@inproceedings{asgarov-etal-2026-nest,
title = "{NEST}: Narrative Event Structures in Time for Long Video Understanding",
author = "Asgarov, Ali and
Narasimhan, Kaushik and
Sarker, Najibul Haque and
Alomari, Hani and
Tang, Chia-Wei and
Sivakumar, Anushka and
Abdul Hakim, Zaber Ibn and
Mallampati, Shaurya and
Thomas, Chris",
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.2512/",
pages = "57094--57129",
ISBN = "979-8-89176-459-0",
abstract = "Recent progress in vision-language models has enabled processing of increasingly long video sequences, but handling extended token streams does not translate to understanding complex narrative structure in long videos. Existing long-video benchmarks focus on needle-in-a-haystack retrieval rather than evaluating how low-level actions form events, interact across time, and drive narratives, for example whether a model can connect an early job loss to a later relationship breakup despite intervening scenes or flashbacks. We introduce NEST (Narrative Event Structures in Time for Long Video Understanding), a dataset of 1,005 full-length movies (avg. 98 minutes), each annotated with about 103 multimodal narrative events grounded in visual content, dialogue, and audio. NEST links these events through relations that reflect narrative structure, including explicit temporal ordering, hierarchical composition, and long-range dependencies. We introduce baselines for event trigger detection (ETD), event localization (EL), event argument extraction (EAE), and event relation extraction (ERE). The benchmark is highly challenging for grounded event discovery, with the best performance across evaluated zero-shot models on main evaluation subsets remaining below 8{\%} for ETD, 6{\%} for EL, and 11{\%} for EAE. In contrast, ERE is more tractable once events are given, reaching 34.23{\%} F1 zero-shot and 49.20{\%} F1 after fine-tuning."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="asgarov-etal-2026-nest">
<titleInfo>
<title>NEST: Narrative Event Structures in Time for Long Video Understanding</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ali</namePart>
<namePart type="family">Asgarov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaushik</namePart>
<namePart type="family">Narasimhan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Najibul</namePart>
<namePart type="given">Haque</namePart>
<namePart type="family">Sarker</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hani</namePart>
<namePart type="family">Alomari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chia-Wei</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anushka</namePart>
<namePart type="family">Sivakumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zaber</namePart>
<namePart type="given">Ibn</namePart>
<namePart type="family">Abdul Hakim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shaurya</namePart>
<namePart type="family">Mallampati</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chris</namePart>
<namePart type="family">Thomas</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>Recent progress in vision-language models has enabled processing of increasingly long video sequences, but handling extended token streams does not translate to understanding complex narrative structure in long videos. Existing long-video benchmarks focus on needle-in-a-haystack retrieval rather than evaluating how low-level actions form events, interact across time, and drive narratives, for example whether a model can connect an early job loss to a later relationship breakup despite intervening scenes or flashbacks. We introduce NEST (Narrative Event Structures in Time for Long Video Understanding), a dataset of 1,005 full-length movies (avg. 98 minutes), each annotated with about 103 multimodal narrative events grounded in visual content, dialogue, and audio. NEST links these events through relations that reflect narrative structure, including explicit temporal ordering, hierarchical composition, and long-range dependencies. We introduce baselines for event trigger detection (ETD), event localization (EL), event argument extraction (EAE), and event relation extraction (ERE). The benchmark is highly challenging for grounded event discovery, with the best performance across evaluated zero-shot models on main evaluation subsets remaining below 8% for ETD, 6% for EL, and 11% for EAE. In contrast, ERE is more tractable once events are given, reaching 34.23% F1 zero-shot and 49.20% F1 after fine-tuning.</abstract>
<identifier type="citekey">asgarov-etal-2026-nest</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2512/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>57094</start>
<end>57129</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T NEST: Narrative Event Structures in Time for Long Video Understanding
%A Asgarov, Ali
%A Narasimhan, Kaushik
%A Sarker, Najibul Haque
%A Alomari, Hani
%A Tang, Chia-Wei
%A Sivakumar, Anushka
%A Abdul Hakim, Zaber Ibn
%A Mallampati, Shaurya
%A Thomas, Chris
%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 asgarov-etal-2026-nest
%X Recent progress in vision-language models has enabled processing of increasingly long video sequences, but handling extended token streams does not translate to understanding complex narrative structure in long videos. Existing long-video benchmarks focus on needle-in-a-haystack retrieval rather than evaluating how low-level actions form events, interact across time, and drive narratives, for example whether a model can connect an early job loss to a later relationship breakup despite intervening scenes or flashbacks. We introduce NEST (Narrative Event Structures in Time for Long Video Understanding), a dataset of 1,005 full-length movies (avg. 98 minutes), each annotated with about 103 multimodal narrative events grounded in visual content, dialogue, and audio. NEST links these events through relations that reflect narrative structure, including explicit temporal ordering, hierarchical composition, and long-range dependencies. We introduce baselines for event trigger detection (ETD), event localization (EL), event argument extraction (EAE), and event relation extraction (ERE). The benchmark is highly challenging for grounded event discovery, with the best performance across evaluated zero-shot models on main evaluation subsets remaining below 8% for ETD, 6% for EL, and 11% for EAE. In contrast, ERE is more tractable once events are given, reaching 34.23% F1 zero-shot and 49.20% F1 after fine-tuning.
%U https://aclanthology.org/2026.emnlp-main.2512/
%P 57094-57129
Markdown (Informal)
[NEST: Narrative Event Structures in Time for Long Video Understanding](https://aclanthology.org/2026.emnlp-main.2512/) (Asgarov et al., EMNLP 2026)
ACL
- Ali Asgarov, Kaushik Narasimhan, Najibul Haque Sarker, Hani Alomari, Chia-Wei Tang, Anushka Sivakumar, Zaber Ibn Abdul Hakim, Shaurya Mallampati, and Chris Thomas. 2026. NEST: Narrative Event Structures in Time for Long Video Understanding. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 57094–57129, Budapest, Hungary. Association for Computational Linguistics.