Vision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations

TL;DR

This paper introduces a unified multimodal framework that aligns visual understanding and generation through a shared text-aligned semantic representation, enabling efficient cross-modal tasks with improved performance.

Contribution

The paper proposes the Text-Aligned Tokenizer (TA-Tok) and a unified multimodal LLM, Tar, which integrate vision and text into a shared space without modality-specific components, enhancing efficiency and performance.

Findings

Tar matches or surpasses existing multimodal LLMs in benchmarks.

The framework achieves faster convergence and greater training efficiency.

Experiments demonstrate improved visual understanding and generation capabilities.

Abstract

This paper presents a multimodal framework that attempts to unify visual understanding and generation within a shared discrete semantic representation. At its core is the Text-Aligned Tokenizer (TA-Tok), which converts images into discrete tokens using a text-aligned codebook projected from a large language model's (LLM) vocabulary. By integrating vision and text into a unified space with an expanded vocabulary, our multimodal LLM, Tar, enables cross-modal input and output through a shared interface, without the need for modality-specific designs. Additionally, we propose scale-adaptive encoding and decoding to balance efficiency and visual detail, along with a generative de-tokenizer to produce high-fidelity visual outputs. To address diverse decoding needs, we utilize two complementary de-tokenizers: a fast autoregressive model and a diffusion-based model. To enhance modality fusion, we investigate advanced pre-training tasks, demonstrating improvements in both visual understanding and generation. Experiments across benchmarks show that Tar matches or surpasses existing multimodal LLM methods, achieving faster convergence and greater training efficiency. Code, models, and data are available at https://tar.csuhan.com