Ask, Solve, Generate: Self-Evolving Unified Multimodal Understanding and Generation via Self-Consistency Rewards
作者: Ritesh Thawkar, Shravan Venkatraman, Omkar Thawakar, Abdelrahman Shaker, Fahad Khan, Hisham Cholakkal, Salman Khan, Rao Muhammad Anwer
分类: cs.CV
发布日期: 2026-06-25
💡 一句话要点
提出自演化框架以提升多模态理解与生成能力
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 多模态模型 自演化训练 视觉理解 图像生成 一致性信号
📋 核心要点
- 现有的统一多模态模型依赖人工标注等后训练监督,限制了其自主学习能力。
- 提出了一种自演化训练框架,通过内部角色协作,利用未标注图像提升理解与生成能力。
- 在多个理解指标上取得显著提升,BAGEL模型的生成性能从82%提升至85%。
📝 摘要(中文)
大多数统一的大型多模态模型(LMM)依赖于人工标注等后训练监督来支持视觉理解和图像生成。本文提出了一种自演化训练框架,利用未标注图像自主提升这两种能力。框架包含三个内部角色:生成视觉问题的Proposer、回答和评估问题的Solver,以及合成图像的Generator。训练过程中仅使用自生成的一致性信号,且引入了解决者令牌熵(STE)作为持续的难度信号,以稳定学习。实验结果表明,该方法在八个理解指标上均优于基线模型,并在BAGEL上实现了3.5%的绝对增益。代码和模型已公开发布。
🔬 方法详解
问题定义:本文旨在解决现有统一多模态模型依赖人工标注的问题,这限制了模型的自主学习和适应能力。
核心思路:提出的自演化训练框架通过内部角色的协作,利用未标注图像生成一致性信号,提升视觉理解和图像生成的能力。
技术框架:框架包含三个主要模块:Proposer生成视觉问题,Solver回答并评估问题,Generator合成图像。训练过程中,使用自生成的一致性信号进行学习。
关键创新:引入了解决者令牌熵(STE)作为持续的难度信号,确保在样本级一致性不可靠时仍能有效学习。此外,设计了多尺度内部评估方案,增强了生成评估的可靠性。
关键设计:框架保持了角色分解、奖励逻辑和训练调度的一致性,适用于不同的基础架构(如BLIP3o、BAGEL和VARGPT-v1.1),仅需各自的原生提示和生成接口。
📊 实验亮点
实验结果显示,提出的方法在八个理解指标上均优于基线模型,特别是在BAGEL模型中,生成性能从82%提升至85%,实现了3.5%的绝对增益,证明了框架的有效性和创新性。
🎯 应用场景
该研究的潜在应用领域包括智能问答系统、自动图像生成和多模态内容创作等。通过提升模型的自主学习能力,可以在缺乏人工标注的情况下,快速适应新任务,降低人工成本,推动多模态技术的广泛应用。
📄 摘要(原文)
Most unified large multimodal models (LMMs) that support both visual understanding and image generation still rely on curated post-training supervision, such as human annotations, preference labels, or external reward models. We ask whether a unified LMM can improve both abilities autonomously using only unlabeled images. We propose a self-evolving training framework with three internal roles: a Proposer that generates visual questions, a Solver that answers and evaluates them, and a Generator that synthesizes images. Training uses only self-derived consistency signals, without human annotations, preference labels, or task-trained external reward/judge models. To stabilize learning, we introduce Solver Token Entropy (STE), a continuous difficulty signal based on token-level prediction uncertainty that remains useful even when sample-level consistency becomes unreliable. For image generation, we design a multi-scale internal evaluation scheme that combines question-answer fidelity scoring with cycle-consistent captioning. This creates a solver-mediated coupling, where better visual understanding enables more reliable generation assessment and stronger internal training signals. The framework preserves the same role decomposition, reward logic, and training schedule across diffusion-based BLIP3o, rectified-flow BAGEL, and autoregressive VARGPT-v1.1 architectures, requiring only each backbone's native prompting and generation interface. Across eight understanding metrics, our method consistently improves over the corresponding base models. On BAGEL, it achieves a $+3.5\%$ absolute gain on MMMU and improves GenEval image generation performance from $82\%$ to $85\%$. Code and models are publicly released.