Trade-offs in Fine-tuned Diffusion Models Between Accuracy and Interpretability

TL;DR

This paper investigates the balance between accuracy and interpretability in fine-tuned diffusion models, especially in medical imaging, revealing a trade-off influenced by the choice of text encoders and proposing design principles for interpretability.

Contribution

It uncovers the trade-off between image fidelity and interpretability in fine-tuned diffusion models and offers design principles to enhance interpretability in generative models.

Findings

Learnable text encoders reduce interpretability.

Trade-off exists between image quality and interpretability.

Design principles can improve model interpretability.

Abstract

Recent advancements in diffusion models have significantly impacted the trajectory of generative machine learning research, with many adopting the strategy of fine-tuning pre-trained models using domain-specific text-to-image datasets. Notably, this method has been readily employed for medical applications, such as X-ray image synthesis, leveraging the plethora of associated radiology reports. Yet, a prevailing concern is the lack of assurance on whether these models genuinely comprehend their generated content. With the evolution of text-conditional image generation, these models have grown potent enough to facilitate object localization scrutiny. Our research underscores this advancement in the critical realm of medical imaging, emphasizing the crucial role of interpretability. We further unravel a consequential trade-off between image fidelity as gauged by conventional metrics and model interpretability in generative diffusion models. Specifically, the adoption of learnable text encoders when fine-tuning results in diminished interpretability. Our in-depth exploration uncovers the underlying factors responsible for this divergence. Consequently, we present a set of design principles for the development of truly interpretable generative models. Code is available at https://github.com/MischaD/chest-distillation.