PathGuide: Dynamic Classifier-Free Guidance via On-Policy Transport Alignment
Abstract
While modern generative models excel at modeling complex data, precise inference-time control in conditional generation remains a critical challenge. Classifier-free guidance (CFG) is a primary mechanism for such control, yet it is typically treated as a static tuning parameter. In flow-based models, however, the guidance scale fundamentally dictates the velocity field and the resulting probability path, making guidance selection a dynamic path-optimization problem. We introduce PathGuide, a framework that reformulates scalar CFG selection as an on-policy transport problem. Leveraging the weak form of the continuity equation, we derive a selection criterion with a direct path-correctness interpretation: we prove that if the guided field is weakly equivalent to the exact conditional field along the generated rollout, the sampler's path coincides with the target conditional law. For scalar CFG, this criterion yields a strictly quadratic local objective with an efficient, closed-form selector for each solver interval. PathGuide enables optimal guidance scales to be computed and used online during generation or fitted offline as a reusable piecewise-constant schedule. We validate our method on low-resolution image manifolds and controlled settings across various continuous-time flow constructions, demonstrating that this transport-based selector improves path alignment and sample fidelity over both fixed and state-of-the-art adaptive guidance baselines.
Related
- Joint Branch-Space Transform Coding for Diffusion Activation Quantization with Classifier-Free Guidance
- Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models
- VGAS: Variance-Reduced Guidance and Adaptive Selection for Training-Free Reward Alignment in Discrete Diffusion
- Dynamic Generalized Gromov-Wasserstein Optimal Transport
- Dissipative solutions for viscoelastic phase separation with degenerate mobility: a dynamic transport approach
BibTeX
@misc{nevo2026pathguide,
title = {{PathGuide: Dynamic Classifier-Free Guidance via On-Policy Transport Alignment}},
author = {Avishag Nevo and Tamir Hazan},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2608.29107},
eprint = {2608.29107},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.29107},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - PathGuide: Dynamic Classifier-Free Guidance via On-Policy Transport Alignment
AU - Avishag Nevo
AU - Tamir Hazan
PY - 2026
DA - 2026/09/05
PB - arXiv
DO - 10.48550/arxiv.2608.29107
UR - https://arxiv.org/abs/2608.29107
AB - While modern generative models excel at modeling complex data, precise inference-time control in conditional generation remains a critical challenge. Classifier-free guidance (CFG) is a primary mechanism for such control, yet it is typically treated as a static tuning parameter. In flow-based models, however, the guidance scale fundamentally dictates the velocity field and the resulting probability path, making guidance selection a dynamic path-optimization problem. We introduce PathGuide, a framework that reformulates scalar CFG selection as an on-policy transport problem. Leveraging the weak form of the continuity equation, we derive a selection criterion with a direct path-correctness interpretation: we prove that if the guided field is weakly equivalent to the exact conditional field along the generated rollout, the sampler's path coincides with the target conditional law. For scalar CFG, this criterion yields a strictly quadratic local objective with an efficient, closed-form selector for each solver interval. PathGuide enables optimal guidance scales to be computed and used online during generation or fitted offline as a reusable piecewise-constant schedule. We validate our method on low-resolution image manifolds and controlled settings across various continuous-time flow constructions, demonstrating that this transport-based selector improves path alignment and sample fidelity over both fixed and state-of-the-art adaptive guidance baselines.
ER -