Towards a Golden Classifier-Free Guidance Path via Foresight Fixed Point Iterations

Tong Wu (Beijing Institute for General Artificial Intelligence) · Yang Xiang (Peng Cheng Laboratory) · Kaibo Wang (The Hong Kong University of Science and Technology) · Jianda Mao (Hong Kong University of Science and Technology)

adaptive designclassifier-free guidance computational efficiencyconditional guidancedesign choicesearly diffusion stagesfixed point iterationsforesight guidanceimage quality latent representationslonger-interval subproblemsmodel architectures.single-step short-interval iterationtext-to-image diffusion modelstheoretical interpretations

Classifier-Free Guidance (CFG) is an essential component of text-to-image diffusion models, and understanding and advancing its operational mechanisms remains a central focus of research. Existing approaches stem from divergent theoretical interpretations, thereby limiting the design space and obscuring key design choices. To address this, we propose a unified perspective that reframes conditional guidance as fixed point iterations, seeking to identify a golden path where latents produce consistent outputs under both conditional and unconditional generation. We demonstrate that CFG and its variants constitute a special case of single-step short-interval iteration, which is theoretically proven to exhibit inefficiency. To this end, we introduce Foresight Guidance (FSG), which prioritizes solving longer-interval subproblems in early diffusion stages with increased iterations. Extensive experiments across diverse datasets and model architectures validate the superiority of FSG over state-of-the-art methods in both image quality and computational efficiency. Our work offers novel perspectives for conditional guidance and unlocks the potential of adaptive design.