重新审视统一多模态模型的视觉生成：面向高效纯图像预训练的掩码建模
Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training
把 UMM 的视觉生成部分拆成「纯图像掩码预训练 + 少量图文对指令对齐」两阶段，用 1000 H800 小时从零训出 3.6B 生成器，为视觉研究者提供了一个低成本复现统一多模态生成能力的训练配方。
原文依据 The first stage pre-trains the visual generative component exclusively using abundant unlabeled image-only data
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论统一多模态模型（UMM）常常受限于其视觉生成组件的预训练，而这一预训练通常依赖低效的范式以及稀缺的高质量图文配对数据。本文系统分析了 UMM 视觉生成的预训练配方，并将上述两个问题认定为主要瓶颈。为解决这些问题，我们提出了面向 UMM 的纯图像训练框架（IOMM），这是一个数据高效的两阶段训练框架。第一阶段仅使用大量无标注的纯图像数据来预训练视觉生成组件，从而在这一高成本阶段摆脱对配对数据的依赖。第二阶段使用无标注图像与少量精选图文对的混合数据对模型进行微调，从而提升指令对齐能力与生成质量。大量实验表明，IOMM 不仅提升了训练效率，还达到了最优（SOTA）性能。例如，我们的 IOMM-B（3.6B）模型从零开始训练仅需约 1050 个 H800 GPU 小时（其中绝大部分即 1000 小时用于高效的纯图像预训练阶段）。它在 GenEval 上达到 0.89、在 WISE 上达到 0.55，超过了 BAGEL-7B（0.82 与 0.55）和 BLIP3-o-4B（0.84 与 0.50）等强基线。代码见：https://github.com/LINs-lab/IOMM
英文原文摘要 1353 字符
Unified Multimodal Models (UMMs) are often constrained by the pre-training of their 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 UMM visual generation and identify these two issues as the major bottlenecks. To address them, we propose Image-Only Training for UMMs (IOMM), a data-efficient two-stage training framework. The first stage pre-trains the visual generative component exclusively using abundant unlabeled image-only data, thereby removing the dependency on paired data 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 about 1050 H800 GPU hours (with the vast majority, 1000 hours, dedicated to the efficient image-only pre-training stage). It achieves 0.89 on GenEval and 0.55 on WISE--surpassing strong baselines such as BAGEL-7B (0.82 and 0.55) and BLIP3-o-4B (0.84 and 0.50). Code is available at: https://github.com/LINs-lab/IOMM