Scaling the Long Video Understanding of Multimodal Large Language Models via Visual Memory Mechanism

TL;DR

This paper introduces FlexMem, a training-free visual memory mechanism that enables multimodal large language models to understand arbitrarily long videos by mimicking human memory recall, significantly improving efficiency and performance.

Contribution

The paper proposes FlexMem, a novel visual memory mechanism that allows MLLMs to process infinite-length videos without training, enhancing understanding and performance on long video tasks.

Findings

FlexMem enables processing over 1,000 frames on a single GPU.

It improves performance of existing MLLMs on long video benchmarks.

FlexMem achieves comparable or better results than SOTA models like GPT-4o and Gemini-1.5 Pro.

Abstract

Long video understanding is a key challenge that plagues the advancement of \emph{Multimodal Large language Models} (MLLMs). In this paper, we study this problem from the perspective of visual memory mechanism, and proposed a novel and training-free approach, termed \emph{Flexible Memory} (\textbf{FlexMem}). In principle, FlexMem aims to mimic human behavior of video watching, \emph{i.e.}, continually watching video content and recalling the most relevant memory fragments to answer the question. In this way, FlexMem can help MLLMs achieve video understanding of infinite lengths, unlike previous methods that process all video information at once and have input upper-limit. Concretely, FlexMem first consider the visual KV caches as the memory sources, and realize the effective memory transfer and writing via a dual-pathway compression design. Afterwards, FlexMem also explores different memory reading strategies for the diverse video understanding tasks, including the popular streaming one. To validate FlexMem, we apply it to two popular video-MLLMs, and conduct extensive experiments on five long video and one streaming video task. The experimental results show that on \textbf{a single 3090 GPU}, our FlexMem can achieve obvious improvements than existing efficient video understanding methods and process more than \textbf{1k frames}, which also helps the base MLLMs achieve comparable or even better performance than SOTA MLLMs on some benchmarks, \emph{e.g.} , GPT-4o and Gemini-1.5 Pro.