasoning, positioning TimeBlind as a vital diagnostic tool for next-generation video understanding.</abstract>
    <identifier type="citekey">li-etal-2026-timeblind</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.1923/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>43801</start>
            <end>43822</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T TimeBlind: A Spatio-Temporal Compositionality Benchmark for Video LLMs
%A Li, Baiqi
%A Zhao, Kangyi
%A Zhang, Ce
%A Mitra, Chancharik
%A Nyandwi, Jean De Dieu
%A Bertasius, Gedas
%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 li-etal-2026-timeblind
%X Fine-grained spatio-temporal understanding is essential for video reasoning and embodied AI. Yet, while Multimodal Large Language Models (MLLMs) master static semantics, their grasp of temporal dynamics remains brittle. We present TimeBlind, a diagnostic benchmark for compositional spatio-temporal understanding. Inspired by cognitive science, TimeBlind categorizes fine-grained temporal understanding into three levels: recognizing atomic events, characterizing event properties, and reasoning about event interdependencies. Unlike benchmarks that conflate recognition with temporal reasoning, TimeBlind leverages a minimal-pairs paradigm: video pairs share identical static visual content but differ solely in temporal structure, utilizing complementary questions to neutralize language priors. Evaluating over 20 state-of-the-art MLLMs (e.g., GPT-5, Gemini 3 Pro) on 600 curated instances (2400 video-question pairs), reveals that the Instance Accuracy (correctly distinguishing both videos in a pair) of the best performing MLLM is only 48.2%, far below the human performance (98.2%). These results demonstrate that even frontier models exhibit significant deficiencies in temporal reasoning, positioning TimeBlind as a vital diagnostic tool for next-generation video understanding.
%U https://aclanthology.org/2026.emnlp-main.1923/
%P 43801-43822

Markdown (Informal)

[TimeBlind: A Spatio-Temporal Compositionality Benchmark for Video LLMs](https://aclanthology.org/2026.emnlp-main.1923/) (Li et al., EMNLP 2026)

• TimeBlind: A Spatio-Temporal Compositionality Benchmark for Video LLMs (Li et al., EMNLP 2026)

ACL

• Baiqi Li, Kangyi Zhao, Ce Zhang, Chancharik Mitra, Jean De Dieu Nyandwi, and Gedas Bertasius. 2026. TimeBlind: A Spatio-Temporal Compositionality Benchmark for Video LLMs. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 43801–43822, Budapest, Hungary. Association for Computational Linguistics.