video-SALMONN S: Memory-Enhanced Streaming Audio-Visual LLM

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sun, Guangzhi, Li, Yixuan, Wu, Xiaodong, Yang, Yudong, Li, Wei, Ma, Zejun, Zhang, Chao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917243144634368
author Sun, Guangzhi
Li, Yixuan
Wu, Xiaodong
Yang, Yudong
Li, Wei
Ma, Zejun
Zhang, Chao
author_facet Sun, Guangzhi
Li, Yixuan
Wu, Xiaodong
Yang, Yudong
Li, Wei
Ma, Zejun
Zhang, Chao
contents Long-duration streaming video understanding is fundamental for future AI agents, yet remains limited by ineffective long-term memory. We introduce video-SALMONN S, a memory-enhanced streaming audio-visual large language model that processes over 3-hour videos at 1 FPS and 360p resolution, outperforming strong non-streaming models under the same memory budget. In addition to token merging or downsampling, video-SALMONN S is the first to employ test-time training (TTT) as a streaming memory mechanism for video understanding. TTT continuously transforms short-term multimodal representations into long-term memory embedded in model parameters. To improve long-range dependency modeling and memory capacity, we propose (i) a TTT_MEM layer with an additional long-span prediction objective, (ii) a two-stage training scheme, and (iii) a modality-aware memory reader. We further introduce the Episodic Learning from Video Memory (ELViM) benchmark, simulating agent-like scenarios where models must learn from videos observed hours earlier. video-SALMONN S consistently outperforms both streaming and non-streaming baselines by 3-7% on long video benchmarks. Notably, video-SALMONN S achieves a 15% absolute accuracy improvement over strong non-streaming models on ELViM, demonstrating strong learning abilities from video memory.
format Preprint
id arxiv_https___arxiv_org_abs_2510_11129
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle video-SALMONN S: Memory-Enhanced Streaming Audio-Visual LLM
Sun, Guangzhi
Li, Yixuan
Wu, Xiaodong
Yang, Yudong
Li, Wei
Ma, Zejun
Zhang, Chao
Computer Vision and Pattern Recognition
Artificial Intelligence
Long-duration streaming video understanding is fundamental for future AI agents, yet remains limited by ineffective long-term memory. We introduce video-SALMONN S, a memory-enhanced streaming audio-visual large language model that processes over 3-hour videos at 1 FPS and 360p resolution, outperforming strong non-streaming models under the same memory budget. In addition to token merging or downsampling, video-SALMONN S is the first to employ test-time training (TTT) as a streaming memory mechanism for video understanding. TTT continuously transforms short-term multimodal representations into long-term memory embedded in model parameters. To improve long-range dependency modeling and memory capacity, we propose (i) a TTT_MEM layer with an additional long-span prediction objective, (ii) a two-stage training scheme, and (iii) a modality-aware memory reader. We further introduce the Episodic Learning from Video Memory (ELViM) benchmark, simulating agent-like scenarios where models must learn from videos observed hours earlier. video-SALMONN S consistently outperforms both streaming and non-streaming baselines by 3-7% on long video benchmarks. Notably, video-SALMONN S achieves a 15% absolute accuracy improvement over strong non-streaming models on ELViM, demonstrating strong learning abilities from video memory.
title video-SALMONN S: Memory-Enhanced Streaming Audio-Visual LLM
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2510.11129