MMG-Vid: Maximizing Marginal Gains at Segment-level and Token-level for Efficient Video LLMs
This addresses efficiency issues for real-world video understanding applications, representing an incremental improvement over existing token pruning methods.
The paper tackles the computational challenge of excessive visual tokens in Video Large Language Models by proposing MMG-Vid, a training-free visual token pruning framework that reduces 75% of visual tokens and accelerates the prefilling stage by 3.9x while maintaining over 99.5% of original performance.
Video Large Language Models (VLLMs) excel in video understanding, but their excessive visual tokens pose a significant computational challenge for real-world applications. Current methods aim to enhance inference efficiency by visual token pruning. However, they do not consider the dynamic characteristics and temporal dependencies of video frames, as they perceive video understanding as a multi-frame task. To address these challenges, we propose MMG-Vid, a novel training-free visual token pruning framework that removes redundancy by Maximizing Marginal Gains at both segment-level and token-level. Specifically, we first divide the video into segments based on frame similarity, and then dynamically allocate the token budget for each segment to maximize the marginal gain of each segment. Subsequently, we propose a temporal-guided DPC algorithm that jointly models inter-frame uniqueness and intra-frame diversity, thereby maximizing the marginal gain of each token. By combining both stages, MMG-Vid can maximize the utilization of the limited token budget, significantly improving efficiency while maintaining strong performance. Extensive experiments demonstrate that MMG-Vid can maintain over 99.5% of the original performance, while effectively reducing 75% visual tokens and accelerating the prefilling stage by 3.9x on LLaVA-OneVision-7B. Code will be released soon.