JEDI：Jensen-Shannon散度在解耦扩散模型中的力量
JEDI: The Force of Jensen-Shannon Divergence in Disentangling Diffusion Models
AI总结：
提出测试时自适应方法JEDI，利用基于Jensen-Shannon散度的目标和对抗优化减少扩散模型注意力语义纠缠，无需重训即可提升主体分离与提示对齐，并给出无CLIP解耦评分。
摘要：
我们提出JEDI，这是一种测试时自适应方法，无需重新训练或外部监督即可增强扩散模型中的主体分离与组合对齐。JEDI通过使用一种新颖的基于Jensen-Shannon散度的目标函数来最小化注意力图中的语义纠缠。为提高效率，我们利用对抗优化，减少所需的更新步数。JEDI与模型无关，适用于Stable Diffusion 1.5和3.5等架构，在复杂场景中持续改善提示对齐和解耦。此外，JEDI提供了一个源自内部注意力分布的轻量级、无CLIP的解耦分数，为测试时条件下的组合对齐提供了原则性基准。代码和结果可在https://ericbill21.github.io/JEDI/获取。
英文摘要：
We introduce JEDI, a test-time adaptation method that enhances subject separation and compositional alignment in diffusion models without requiring retraining or external supervision. JEDI operates by minimizing semantic entanglement in attention maps using a novel Jensen-Shannon divergence based objective. To improve efficiency, we leverage adversarial optimization, reducing the number of updating steps required. JEDI is model-agnostic and applicable to architectures such as Stable Diffusion 1.5 and 3.5, consistently improving prompt alignment and disentanglement in complex scenes. Additionally, JEDI provides a lightweight, CLIP-free disentanglement score derived from internal attention distributions, offering a principled benchmark for compositional alignment under test-time conditions. Code and results are available at https://ericbill21.github.io/JEDI/.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
