Semantic Alignment for Multimodal Large Language Models

TL;DR

This paper introduces SAM, a semantic alignment method for Multi-modal Large Language Models, which improves cross-image semantic coherence by bidirectional guidance during visual token extraction, significantly enhancing multi-image understanding tasks.

Contribution

The paper proposes a novel semantic alignment approach that incorporates bidirectional semantic guidance to better preserve linking information among images in MLLMs.

Findings

SAM outperforms state-of-the-art methods by +37% on group captioning and +22% on storytelling tasks.

The proposed dataset MmLINK contains 69K diverse multi-modal instruction samples.

Extensive experiments demonstrate the effectiveness of SAM in multi-image understanding.

Abstract

Research on Multi-modal Large Language Models (MLLMs) towards the multi-image cross-modal instruction has received increasing attention and made significant progress, particularly in scenarios involving closely resembling images (e.g., change captioning). Existing MLLMs typically follow a two-step process in their pipelines: first, extracting visual tokens independently for each input image, and then aligning these visual tokens from different images with the Large Language Model (LLM) in its textual feature space. However, the independent extraction of visual tokens for each image may result in different semantics being prioritized for different images in the first step, leading to a lack of preservation of linking information among images for subsequent LLM analysis. This issue becomes more serious in scenarios where significant variations exist among the images (e.g., visual storytelling). To address this challenge, we introduce Semantic Alignment for Multi-modal large language models (SAM). By involving the bidirectional semantic guidance between different images in the visual-token extraction process, SAM aims to enhance the preservation of linking information for coherent analysis and align the semantics of different images before feeding them into LLM. As the test bed, we propose a large-scale dataset named MmLINK consisting of 69K samples. Different from most existing datasets for MLLMs fine-tuning, our MmLINK dataset comprises multi-modal instructions with significantly diverse images. Extensive experiments on the group captioning task and the storytelling task prove the effectiveness of our SAM model, surpassing the state-of-the-art methods by a large margin (+37% for group captioning and +22% for storytelling on CIDEr score). Project page: https://mccartney01.github.io/SAM.