@inproceedings{zhao-etal-2026-aemagent,
title = "{AEMA}gent: Cognition-Inspired Active Memory Formation for Very-Long Video Understanding",
author = "Zhao, Xian and
Zhang, Jiaming and
Chu, Haonan and
Bo, Weihao and
Chen, Xu and
Qiu, Delai and
Wang, Yining and
Zhang, Yi and
Liu, Shengping and
Yu, Jian and
Sang, Jitao",
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.915/",
pages = "19342--19355",
ISBN = "979-8-89176-462-0",
abstract = "Embodied AI and wearable assistants require agents to retrieve sparse evidence from continuous days-level videos. Existing memory-based methods typically follow a passive recording paradigm based on fixed-window captioning and periodic memory construction, which produces redundant memories and inefficient long-horizon reasoning. Inspired by human memory formation, we propose Active Explore Memory Agent (AEMAgent), which transforms passive recording into active memory formation by deciding when to remember, how to explore events, and when to reflect. These decisions are instantiated as a coherent pipeline: Surprise-guided Event Selection detects salient event changes for selective memory writing, Exploratory Episodic Memory Encoding enriches accepted events through self-questioning, Predictive-Coding-Guided Reflection updates knowledge from expectation-observation mismatches, and Semantic Memory Consolidation distills noisy episodic updates into compact long-term memory. Experiments on multiple benchmarks show that AEMAgent achieves the best overall performance among all compared methods while maintaining lower storage overhead than strong full-memory baselines, demonstrating the effectiveness of selective and efficient memory formation for scalable very-long video understanding."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhao-etal-2026-aemagent">
<titleInfo>
<title>AEMAgent: Cognition-Inspired Active Memory Formation for Very-Long Video Understanding</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xian</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiaming</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haonan</namePart>
<namePart type="family">Chu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weihao</namePart>
<namePart type="family">Bo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xu</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Delai</namePart>
<namePart type="family">Qiu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yining</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shengping</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jian</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jitao</namePart>
<namePart type="family">Sang</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>Embodied AI and wearable assistants require agents to retrieve sparse evidence from continuous days-level videos. Existing memory-based methods typically follow a passive recording paradigm based on fixed-window captioning and periodic memory construction, which produces redundant memories and inefficient long-horizon reasoning. Inspired by human memory formation, we propose Active Explore Memory Agent (AEMAgent), which transforms passive recording into active memory formation by deciding when to remember, how to explore events, and when to reflect. These decisions are instantiated as a coherent pipeline: Surprise-guided Event Selection detects salient event changes for selective memory writing, Exploratory Episodic Memory Encoding enriches accepted events through self-questioning, Predictive-Coding-Guided Reflection updates knowledge from expectation-observation mismatches, and Semantic Memory Consolidation distills noisy episodic updates into compact long-term memory. Experiments on multiple benchmarks show that AEMAgent achieves the best overall performance among all compared methods while maintaining lower storage overhead than strong full-memory baselines, demonstrating the effectiveness of selective and efficient memory formation for scalable very-long video understanding.</abstract>
<identifier type="citekey">zhao-etal-2026-aemagent</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.915/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>19342</start>
<end>19355</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T AEMAgent: Cognition-Inspired Active Memory Formation for Very-Long Video Understanding
%A Zhao, Xian
%A Zhang, Jiaming
%A Chu, Haonan
%A Bo, Weihao
%A Chen, Xu
%A Qiu, Delai
%A Wang, Yining
%A Zhang, Yi
%A Liu, Shengping
%A Yu, Jian
%A Sang, Jitao
%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 zhao-etal-2026-aemagent
%X Embodied AI and wearable assistants require agents to retrieve sparse evidence from continuous days-level videos. Existing memory-based methods typically follow a passive recording paradigm based on fixed-window captioning and periodic memory construction, which produces redundant memories and inefficient long-horizon reasoning. Inspired by human memory formation, we propose Active Explore Memory Agent (AEMAgent), which transforms passive recording into active memory formation by deciding when to remember, how to explore events, and when to reflect. These decisions are instantiated as a coherent pipeline: Surprise-guided Event Selection detects salient event changes for selective memory writing, Exploratory Episodic Memory Encoding enriches accepted events through self-questioning, Predictive-Coding-Guided Reflection updates knowledge from expectation-observation mismatches, and Semantic Memory Consolidation distills noisy episodic updates into compact long-term memory. Experiments on multiple benchmarks show that AEMAgent achieves the best overall performance among all compared methods while maintaining lower storage overhead than strong full-memory baselines, demonstrating the effectiveness of selective and efficient memory formation for scalable very-long video understanding.
%U https://aclanthology.org/2026.findings-emnlp.915/
%P 19342-19355
Markdown (Informal)
[AEMAgent: Cognition-Inspired Active Memory Formation for Very-Long Video Understanding](https://aclanthology.org/2026.findings-emnlp.915/) (Zhao et al., Findings 2026)
ACL
- Xian Zhao, Jiaming Zhang, Haonan Chu, Weihao Bo, Xu Chen, Delai Qiu, Yining Wang, Yi Zhang, Shengping Liu, Jian Yu, and Jitao Sang. 2026. AEMAgent: Cognition-Inspired Active Memory Formation for Very-Long Video Understanding. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 19342–19355, Budapest, Hungary. Association for Computational Linguistics.