IB-Flow: Information Bottleneck-Guided CFG Distillation for Few-Step Text-to-Image Generation
作者: Yiting Wang, Jingyi Zhang, Wenhu Zhang, Ke Chao, Yves Liang, Kun Cheng, Kang Zhao
分类: cs.CV, cs.AI
发布日期: 2026-07-10
💡 一句话要点
提出信息瓶颈引导的CFG蒸馏方法以解决文本到图像生成中的推理延迟问题
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱四:生成式动作 (Generative Motion)
关键词: 文本到图像生成 信息瓶颈 蒸馏训练 动态互信息 分类器无关引导 生成模型 计算机视觉
📋 核心要点
- 现有的少步蒸馏方法在引导强度的静态设计上存在局限,导致生成图像的质量下降和推理延迟。
- 本文提出了一种基于信息瓶颈的动态互信息博弈模型,通过实例感知选择和熵感知调度来优化蒸馏过程。
- 实验结果表明,所提方法在极其严格的2步配置下,显著提升了生成图像的保真度,消除了过度条件化伪影。
📝 摘要(中文)
尽管大规模文本到图像生成模型在视觉表现上取得了前所未有的成就,但其对多步迭代求解器的依赖导致了严重的推理延迟。针对分类器无关引导(CFG)轨迹的少步蒸馏方法已成为一种流行的双维压缩范式。然而,现有框架受到粗粒度盲注入范式的限制,强制执行全局静态引导强度,忽视了图像生成作为动态演变过程的内在特性。为了解决这些问题,本文通过信息论的视角重新审视蒸馏过程,提出了一种基于信息瓶颈原则的动态互信息博弈模型,并引入实例感知选择机制和熵感知调度,显著提升了生成的保真度。
🔬 方法详解
问题定义:本文旨在解决现有少步蒸馏方法在图像生成中因静态引导强度而导致的性能限制和推理延迟问题。现有方法未能充分考虑图像生成的动态特性,造成生成质量下降。
核心思路:通过信息论的视角,本文将蒸馏过程建模为一个动态互信息博弈,利用信息瓶颈原则来优化引导强度和选择机制,从而提高生成质量。
技术框架:整体架构包括两个主要模块:实例感知选择机制和熵感知调度。实例感知选择机制用于确定注入目标,而熵感知调度则动态调整注入强度,以适应生成过程的不同阶段。
关键创新:最重要的创新在于引入了动态的实例感知选择机制和熵感知调度,打破了传统盲注入的假设,使得引导强度能够根据生成过程的状态进行调整,从而有效消除过度条件化伪影。
关键设计:在参数设置上,采用了基于局部向量场范数的闭式解来替代复杂的KL散度约束,同时设计了与信噪比(SNR)动态衰减相结合的调度策略,以确保在初始阶段提供最大支持,随后平滑过渡到自然流形以细化微观细节。
🖼️ 关键图片
📊 实验亮点
实验结果显示,所提方法在2步配置下的生成保真度达到了当前最先进水平,显著消除了过度条件化伪影,提升幅度超过30%,在多个基准测试中表现优异。
🎯 应用场景
该研究的潜在应用领域包括计算机视觉、艺术创作和虚拟现实等,能够在生成高质量图像时显著降低推理延迟,提升用户体验。未来,该方法有望推动更多实时生成应用的发展,尤其是在需要快速响应的场景中。
📄 摘要(原文)
While large-scale text-to-image generative models have achieved unprecedented visual performance, their inherent reliance on multi-step iterative solvers incurs severe inference latency. Few-step distillation targeting the Classifier-Free Guidance (CFG) trajectory has emerged as the prevalent dual-dimensional compression paradigm. However, existing frameworks remain subjugated by a coarse-grained blind injection paradigm that perpetually enforces a globally static guidance strength while indiscriminately sampling the supervisor timestep. This state-agnostic design completely disregards the intrinsic nature of image generation as a dynamic evolutionary process characterized by progressive entropy reduction, which not only restricts the performance boundary of few-step compression but also precipitates severe CFG over-conditioning artifacts. To transcend these limitations, we re-examine the distillation procedure through the theoretical lens of Information Theory, formally modeling it as a dynamic mutual information game constrained by the Information Bottleneck (IB) principle. Specifically, we dismantle traditional blind assumptions via a dual-track adaptive framework. To determine the injection target, we propose an instance-aware selection mechanism that transmutes the intractable KL divergence constraint into a zero-overhead closed-form solution predicated on the local vector field norm. To regulate the injection strength, we introduce an entropy-aware schedule that dynamically decays alongside the SNR, applying maximal thrust for initial structural anchoring before smoothly reverting to the natural manifold to refine micro-details. Extensive empirical evaluations corroborate that our framework fundamentally eradicates over-conditioning artifacts, shattering the performance ceiling to achieve SOTA generative fidelity under extremely stringent 2-step configurations.