MaVEn: An Effective Multi-granularity Hybrid Visual Encoding Framework for Multimodal Large Language Model

TL;DR

MaVEn is a novel multi-granularity visual encoding framework that improves multimodal large language models' multi-image reasoning by combining semantic symbols with detailed features, enhancing understanding and efficiency.

Contribution

MaVEn introduces a dual encoding approach with a dynamic reduction mechanism, advancing multi-image reasoning capabilities in MLLMs beyond existing single-image focused models.

Findings

Significantly improves multi-image reasoning accuracy.

Enhances single-image understanding performance.

Increases processing efficiency for long visual sequences.

Abstract

This paper presents MaVEn, an innovative Multi-granularity Visual Encoding framework designed to enhance the capabilities of Multimodal Large Language Models (MLLMs) in multi-image reasoning. Current MLLMs primarily focus on single-image visual understanding, limiting their ability to interpret and integrate information across multiple images. MaVEn addresses this limitation by combining discrete visual symbol sequences, which abstract coarse-grained semantic concepts, with traditional continuous representation sequences that model fine-grained features. This dual approach bridges the semantic gap between visual and textual data, thereby improving the model's ability to process and interpret information from multiple images effectively. Additionally, we design a dynamic reduction mechanism by for long-sequence continuous features to enhance multi-image processing efficiency. Experimental results demonstrate that MaVEn significantly enhances MLLMs' understanding in complex multi-image scenarios, while also improving performance in single-image contexts.