Guided Path Sampling: Steering Diffusion Models Back on Track with Principled Path Guidance
This addresses a fundamental stability issue in diffusion model refinement for image generation, offering a new paradigm that is not incremental.
The paper tackles the problem that Classifier-Free Guidance (CFG) causes sampling paths to diverge from the data manifold in diffusion models, undermining iterative refinement, and proposes Guided Path Sampling (GPS) to stabilize paths through manifold-constrained interpolation, achieving improved perceptual quality and semantic alignment, such as an ImageReward of 0.79 and HPS v2 of 0.2995 on SDXL.
Iterative refinement methods based on a denoising-inversion cycle are powerful tools for enhancing the quality and control of diffusion models. However, their effectiveness is critically limited when combined with standard Classifier-Free Guidance (CFG). We identify a fundamental limitation: CFG's extrapolative nature systematically pushes the sampling path off the data manifold, causing the approximation error to diverge and undermining the refinement process. To address this, we propose Guided Path Sampling (GPS), a new paradigm for iterative refinement. GPS replaces unstable extrapolation with a principled, manifold-constrained interpolation, ensuring the sampling path remains on the data manifold. We theoretically prove that this correction transforms the error series from unbounded amplification to strictly bounded, guaranteeing stability. Furthermore, we devise an optimal scheduling strategy that dynamically adjusts guidance strength, aligning semantic injection with the model's natural coarse-to-fine generation process. Extensive experiments on modern backbones like SDXL and Hunyuan-DiT show that GPS outperforms existing methods in both perceptual quality and complex prompt adherence. For instance, GPS achieves a superior ImageReward of 0.79 and HPS v2 of 0.2995 on SDXL, while improving overall semantic alignment accuracy on GenEval to 57.45%. Our work establishes that path stability is a prerequisite for effective iterative refinement, and GPS provides a robust framework to achieve it.