通过自适应选择注意力层，生成更精细的负样本以提升扩散模型生成质量。
EMAG: Self-Rectifying Diffusion Sampling with Exponential Moving Average Guidance
- 推理时基于统计自适应选择注意力层，无需训练即可调整负样本难度。
- 相比分类器自由引导（CFG），人类偏好得分提升0.54，细节瑕疵显著减少。
- 可与APG、CADS等正交引导技术兼容，适合追求高保真图像生成的研究者。
在扩散和流匹配生成模型中，引导技术广泛用于提升样本质量和一致性。分类器自由引导（CFG）是当前主流方法，通过对比条件与无条件样本实现优化。近期工作尝试在推理阶段使用较弱模型生成负样本，如强/弱模型对、基于注意力的掩码、随机块丢弃或对自注意力能量景观的扰动。尽管这些策略提升了生成质量，但仍缺乏对负样本粒度或难度的可靠控制，且目标层选择常固定不变。本文提出指数移动平均引导（EMAG），一种无需训练的推理时机制，在扩散变换器中修改注意力，采用基于统计的自适应层选择规则。不同于以往方法，EMAG生成更具挑战性且语义一致的负样本（细粒度退化），暴露困难失败模式，使去噪器能修复细微缺陷，相较CFG提升人类偏好得分（HPS）0.54。进一步证明，EMAG可自然与先进正交引导技术（如APG和CADS）结合，进一步提高HPS。
原文摘要 · Abstract (English)
In diffusion and flow-matching generative models, guidance techniques are widely used to improve sample quality and consistency. Classifier-free guidance (CFG) is the de facto choice in modern systems and achieves this by contrasting conditional and unconditional samples. Recent work explores contrasting negative samples at inference using a weaker model, via strong/weak model pairs, attention-based masking, stochastic block dropping, or perturbations to the self-attention energy landscape. While these strategies refine the generation quality, they still lack a reliable control over the granularity or difficulty of the negative samples, and target-layer selection is often fixed. We propose Exponential Moving Average Guidance (EMAG), a training-free mechanism that modifies attention at inference time in diffusion transformers, with a statistics-based, adaptive layer-selection rule. Unlike prior methods, EMAG produces harder, semantically faithful negatives (fine-grained degradations), surfacing difficult failure modes, enabling the denoiser to refine subtle artifacts, boosting the quality and human preference score (HPS) by +0.54 over CFG. We further demonstrate that EMAG naturally composes with advanced orthogonal guidance techniques, such as APG and CADS, further improving HPS.
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。