OpenUni: A Simple Baseline for Unified Multimodal Understanding and Generation

TL;DR

OpenUni introduces a lightweight, open-source baseline that unifies multimodal understanding and generation, achieving high-quality image synthesis and strong benchmark performance with minimal complexity.

Contribution

It presents a simple, efficient architecture bridging multimodal LLMs and diffusion models, with released code, weights, and datasets to foster open research.

Findings

High-quality, instruction-aligned image generation.

Exceptional benchmark performance with few activated parameters.

Open-source release of models, code, and datasets.

Abstract

In this report, we present OpenUni, a simple, lightweight, and fully open-source baseline for unifying multimodal understanding and generation. Inspired by prevailing practices in unified model learning, we adopt an efficient training strategy that minimizes the training complexity and overhead by bridging the off-the-shelf multimodal large language models (LLMs) and diffusion models through a set of learnable queries and a light-weight transformer-based connector. With a minimalist choice of architecture, we demonstrate that OpenUni can: 1) generate high-quality and instruction-aligned images, and 2) achieve exceptional performance on standard benchmarks such as GenEval, DPG- Bench, and WISE, with only 1.1B and 3.1B activated parameters. To support open research and community advancement, we release all model weights, training code, and our curated training datasets (including 23M image-text pairs) at https://github.com/wusize/OpenUni.