Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models
stat.ML, cs.LG
Submitted: 2026-09-30
Updated: 2026-10-05
Code: https://github.com/black-forest-labs/flux
Terminology
Sources
- Reinforce Adjoint Matching: Scaling RL Post-Training of Diffusion and Flow-Matching Models
- Greed is Good: A Unifying Perspective on Guided Generation
- Particle Guidance: non-I.I.D. Diverse Sampling with Diffusion Models
- Are we really tilting? The mechanics of reward guidance in flow and diffusion models
- Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Control
- Classifier-Free Diffusion Guidance
- GLASS Flows: Transition Sampling for Alignment of Flow and Diffusion Models
- Diamond Maps: Efficient Reward Alignment via Stochastic Flow Maps
- Towards a unified framework for guided diffusion models
- Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-Based Decoding
- Flow Matching for Generative Modeling
- Implicit Diffusion: Efficient Optimization through Stochastic Sampling
- WebGPT: Browser-assisted question-answering with human feedback
- Slowly Annealed Langevin Dynamics: Theory and Applications to Training-Free Guided Generation
- Inference-Aware Meta-Alignment of LLMs via Non-Linear GRPO
- Fine-Tuning of Continuous-Time Diffusion Models as Entropy-Regularized Control
- Diverse Sampling in Diffusion Models with Divergence-Free Particle Guidance
- Efficient Tail-Aware Generative Optimization via Flow Model Fine-Tuning
- Training-Free Adaptation of Diffusion Models via Doob's $h$-Transform
- Repulsive Latent Score Distillation for Solving Inverse Problems
Related papers
- Behavior of prediction performance metrics with rare events
- Optimal Estimation of Generic Dynamics by Path-Dependent Neural Jump ODEs
- A Posterior-Dynamics Framework for Imaging Inverse Problems with Pretrained Diffusion Priors
- One Permutation Is All You Need: Fast, Deterministic Feature Importance and Model Stress-Testing
- Online Conformal Prediction for Non-Exchangeable Panel Data
- Deep Time-Series Forecasting in 10 Years: A Survey