arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2510.17136cs.LG
收藏

原位自引导:激发扩散模型的自校正能力

In-situ Autoguidance: Eliciting Self-Correction in Diffusion Models

Xinjiang University, Urumqi, Xinjiang, China(新疆大学);Ocean University of China, Qingdao, Shandong, China(中国海洋大学)

Enhao Gu, Haolin Hou

AI总结:

该研究提出原位自引导方法,无需辅助模型,通过随机前向传播动态生成次优预测,实现零成本的扩散模型自校正引导,为高性价比引导建立了新基准。

摘要:

生成高质量、多样化且与提示对齐的图像是图像生成扩散模型的核心目标。流行的无分类器引导(CFG)方法在提升质量和对齐性的同时,会降低多样性,导致这些效应存在固有纠缠。近期研究通过用单独训练的次优对应模型引导模型,成功解开了这些属性,但该方案需要辅助模型,带来了可观的开销。我们挑战这一前提,提出原位自引导(In-situ Autoguidance)方法,无需任何辅助组件即可从模型自身生成引导信号。该方法通过随机前向传播动态生成次优预测,将引导重新定义为推理时的自校正形式。我们证明这种零成本方法不仅可行,还为高性价比引导建立了强大的新基准,证实无需外部模型即可实现自引导的优势。

英文摘要:

The generation of high-quality, diverse, and prompt-aligned images is a central goal in image-generating diffusion models. The popular classifier-free guidance (CFG) approach improves quality and alignment at the cost of reduced variation, creating an inherent entanglement of these effects. Recent work has successfully disentangled these properties by guiding a model with a separately trained, inferior counterpart; however, this solution introduces the considerable overhead of requiring an auxiliary model. We challenge this prerequisite by introducing In-situ Autoguidance, a method that elicits guidance from the model itself without any auxiliary components. Our approach dynamically generates an inferior prediction on the fly using a stochastic forward pass, reframing guidance as a form of inference-time self-correction. We demonstrate that this zero-cost approach is not only viable but also establishes a powerful new baseline for cost-efficient guidance, proving that the benefits of self-guidance can be achieved without external models.

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。

↑