@inproceedings{liu-etal-2026-task-oriented,
title = "Task-Oriented Moment Localization for Long Video Understanding with Near-Second Coverage",
author = "Liu, Xiangrui and
Qin, Minghao and
Shu, Yan and
Liang, Zhengyang and
Tian, Yang and
Zhang, Chen Jason and
Zhao, Bo and
Liu, Zheng",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2144/",
pages = "48748--48769",
ISBN = "979-8-89176-459-0",
abstract = "Long-video understanding (LVU) is increasingly framed as an agentic process that iteratively localizes task-critical moments before perception and reasoning. Moment localization is therefore a key bottleneck: under long-video context constraints, sparse sampling can easily miss decisive evidence, while similarity-based retrieval often returns semantically related but task-irrelevant clips. We present EviScan, an evidence-seeking moment localization model built on an MLLM, designed for task-oriented localization and near-second temporal scanning on hour-long videos. To learn task-oriented localization, we construct ToMo-Pile, a training set that pairs temporal intervals with rationale traces explaining why the interval resolves the task. To make near-second scanning feasible on long videos, EviScan uses DualKV Prefill for memory-efficient streaming inference, and Progressive Localization to first narrow down candidate intervals and then refine boundaries using high-resolution evidence. We further introduce ToMo-Bench, a benchmark for task-oriented moment localization across diverse long-video scenarios. EviScan consistently outperforms specialized localizers and general-purpose MLLMs on both ToMo-Bench and standard temporal grounding benchmarks. When used as a plug-in localization module, it yields substantial gains on downstream long-video understanding benchmarks (MLVU and LongVideoBench), and these gains are larger than those brought by other moment-localization methods. We will release the model, dataset, benchmark, and code."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-task-oriented">
<titleInfo>
<title>Task-Oriented Moment Localization for Long Video Understanding with Near-Second Coverage</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xiangrui</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghao</namePart>
<namePart type="family">Qin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yan</namePart>
<namePart type="family">Shu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhengyang</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Tian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chen</namePart>
<namePart type="given">Jason</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bo</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zheng</namePart>
<namePart type="family">Liu</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>Long-video understanding (LVU) is increasingly framed as an agentic process that iteratively localizes task-critical moments before perception and reasoning. Moment localization is therefore a key bottleneck: under long-video context constraints, sparse sampling can easily miss decisive evidence, while similarity-based retrieval often returns semantically related but task-irrelevant clips. We present EviScan, an evidence-seeking moment localization model built on an MLLM, designed for task-oriented localization and near-second temporal scanning on hour-long videos. To learn task-oriented localization, we construct ToMo-Pile, a training set that pairs temporal intervals with rationale traces explaining why the interval resolves the task. To make near-second scanning feasible on long videos, EviScan uses DualKV Prefill for memory-efficient streaming inference, and Progressive Localization to first narrow down candidate intervals and then refine boundaries using high-resolution evidence. We further introduce ToMo-Bench, a benchmark for task-oriented moment localization across diverse long-video scenarios. EviScan consistently outperforms specialized localizers and general-purpose MLLMs on both ToMo-Bench and standard temporal grounding benchmarks. When used as a plug-in localization module, it yields substantial gains on downstream long-video understanding benchmarks (MLVU and LongVideoBench), and these gains are larger than those brought by other moment-localization methods. We will release the model, dataset, benchmark, and code.</abstract>
<identifier type="citekey">liu-etal-2026-task-oriented</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2144/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>48748</start>
<end>48769</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Task-Oriented Moment Localization for Long Video Understanding with Near-Second Coverage
%A Liu, Xiangrui
%A Qin, Minghao
%A Shu, Yan
%A Liang, Zhengyang
%A Tian, Yang
%A Zhang, Chen Jason
%A Zhao, Bo
%A Liu, Zheng
%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 liu-etal-2026-task-oriented
%X Long-video understanding (LVU) is increasingly framed as an agentic process that iteratively localizes task-critical moments before perception and reasoning. Moment localization is therefore a key bottleneck: under long-video context constraints, sparse sampling can easily miss decisive evidence, while similarity-based retrieval often returns semantically related but task-irrelevant clips. We present EviScan, an evidence-seeking moment localization model built on an MLLM, designed for task-oriented localization and near-second temporal scanning on hour-long videos. To learn task-oriented localization, we construct ToMo-Pile, a training set that pairs temporal intervals with rationale traces explaining why the interval resolves the task. To make near-second scanning feasible on long videos, EviScan uses DualKV Prefill for memory-efficient streaming inference, and Progressive Localization to first narrow down candidate intervals and then refine boundaries using high-resolution evidence. We further introduce ToMo-Bench, a benchmark for task-oriented moment localization across diverse long-video scenarios. EviScan consistently outperforms specialized localizers and general-purpose MLLMs on both ToMo-Bench and standard temporal grounding benchmarks. When used as a plug-in localization module, it yields substantial gains on downstream long-video understanding benchmarks (MLVU and LongVideoBench), and these gains are larger than those brought by other moment-localization methods. We will release the model, dataset, benchmark, and code.
%U https://aclanthology.org/2026.emnlp-main.2144/
%P 48748-48769
Markdown (Informal)
[Task-Oriented Moment Localization for Long Video Understanding with Near-Second Coverage](https://aclanthology.org/2026.emnlp-main.2144/) (Liu et al., EMNLP 2026)
ACL
- Xiangrui Liu, Minghao Qin, Yan Shu, Zhengyang Liang, Yang Tian, Chen Jason Zhang, Bo Zhao, and Zheng Liu. 2026. Task-Oriented Moment Localization for Long Video Understanding with Near-Second Coverage. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 48748–48769, Budapest, Hungary. Association for Computational Linguistics.