Accelerating Multimodal Large Language Models by Searching Optimal Vision Token Reduction

TL;DR

This paper introduces a novel method to reduce vision tokens in Multimodal Large Language Models, significantly accelerating them while maintaining or improving performance through optimal token selection strategies.

Contribution

It proposes a greedy search algorithm and a parametric sigmoid function to optimize vision token reduction, enhancing efficiency and effectiveness of MLLMs.

Findings

Achieves over 2x speedup in models like LLaVA and InternVL2.

Outperforms existing token reduction methods under limited computational budgets.

Maintains performance levels despite significant token reduction.

Abstract

Prevailing Multimodal Large Language Models (MLLMs) encode the input image(s) as vision tokens and feed them into the language backbone, similar to how Large Language Models (LLMs) process the text tokens. However, the number of vision tokens increases quadratically as the image resolutions, leading to huge computational costs. In this paper, we consider improving MLLM's efficiency from two scenarios, (I) Reducing computational cost without degrading the performance. (II) Improving the performance with given budgets. We start with our main finding that the ranking of each vision token sorted by attention scores is similar in each layer except the first layer. Based on it, we assume that the number of essential top vision tokens does not increase along layers. Accordingly, for Scenario I, we propose a greedy search algorithm (G-Search) to find the least number of vision tokens to keep at each layer from the shallow to the deep. Interestingly, G-Search is able to reach the optimal reduction strategy based on our assumption. For Scenario II, based on the reduction strategy from G-Search, we design a parametric sigmoid function (P-Sigmoid) to guide the reduction at each layer of the MLLM, whose parameters are optimized by Bayesian Optimization. Extensive experiments demonstrate that our approach can significantly accelerate those popular MLLMs, e.g. LLaVA, and InternVL2 models, by more than $2 \times$ without performance drops. Our approach also far outperforms other token reduction methods when budgets are limited, achieving a better trade-off between efficiency and effectiveness.