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 while maintaining competitive image quality.
The following videos showcase the effects of our VSF method applied in Wan 2.1 14B various generative scenarios. Left and right videos illustrate comparisons between baseline methods and VSF-guided outputs. All videos are generated in 6 steps with lightx2v lora.
Positive Prompt: An astronaut hatching from an egg, on the surface of the moon, the darkness and depth of space realised in the background. High quality, ultrarealistic detail and breath-taking movie-like camera shot.
Negative Prompt: american flag, low quality, blurry, distorted, low resolution, unatural motion, unnatural lighting
Original
VSF
Positive Prompt: An astronaut hatching from an egg, on the surface of the moon, the darkness and depth of space realised in the background. High quality, ultrarealistic detail and breath-taking movie-like camera shot.
Negative Prompt: stars in the sky, low quality, blurry, distorted, low resolution, unatural motion, unnatural lighting
Original
VSF
Positive Prompt: A short-haired gray cat sitting alert on a windowsill, its cheeks unusually smooth beneath attentive eyes.
Negative Prompt: curtain, low quality, blurry, distorted, low resolution, unatural motion, unnatural lighting
Original
VSF
Positive Prompt: A mountain bike rides along a winding countryside trail under a cloudy night sky. The moon is clearly visible through gaps in the clouds, casting faint silver light over the uneven terrain. The bike's headlamp cuts through the darkness, illuminating the rocky path ahead as it speeds forward. The ambient sounds of distant winds and gravel crunching beneath tires accompany the scene.
Negative Prompt: tree, low quality, blurry, distorted, low resolution, unatural motion, unnatural lighting
Original
VSF
Positive Prompt: CG animation style, a small blue bird takes off from the ground, flapping its wings. The bird's feathers are delicate, with a unique pattern on its chest. The background shows a blue sky with white clouds under bright sunshine. The camera follows the bird upward, capturing its flight and the vastness of the sky from a close-up, low-angle perspective.
Negative Prompt: tree, low quality, blurry, distorted, low resolution, unatural motion, unnatural lighting
Original
VSF
Positive Prompt: A sharp stainless steel knife cutting through a ripe orange tomato, with the blade halfway in and tomato juice and seeds visible. Sliced tomato pieces are nearby, and the scene is well-lit with a soft kitchen background.
Negative Prompt: wooden board, low quality, blurry, distorted, low resolution, unatural motion, unnatural lighting
Original
VSF
@article{du_vsf:_2025,
title = {{VSF}: {Simple}, {Efficient}, and {Effective} {Negative} {Guidance} in {Few}-{Step} {Image} {Generation} {Models} {By} {Value} {Sign} {Flip}},
shorttitle = {{VSF}},
url = {https://rgdoi.net/10.13140/RG.2.2.18760.43520},
doi = {10.13140/RG.2.2.18760.43520},
language = {en},
urldate = {2025-07-27},
author = {Guo, Wenqi and Du, Shan},
year = {2025},
}