VSF: Simple, Efficient, and Effective Negative Guidance in Few-Step Image Generation Models By Value Sign Flip

TL;DR

The paper presents VSF, a simple and efficient method that improves negative prompt guidance in few-step diffusion models by flipping attention signs, leading to better adherence and quality in image and video generation.

Contribution

Introducing Value Sign Flip (VSF), a novel technique that enhances negative guidance in diffusion models with minimal computational overhead and broad compatibility.

Findings

VSF outperforms prior negative guidance methods in few-step models.

VSF improves negative prompt adherence in static and video generation.

VSF maintains competitive image quality while enhancing guidance effectiveness.

Abstract

We introduce Value Sign Flip (VSF), a simple and efficient method for incorporating negative prompt guidance in few-step diffusion and flow-matching image generation models. Unlike existing approaches such as classifier-free guidance (CFG), NASA, and NAG, VSF dynamically suppresses undesired content by flipping the sign of attention values from negative prompts. Our method requires only small computational overhead and integrates effectively with MMDiT-style architectures such as Stable Diffusion 3.5 Turbo, as well as cross-attention-based models like Wan. We validate VSF on challenging datasets with complex prompt pairs and demonstrate superior performance in both static image and video generation tasks. Experimental results show that VSF significantly improves negative prompt adherence compared to prior methods in few-step models, and even CFG in non-few-step models, while maintaining competitive image quality. Code and ComfyUI node are available in https://github.com/weathon/VSF/tree/main.