Adaptive Diffusion Guidance via Stochastic Optimal Control

2025-05-25

AI Summary

Current diffusion models employ static classifier guidance weights, limiting generation quality and controllability while lacking theoretical grounding. Method: We propose the first adaptive guidance scheduling framework grounded in stochastic optimal control: (i) we establish a theoretical connection between guidance strength and classifier confidence; (ii) we formulate guidance scheduling as a dynamic optimal control problem dependent on time, sample state, and conditional class; and (iii) we derive principled, sample-adaptive weight optimization via diffusion process modeling and conditional generation theory. Contribution/Results: Our approach significantly improves fidelity and robustness of conditional generation without compromising sampling efficiency. It provides an interpretable, theoretically justified foundation for guidance in diffusion models—enabling both explainability and optimization of the guidance mechanism.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Reasoning under Uncertainty: Stochastic Optimization
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Guidance is a cornerstone of modern diffusion models, playing a pivotal role in conditional generation and enhancing the quality of unconditional samples. However, current approaches to guidance scheduling--determining the appropriate guidance weight--are largely heuristic and lack a solid theoretical foundation. This work addresses these limitations on two fronts. First, we provide a theoretical formalization that precisely characterizes the relationship between guidance strength and classifier confidence. Second, building on this insight, we introduce a stochastic optimal control framework that casts guidance scheduling as an adaptive optimization problem. In this formulation, guidance strength is not fixed but dynamically selected based on time, the current sample, and the conditioning class, either independently or in combination. By solving the resulting control problem, we establish a principled foundation for more effective guidance in diffusion models.

Problem

• Theoretical formalization of guidance strength and classifier confidence relationship
• Dynamic guidance scheduling via stochastic optimal control framework
• Adaptive optimization for effective diffusion model guidance

Innovation

• Theoretical formalization of guidance strength relationship
• Stochastic optimal control for guidance scheduling
• Dynamic guidance strength based on multiple factors