Linear Scaling Video VLMs for Long Video Understanding

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Eyzaguirre, Cristobal, Wu, Jiajun, Niebles, Juan Carlos
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917548321144832
author Eyzaguirre, Cristobal
Wu, Jiajun
Niebles, Juan Carlos
author_facet Eyzaguirre, Cristobal
Wu, Jiajun
Niebles, Juan Carlos
contents Video vision-language models (VLMs) are increasingly used in long-horizon and streaming settings, yet most video encoders still rely on spatiotemporal self-attention, causing compute and latency to grow quadratically with the number of frames. Existing efficiency methods improve scalability but often lose accuracy relative to full self-attention, for example through aggressive frame/token dropping or coarse attention approximations. We introduce StateKV, an inference-time method that adapts pretrained long-video VLMs to linear-time video prefill by carrying cross-frame context in a fixed-capacity, importance-based recurrent state, paired with a second full per-frame cache used for decoding. Across three long-video benchmarks and seven models spanning three families and multiple scales, StateKV remains close to full self-attention and consistently outperforms dominant sliding-window / recency-based streaming approximations, without fine-tuning or architectural changes. StateKV also reduces video-prefill cost measured FLOPs, enabling stronger accuracy at a fixed compute budget by running larger models. These results suggest a practical step toward scalable long-video understanding.
format Preprint
id arxiv_https___arxiv_org_abs_2605_31598
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Linear Scaling Video VLMs for Long Video Understanding
Eyzaguirre, Cristobal
Wu, Jiajun
Niebles, Juan Carlos
Computer Vision and Pattern Recognition
Video vision-language models (VLMs) are increasingly used in long-horizon and streaming settings, yet most video encoders still rely on spatiotemporal self-attention, causing compute and latency to grow quadratically with the number of frames. Existing efficiency methods improve scalability but often lose accuracy relative to full self-attention, for example through aggressive frame/token dropping or coarse attention approximations. We introduce StateKV, an inference-time method that adapts pretrained long-video VLMs to linear-time video prefill by carrying cross-frame context in a fixed-capacity, importance-based recurrent state, paired with a second full per-frame cache used for decoding. Across three long-video benchmarks and seven models spanning three families and multiple scales, StateKV remains close to full self-attention and consistently outperforms dominant sliding-window / recency-based streaming approximations, without fine-tuning or architectural changes. StateKV also reduces video-prefill cost measured FLOPs, enabling stronger accuracy at a fixed compute budget by running larger models. These results suggest a practical step toward scalable long-video understanding.
title Linear Scaling Video VLMs for Long Video Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.31598