Rectified-CFG++ for Flow Based Models
This addresses a key issue for users of text-to-image models like Flux and Stable Diffusion by enhancing guidance stability, though it appears incremental as it builds on existing rectified flow and CFG methods.
The paper tackled the problem of off-manifold drift causing visual artifacts and text misalignment when applying classifier-free guidance to rectified flow models, and introduced Rectified-CFG++, which improved performance on benchmark datasets like MS-COCO and LAION-Aesthetic.
Classifier-free guidance (CFG) is the workhorse for steering large diffusion models toward text-conditioned targets, yet its native application to rectified flow (RF) based models provokes severe off-manifold drift, yielding visual artifacts, text misalignment, and brittle behaviour. We present Rectified-CFG++, an adaptive predictor-corrector guidance that couples the deterministic efficiency of rectified flows with a geometry-aware conditioning rule. Each inference step first executes a conditional RF update that anchors the sample near the learned transport path, then applies a weighted conditional correction that interpolates between conditional and unconditional velocity fields. We prove that the resulting velocity field is marginally consistent and that its trajectories remain within a bounded tubular neighbourhood of the data manifold, ensuring stability across a wide range of guidance strengths. Extensive experiments on large-scale text-to-image models (Flux, Stable Diffusion 3/3.5, Lumina) show that Rectified-CFG++ consistently outperforms standard CFG on benchmark datasets such as MS-COCO, LAION-Aesthetic, and T2I-CompBench. Project page: https://rectified-cfgpp.github.io/