---
id: 20260927-T0-06
title: "扩散模型新正则化方法CARE：让条件信息参与表征约束"
title_en: "CARE: Condition-Aware Representation Regularization for Diffusion Models"
url: https://ai.daily.yangsir.net/daily/20260927-T0-06
issue_date: 2026-09-27
publish_date: 2026-09-26T04:00:00.000Z
category: research
source_name: "arXiv cs.LG (ML)"
source_url: https://arxiv.org/abs/2609.28561
---

# 扩散模型新正则化方法CARE：让条件信息参与表征约束

arXiv新论文提出CARE（Condition-Aware Representation Regularization），用于扩散模型的表征正则化。现有方法常忽略模型内置条件（如标签或文本），CARE将条件信息纳入正则化过程，以提升样本质量与训练效率。该方法对图像生成等依赖条件控制的扩散任务有潜在改进价值。

## English Version

**CARE: Condition-Aware Representation Regularization for Diffusion Models**

A new arXiv paper proposes CARE (Condition-Aware Representation Regularization) for diffusion models. Existing regularization methods often overlook built-in conditions such as labels or text; CARE incorporates condition information into the regularization process to improve sample quality and training efficiency. It has potential to enhance conditional diffusion tasks like image generation.

---

**来源**：[arXiv cs.LG (ML)](https://arxiv.org/abs/2609.28561)

**详情页**：https://ai.daily.yangsir.net/daily/20260927-T0-06

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*