English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

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

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