Modality-Specialized Synergizers for Interleaved Vision-Language Generalists

TL;DR

This paper introduces modality-specialized adaptation layers for vision-language models, significantly improving their ability to generate interleaved text and images by leveraging modality-specific features and a new instruction tuning dataset.

Contribution

The paper proposes MOSS, a novel modality-specialized synergizer with adaptation layers for VLGs, and LEAFINSTRUCT, a large interleaved instruction tuning dataset, enhancing interleaved generation capabilities.

Findings

Achieves state-of-the-art performance on complex interleaved tasks.

Demonstrates strong generalizability across different VLG architectures.

Significantly surpasses baseline models in interleaved text-image generation.

Abstract

Recent advancements in Vision-Language Models (VLMs) have led to the emergence of Vision-Language Generalists (VLGs) capable of understanding and generating both text and images. However, seamlessly generating an arbitrary sequence of text and images remains a challenging task for the current VLGs. One primary limitation lies in applying a unified architecture and the same set of parameters to simultaneously model discrete text tokens and continuous image features. Recent works attempt to tackle this fundamental problem by introducing modality-aware expert models. However, they employ identical architectures to process both text and images, disregarding the intrinsic inductive biases in these two modalities. In this work, we introduce MODALITY-SPECIALIZED SYNERGIZERS (MOSS), a novel design that efficiently optimizes existing unified architectures of VLGs with modality-specialized adaptation layers, i.e., a Convolutional LoRA for modeling the local priors of image patches and a Linear LoRA for processing sequential text. This design enables more effective modeling of modality-specific features while maintaining the strong cross-modal integration gained from pretraining. In addition, to improve the instruction-following capability on interleaved text-and-image generation, we introduce LEAFINSTRUCT, the first open-sourced interleaved instruction tuning dataset comprising 184,982 high-quality instances on more than 10 diverse domains. Extensive experiments show that VLGs integrated with M OSS achieve state-of-the-art performance, significantly surpassing baseline VLGs in complex interleaved generation tasks. Furthermore, our method exhibits strong generalizability on different VLGs.