Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models

Negative guidance -explicitly suppressing unwanted attributes -remains a fundamental challenge in diffusion models, particularly in few-step sampling regimes. While Classifier-Free Guidance (CFG) works well in standard settings, it fails under aggressive sampling step compression due to divergent predictions between positive and negative branches. We present Normalized Attention Guidance (NAG), an efficient, training-free mechanism that applies extrapolation in attention space with L1-based normalization and refinement. NAG restores effective negative guidance where CFG collapses while maintaining fidelity. Unlike existing approaches, NAG generalizes across architectures (UNet, DiT), sampling regimes (few-step, multi-step), and modalities (image, video), functioning as a universal plug-in with minimal computational overhead. Through extensive experimentation, we demonstrate consistent improvements in text alignment (CLIP Score), fidelity (FID, PFID), and human-perceived quality (ImageReward). Our ablation studies validate each design component, while user studies confirm significant preference for NAG-guided outputs. As a model-agnostic inference-time approach requiring no retraining, NAG provides effortless negative guidance for all modern diffusion frameworks -pseudocode in the Appendix! -Cow A tiger cow -Tiger Sketch of UFO over pyramid -Realistic, complex -Black and white -Blurry, low contrast Photo of aurora -Green -Male Portrait of AI researcher -Glasses Negative Prompting Figure 1: Negative prompting on 4-step Flux-Schnell [1]. CFG fails in few-step models. NAG restores effective negative prompting, enabling direct suppression of visual, semantic, and stylistic attributes, such as "glasses," "tiger," "realistic," or "blurry." This enhances controllability and expands creative freedom across composition, style, and quality-including prompt-based debiasing. 39th Conference on Neural Information Processing Systems (NeurIPS 2025).