Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.16139 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908893131571200 |
|---|---|
| author | Sun, Peng Xie, Jun Lin, Tao |
| author_facet | Sun, Peng Xie, Jun Lin, Tao |
| contents | Unified Multimodal Models (UMMs) are often constrained by the pre-training of their $\textbf{visual generation components}$, which typically relies on inefficient paradigms and scarce, high-quality text-image paired data. In this paper, we systematically analyze pre-training recipes for $\textbf{UMM visual generation}$ and identify these two issues as the major bottlenecks.
To address them, we propose $\textbf{Image-Only Training for UMMs (IOMM)}$, a data-efficient two-stage training framework.
The first stage pre-trains the visual generative component $\textbf{exclusively}$ using abundant unlabeled image-only data, thereby removing the dependency on paired data $\textbf{for this costly phase}$. The second stage fine-tunes the model using a mixture of unlabeled images and a small curated set of text-image pairs, leading to improved instruction alignment and generative quality.
Extensive experiments show that IOMM not only improves training efficiency but also achieves state-of-the-art (SOTA) performance.
For example, our IOMM-B (3.6B) model was trained from scratch using only $\sim \textbf{1050}$ H800 GPU hours (with the vast majority, $\textbf{1000}$ hours, dedicated to the efficient $\textbf{image-only pre-training stage}$). It achieves $\textbf{0.89}$ on GenEval and $\textbf{0.55}$ on WISE--surpassing strong baselines such as BAGEL-7B (0.82 & 0.55) and BLIP3-o-4B (0.84 & 0.50).
Code is available $\href{https://github.com/LINs-lab/IOMM}{https://github.com/LINs-lab/IOMM}$. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_16139 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training Sun, Peng Xie, Jun Lin, Tao Computer Vision and Pattern Recognition Unified Multimodal Models (UMMs) are often constrained by the pre-training of their $\textbf{visual generation components}$, which typically relies on inefficient paradigms and scarce, high-quality text-image paired data. In this paper, we systematically analyze pre-training recipes for $\textbf{UMM visual generation}$ and identify these two issues as the major bottlenecks. To address them, we propose $\textbf{Image-Only Training for UMMs (IOMM)}$, a data-efficient two-stage training framework. The first stage pre-trains the visual generative component $\textbf{exclusively}$ using abundant unlabeled image-only data, thereby removing the dependency on paired data $\textbf{for this costly phase}$. The second stage fine-tunes the model using a mixture of unlabeled images and a small curated set of text-image pairs, leading to improved instruction alignment and generative quality. Extensive experiments show that IOMM not only improves training efficiency but also achieves state-of-the-art (SOTA) performance. For example, our IOMM-B (3.6B) model was trained from scratch using only $\sim \textbf{1050}$ H800 GPU hours (with the vast majority, $\textbf{1000}$ hours, dedicated to the efficient $\textbf{image-only pre-training stage}$). It achieves $\textbf{0.89}$ on GenEval and $\textbf{0.55}$ on WISE--surpassing strong baselines such as BAGEL-7B (0.82 & 0.55) and BLIP3-o-4B (0.84 & 0.50). Code is available $\href{https://github.com/LINs-lab/IOMM}{https://github.com/LINs-lab/IOMM}$. |
| title | Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.16139 |