VideoARM: Agentic Reasoning over Hierarchical Memory for Long-Form Video Understanding

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yin, Yufei, Meng, Qianke, Chen, Minghao, Ding, Jiajun, Shao, Zhenwei, Yu, Zhou
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915897866715136
author Yin, Yufei
Meng, Qianke
Chen, Minghao
Ding, Jiajun
Shao, Zhenwei
Yu, Zhou
author_facet Yin, Yufei
Meng, Qianke
Chen, Minghao
Ding, Jiajun
Shao, Zhenwei
Yu, Zhou
contents Long-form video understanding remains challenging due to the extended temporal structure and dense multimodal cues. Despite recent progress, many existing approaches still rely on hand-crafted reasoning pipelines or employ token-consuming video preprocessing to guide MLLMs in autonomous reasoning. To overcome these limitations, we introduce VideoARM, an Agentic Reasoning-over-hierarchical-Memory paradigm for long-form video understanding. Instead of static, exhaustive preprocessing, VideoARM performs adaptive, on-the-fly agentic reasoning and memory construction. Specifically, VideoARM performs an adaptive and continuous loop of observing, thinking, acting, and memorizing, where a controller autonomously invokes tools to interpret the video in a coarse-to-fine manner, thereby substantially reducing token consumption. In parallel, a hierarchical multimodal memory continuously captures and updates multi-level clues throughout the operation of the agent, providing precise contextual information to support the controller in decision-making. Experiments on prevalent benchmarks demonstrate that VideoARM outperforms the state-of-the-art method, DVD, while significantly reducing token consumption for long-form videos.
format Preprint
id arxiv_https___arxiv_org_abs_2512_12360
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle VideoARM: Agentic Reasoning over Hierarchical Memory for Long-Form Video Understanding
Yin, Yufei
Meng, Qianke
Chen, Minghao
Ding, Jiajun
Shao, Zhenwei
Yu, Zhou
Computer Vision and Pattern Recognition
Computation and Language
Long-form video understanding remains challenging due to the extended temporal structure and dense multimodal cues. Despite recent progress, many existing approaches still rely on hand-crafted reasoning pipelines or employ token-consuming video preprocessing to guide MLLMs in autonomous reasoning. To overcome these limitations, we introduce VideoARM, an Agentic Reasoning-over-hierarchical-Memory paradigm for long-form video understanding. Instead of static, exhaustive preprocessing, VideoARM performs adaptive, on-the-fly agentic reasoning and memory construction. Specifically, VideoARM performs an adaptive and continuous loop of observing, thinking, acting, and memorizing, where a controller autonomously invokes tools to interpret the video in a coarse-to-fine manner, thereby substantially reducing token consumption. In parallel, a hierarchical multimodal memory continuously captures and updates multi-level clues throughout the operation of the agent, providing precise contextual information to support the controller in decision-making. Experiments on prevalent benchmarks demonstrate that VideoARM outperforms the state-of-the-art method, DVD, while significantly reducing token consumption for long-form videos.
title VideoARM: Agentic Reasoning over Hierarchical Memory for Long-Form Video Understanding
topic Computer Vision and Pattern Recognition
Computation and Language
url https://arxiv.org/abs/2512.12360