CALM：文本到图像生成中全局不安全性的局限分析
原标题：Keep It CALM: Analyzing the Limits of Global Unsafety in Text-to-Image Generation
AI 摘要
该论文对文本到图像生成中「全局不安全信号」假设进行了受控几何分析，发现紧凑的不安全子空间无法覆盖异构的不安全语义，而扩大聚合范围又会扭曲安全邻近的良性提示，形成覆盖度与选择性的权衡。为此作者提出免训练的 CALM（反事实自适应局部调制）方法，利用匹配的不安全-良性锚点，将每个提示路由到活跃的不安全类别，仅对违规 token 表示做最小编辑并抑制正向对齐的不安全残差。实验表明 CALM 在显著提升不安全内容抑制的同时保留了良性生成效用，代码已开源。
正文节选
Keep It CALM: Analyzing the Limits of Global Unsafety in Text-to-Image Generation Abstract Training-free safeguards for text-to-image generation often rely on a reusable safety signal, such as an unsafe direction or global toxic subspace, applied broadly across prompts. We provide a controlled geometric analysis of this global-unsafety assumption and reveal a consistent coverage-selectivity trade-off: compact unsafe subspaces fail to cover heterogeneous unsafe semantics, whereas broader aggregat