Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.06124 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917467987640320 |
|---|---|
| author | Peng, Xin Gao, Ang |
| author_facet | Peng, Xin Gao, Ang |
| contents | Classifier-Free Guidance (CFG) is essential for high-fidelity conditional generation in flow matching, yet it imposes significant computational overhead by requiring dual forward passes at each sampling step. In this work, we address this bottleneck by introducing \textbf{P-Guide}, a framework that achieves high-quality guidance through a single inference pass by modulating only the initial latent state. We further show that, under a first-order approximation, P-Guide is equivalent to CFG in the sense that it steers generation from the prior space, without requiring explicit velocity field extrapolation during sampling. We consider both homoscedastic and \textbf{heteroscedastic} priors, and find that jointly modeling the mean and variance enables adaptive loss attenuation and improved robustness to data uncertainty. Extensive experiments demonstrate that P-Guide reduces inference latency by approximately 50\% while maintaining fidelity and prompt alignment competitive with standard dual-pass CFG baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_06124 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | P-Guide: Parameter-Efficient Prior Steering for Single-Pass CFG Inference Peng, Xin Gao, Ang Artificial Intelligence Classifier-Free Guidance (CFG) is essential for high-fidelity conditional generation in flow matching, yet it imposes significant computational overhead by requiring dual forward passes at each sampling step. In this work, we address this bottleneck by introducing \textbf{P-Guide}, a framework that achieves high-quality guidance through a single inference pass by modulating only the initial latent state. We further show that, under a first-order approximation, P-Guide is equivalent to CFG in the sense that it steers generation from the prior space, without requiring explicit velocity field extrapolation during sampling. We consider both homoscedastic and \textbf{heteroscedastic} priors, and find that jointly modeling the mean and variance enables adaptive loss attenuation and improved robustness to data uncertainty. Extensive experiments demonstrate that P-Guide reduces inference latency by approximately 50\% while maintaining fidelity and prompt alignment competitive with standard dual-pass CFG baselines. |
| title | P-Guide: Parameter-Efficient Prior Steering for Single-Pass CFG Inference |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2605.06124 |