Script: Graph-Structured and Query-Conditioned Semantic Token Pruning for Multimodal Large Language Models
By: Zhongyu Yang , Dannong Xu , Wei Pang and more
Potential Business Impact:
Makes AI see pictures and videos faster.
The rapid growth of visual tokens in multimodal large language models (MLLMs) leads to excessive memory consumption and inference latency, especially when handling high-resolution images and videos. Token pruning is a technique used to mitigate this issue by removing redundancy, but existing methods often ignore relevance to the user query or suffer from the limitations of attention mechanisms, reducing their adaptability and effectiveness. To address these challenges, we propose Script, a plug-and-play pruning method that requires no retraining and generalizes across diverse MLLMs. Script comprises two modules: a graph-structured pruning module that removes visually redundant tokens, and a query-conditioned semantic pruning module that preserves query-relevant visual information. Together, they enhance performance on multimodal tasks. Experiments on fourteen benchmarks across image and video understanding tasks show that Script consistently achieves higher model efficiency and predictive accuracy compared to existing pruning methods. On LLaVA-NeXT-7B, it achieves up to 6.8x prefill speedup and 10x FLOP reduction, while retaining 96.88% of the original performance.
Similar Papers
GridPrune: From "Where to Look" to "What to Select" in Visual Token Pruning for MLLMs
CV and Pattern Recognition
Makes AI understand pictures faster by focusing on important parts.
What Kind of Visual Tokens Do We Need? Training-free Visual Token Pruning for Multi-modal Large Language Models from the Perspective of Graph
CV and Pattern Recognition
Makes AI understand pictures faster and better.
Language-Guided Temporal Token Pruning for Efficient VideoLLM Processing
CV and Pattern Recognition
Lets computers watch long videos faster.