Saved in:
Bibliographic Details
Main Authors: Reza, Sakib, Song, Xiyun, Yu, Heather, Lin, Zongfang, Moghaddam, Mohsen, Camps, Octavia
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.05491
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916676848582656
author Reza, Sakib
Song, Xiyun
Yu, Heather
Lin, Zongfang
Moghaddam, Mohsen
Camps, Octavia
author_facet Reza, Sakib
Song, Xiyun
Yu, Heather
Lin, Zongfang
Moghaddam, Mohsen
Camps, Octavia
contents Integrating vision models into large language models (LLMs) has sparked significant interest in creating vision-language foundation models, especially for video understanding. Recent methods often utilize memory banks to handle untrimmed videos for video-level understanding. However, they typically compress visual memory using similarity-based greedy approaches, which can overlook the contextual importance of individual tokens. To address this, we introduce an efficient LLM adapter designed for video-level understanding of untrimmed videos that prioritizes the contextual relevance of spatio-temporal tokens. Our framework leverages scorer networks to selectively compress the visual memory bank and filter spatial tokens based on relevance, using a differentiable Top-K operator for end-to-end training. Across three key video-level understanding tasks$\unicode{x2013}$ untrimmed video classification, video question answering, and video captioning$\unicode{x2013}$our method achieves competitive or superior results on four large-scale datasets while reducing computational overhead by up to 34%. The code will be available soon on GitHub.
format Preprint
id arxiv_https___arxiv_org_abs_2504_05491
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle REEF: Relevance-Aware and Efficient LLM Adapter for Video Understanding
Reza, Sakib
Song, Xiyun
Yu, Heather
Lin, Zongfang
Moghaddam, Mohsen
Camps, Octavia
Computer Vision and Pattern Recognition
Integrating vision models into large language models (LLMs) has sparked significant interest in creating vision-language foundation models, especially for video understanding. Recent methods often utilize memory banks to handle untrimmed videos for video-level understanding. However, they typically compress visual memory using similarity-based greedy approaches, which can overlook the contextual importance of individual tokens. To address this, we introduce an efficient LLM adapter designed for video-level understanding of untrimmed videos that prioritizes the contextual relevance of spatio-temporal tokens. Our framework leverages scorer networks to selectively compress the visual memory bank and filter spatial tokens based on relevance, using a differentiable Top-K operator for end-to-end training. Across three key video-level understanding tasks$\unicode{x2013}$ untrimmed video classification, video question answering, and video captioning$\unicode{x2013}$our method achieves competitive or superior results on four large-scale datasets while reducing computational overhead by up to 34%. The code will be available soon on GitHub.
title REEF: Relevance-Aware and Efficient LLM Adapter for Video Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.05491