DyMO: Training-Free Diffusion Model Alignment with Dynamic Multi-Objective Scheduling

TL;DR

DyMO is a training-free, plug-and-play method that improves text-to-image diffusion model alignment with human preferences by dynamically scheduling multiple objectives during inference.

Contribution

It introduces a novel dynamic scheduling approach for multiple objectives and a semantic alignment objective, enhancing inference-time alignment without additional training.

Findings

Effective across diverse diffusion models

Robust in various evaluation metrics

Improves semantic and human preference alignment

Abstract

Text-to-image diffusion model alignment is critical for improving the alignment between the generated images and human preferences. While training-based methods are constrained by high computational costs and dataset requirements, training-free alignment methods remain underexplored and are often limited by inaccurate guidance. We propose a plug-and-play training-free alignment method, DyMO, for aligning the generated images and human preferences during inference. Apart from text-aware human preference scores, we introduce a semantic alignment objective for enhancing the semantic alignment in the early stages of diffusion, relying on the fact that the attention maps are effective reflections of the semantics in noisy images. We propose dynamic scheduling of multiple objectives and intermediate recurrent steps to reflect the requirements at different steps. Experiments with diverse pre-trained diffusion models and metrics demonstrate the effectiveness and robustness of the proposed method.