通过理解来生成：面向统一多模态模型的理解驱动内在奖励
Learning to Generate via Understanding: Understanding-Driven Intrinsic Rewarding for Unified Multimodal Models
把统一多模态模型的理解分支当作奖励模型，用 token 级文本-图像对齐内在奖励做自监督强化学习来提升 T2I 生成，为视觉研究者提供了一种无需外部偏好数据即可微调生成质量的思路。
原文依据 We propose a token-level intrinsic text-image alignment reward mechanism, GvU
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期，统一多模态模型（UMM）在整合视觉理解与生成方面取得了显著进展，在复杂文本到图像（T2I）任务上展现出强大潜力。尽管在理论上颇具前景，但仍存在持续的能力差距：UMM 通常表现出更优的视觉理解能力，而生成能力相对较弱。这一差异在很大程度上源于理解过程与生成过程之间的内在解耦。UMM 虽能准确解读细粒度视觉细节，却常常难以根据复杂文本提示生成语义连贯的图像。为应对这一挑战，我们探索利用 UMM 的内部理解能力来提升生成质量。我们提出一种 token 级的内在文本-图像对齐奖励机制 GvU，使 UMM 能够同时充当教师与学生：它利用理解分支评估自身输出，从而相应地引导生成。在此基础上，我们设计了一个自监督强化学习框架，使 UMM 能够借助基于理解的内在奖励信号迭代提升生成质量，而无需依赖外部监督。实验结果表明，我们的方法大幅提升了 UMM 的生成能力，这反过来又增强了其细粒度视觉理解能力，缩小了 UMM 视觉理解与生成之间的能力差距。项目主页为 https://matrix0721.github.io/gvu.github.io/。
英文原文摘要 1519 字符
Recently, unified multimodal models (UMMs) have made remarkable progress in integrating visual understanding and generation, demonstrating strong potential for complex text-to-image (T2I) tasks. Despite their theoretical promise, a persistent capability gap exists: UMMs typically exhibit superior visual understanding but comparatively weaker generative capabilities. This discrepancy arises largely from the intrinsic decoupling between the understanding and generation processes. While a UMM can accurately interpret fine-grained visual details, it often struggles to produce semantically coherent images from complex textual prompts. To address this challenge, we explore UMMs' internal understanding capability to enhance generation quality. We propose a token-level intrinsic text-image alignment reward mechanism, GvU, enabling the UMM to act simultaneously as teacher and student: it evaluates its own outputs using the understanding branch to guide the generations accordingly. Building upon this, we design a self-supervised reinforcement learning framework, allowing UMMs to iteratively improve their generation quality through understanding-based intrinsic reward signals--without reliance on external supervision. Experimental results show that our method substantially boosts UMMs' generation, which in turn strengthens their fine-grained visual understanding, narrowing the capability gap between UMMs' visual understanding and generation. The project page is https://matrix0721.github.io/gvu.github.io/.