提出AdaMaG方法，让生成模型在强引导下仍保持概率守恒，减少幻觉。
Probability-Conserving Flow Guidance
- 基于连续性方程分解引导项，分离出破坏概率守恒的发散项
- 设计时变调度与分数平行衰减，有效控制发散项和得分项
- 无需额外计算成本，可直接提升图像真实感与可控去饱和
扩散模型与流模型主导视觉生成，引导机制能对齐用户输入并提升感知质量。然而，无分类器引导（CFG）等方法基于速度/得分的启发式线性组合，忽略生成流形几何结构，导致概率不守恒，在强引导下使样本偏离学习到的流形。本文通过连续性方程分析引导机制，发现其效应可分解为发散项与与得分平行项，且该分解在参数化下保持不变。我们证明发散项在采样趋近数据流形时会结构性发散，因此提出时间依赖调度与得分平行衰减策略。由此得到的即插即用规则——自适应流形引导（AdaMaG），可在不增加推理成本的前提下同时约束两项。实验表明，多数用于缓解饱和或提升生成质量的经验法则，均直接对应上述两项。在多个图像生成基准上，AdaMaG显著提升真实感、减少幻觉，并在高引导强度下实现可控去饱和。
原文摘要 · Abstract (English)
Diffusion and flow-based generative models dominate visual synthesis, with guidance aligning samples to user input and improving perceptual quality. However, Classifier-Free Guidance (CFG) and extrapolation-based methods are heuristic linear combinations of velocities/scores that ignore the generative manifold geometry, breaking probability conservation and driving samples off the learned manifold under strong guidance. We analyse guidance through the continuity equation and show its effect decomposes into a divergence term and a score-parallel term defined invariantly across parameterisations. We prove the divergence term blows up structurally as sampling approaches the data manifold, motivating a time-dependent schedule alongside score-parallel attenuation. The resulting plug-and-play rule, Adaptive Manifold Guidance (AdaMaG), bounds both terms at no additional inference cost. Finally, we show that most empirical heuristics for reducing saturation or improving generation quality correspond directly to the two terms in our decomposition. Across image generation benchmarks, AdaMaG improves realism, reduces hallucinations, and induces controlled desaturation in high-guidance regimes.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。