Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models
Abstract
Inference-time alignment of flow and diffusion-based models is critical for achieving flexible generative modeling. Theoretically, Doob's h-transform provides an elegant solution to this problem, and most existing methods are based on this principle. However, in practice, estimating the optimal guidance derived from Doob's h-transform at inference time is challenging. To deal with this issue, we regard inference-time alignment as a sequential optimization problem in the space of probability measures and propose a novel framework called *Steepest Guidance*, based on the principle of maximizing local improvement in the objective. We provide a theoretical analysis of the proposed method and demonstrate its effectiveness through extensive experiments.
Related
- Risk-Averse Inference-Time Diffusion Model Alignment with Finite-Sample Guarantees
- A Tutorial on Safe Anytime-Valid Inference: Practical Maximally Flexible Sampling Designs for Experiments Based on e-Values
- VGAS: Variance-Reduced Guidance and Adaptive Selection for Training-Free Reward Alignment in Discrete Diffusion
- DCRA: Diffusion-Conditioned Representation Alignment for Robust Time-Series Learning
- Efficient Best-of-N policy evaluation for inference-time alignment
BibTeX
@misc{takakura2026steepest,
title = {{Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models}},
author = {Shokichi Takakura and Akifumi Wachi and Rei Higuchi and Kohei Miyaguchi and Taiji Suzuki},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2609.39091},
eprint = {2609.39091},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.39091},
}
RIS
TY - UNPB
TI - Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models
AU - Shokichi Takakura
AU - Akifumi Wachi
AU - Rei Higuchi
AU - Kohei Miyaguchi
AU - Taiji Suzuki
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2609.39091
UR - https://arxiv.org/abs/2609.39091
AB - Inference-time alignment of flow and diffusion-based models is critical for achieving flexible generative modeling. Theoretically, Doob's h-transform provides an elegant solution to this problem, and most existing methods are based on this principle. However, in practice, estimating the optimal guidance derived from Doob's h-transform at inference time is challenging. To deal with this issue, we regard inference-time alignment as a sequential optimization problem in the space of probability measures and propose a novel framework called *Steepest Guidance*, based on the principle of maximizing local improvement in the objective. We provide a theoretical analysis of the proposed method and demonstrate its effectiveness through extensive experiments.
ER -