Less is More: A Simple yet Effective Token Reduction Method for Efficient Multi-modal LLMs

TL;DR

This paper introduces TRIM, a token reduction method for multimodal large language models that significantly decreases resource consumption while maintaining performance, inspired by human attention patterns in VQA tasks.

Contribution

The paper proposes TRIM, a novel token reduction technique using CLIP metrics, improving efficiency of MLLMs without performance loss.

Findings

Significant reduction in computational overhead across 12 datasets.

Maintains performance levels comparable to full models.

Promotes sustainability and accessibility of MLLMs.

Abstract

The rapid advancement of Multimodal Large Language Models (MLLMs) has led to remarkable performances across various domains. However, this progress is accompanied by a substantial surge in the resource consumption of these models. We address this pressing issue by introducing a new approach, Token Reduction using CLIP Metric (TRIM), aimed at improving the efficiency of MLLMs without sacrificing their performance. Inspired by human attention patterns in Visual Question Answering (VQA) tasks, TRIM presents a fresh perspective on the selection and reduction of image tokens. The TRIM method has been extensively tested across 12 datasets, and the results demonstrate a significant reduction in computational overhead while maintaining a consistent level of performance. This research marks a critical stride in efficient MLLM development, promoting greater accessibility and sustainability of high-performing models.