Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 32 of its 41 references that are in the map. First posted on arXiv 30 Sep 2026.
Builds on
32 of 41 references.
References · 32
- 2020Denoising Diffusion Probabilistic Modelscited by 21,076
- 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
- 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,255
- 2022Classifier-Free Diffusion Guidancecited by 5,847
- 2022Flow Matching for Generative Modelingcited by 5,702
- 2021WebGPT: Browser-assisted question-answering with human feedbackcited by 1,745
- 2022Diffusion-LM Improves Controllable Text Generationcited by 1,027
- 2023Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generationcited by 977
- 2023Stochastic Interpolants: A Unifying Framework for Flows and Diffusionscited by 939
- 2022Equivariant Diffusion for Molecule Generation in 3Dcited by 677
- 2024Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Controlcited by 233
- 2023Practical and Asymptotically Exact Conditional Sampling in Diffusion Modelscited by 232
20 more
- 2024Fine-Tuning of Continuous-Time Diffusion Models as Entropy-Regularized Controlcited by 143
- 2024Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-Based Decodingcited by 136
- 2020Variational Policy Gradient Method for Reinforcement Learning with General Utilitiescited by 111
- 2023Particle Guidance: non-I.I.D. Diverse Sampling with Diffusion Modelscited by 71
- 2022Mean-Field Langevin Dynamics: Exponential Convergence and Annealingcited by 61
- 2024Gradient Guidance for Diffusion Models: An Optimization Perspectivecited by 61
- 2022Convex Analysis of the Mean Field Langevin Dynamicscited by 55
- 2026Diamond Maps: Efficient Reward Alignment via Stochastic Flow Mapscited by 29
- 2024Implicit Diffusion: Efficient Optimization through Stochastic Samplingcited by 24
- 2026Reinforce Adjoint Matching: Scaling RL Post-Training of Diffusion and Flow-Matching Modelscited by 19
- 2025GLASS Flows: Transition Sampling for Alignment of Flow and Diffusion Modelscited by 17
- 2025Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuningcited by 17
- 2026Training-Free Adaptation of Diffusion Models via Doob's h-Transformcited by 11
- 2025Towards a unified framework for guided diffusion modelscited by 8
- 2026Efficient Tail-Aware Generative Optimization via Flow Model Fine-Tuningcited by 8
- 2025Direct Distributional Optimization for Provable Alignment of Diffusion Modelscited by 5
- 2026Are we really tilting? The mechanics of reward guidance in flow and diffusion modelscited by 3
- 2025Greed is Good: A Unifying Perspective on Guided Generationcited by 2
- 2026Inference-Aware Meta-Alignment of LLMs via Non-Linear GRPOcited by 1
- 2026Slowly Annealed Langevin Dynamics: Theory and Applications to Training-Free Guided Generationcited by 1
Built on by
Not cited yet.