Accelerated Likelihood Maximization for Diffusion-based Versatile Content Generation

2026-06-30

AI Summary

Existing diffusion models face significant challenges in content generation under partial observation: training-based approaches suffer from limited generalization, while training-free methods often produce globally inconsistent results. This work proposes a novel training-free sampling strategy that explicitly optimizes unobserved regions during the reverse diffusion process and incorporates an efficient acceleration mechanism. For the first time, it enables direct, globally consistent optimization of missing content without any additional training. Extensive experiments demonstrate that the method consistently outperforms state-of-the-art approaches across diverse data domains and tasks, achieving superior quality, high efficiency, and strong global consistency. These advances substantially broaden the applicability of diffusion models in general-purpose content generation.

Abstract

Generating diverse, coherent, and plausible content from partially given inputs remains a fundamental challenge for diffusion models. Existing approaches face clear limitations: training-based approaches offer strong task-specific results but require costly computation, and they generalize poorly across tasks. Training-free approaches offer better efficiency, but they do not explicitly optimize over unobserved variables, leading to globally inconsistent results. To address these limitations, we introduce Accelerated Likelihood Maximization (ALM), a novel training-free sampling strategy integrated into the reverse diffusion process that significantly extends the applicability of diffusion models beyond simple generation tasks. Unlike previous methods that implicitly influence missing regions through pre-generated region constraints, we directly optimize the unobserved region during the sampling process, enabling globally coherent and plausible generation. Furthermore, we incorporate an acceleration strategy that significantly improves computational efficiency without sacrificing performance. Experimental results demonstrate that ALM consistently outperforms state-of-the-art methods in various data domains and tasks, establishing a powerful paradigm for versatile content generation.

Problem

• diffusion models
• content generation
• likelihood maximization
• training-free methods
• global consistency

Innovation

• Accelerated Likelihood Maximization
• diffusion models
• training-free sampling
• global coherence
• versatile content generation