Diffusion models overfit denoising loss at intermediate noise but generalize in inference as model error smooths the flow field and sampling paths avoid memorized noisy training data.
arXiv preprint (2025),https: //arxiv.org/abs/2502.07516
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
The UPDP pipeline filters privacy terms and generates de-identified radiology images that preserve diagnostic pathology information, enabling models with competitive disease detection accuracy but reduced identity leakage and improved cross-hospital performance.
citing papers explorer
-
Diffusion Models Memorize in Training -- and Generalize in Inference
Diffusion models overfit denoising loss at intermediate noise but generalize in inference as model error smooths the flow field and sampling paths avoid memorized noisy training data.
-
A Utility-preserving De-identification Pipeline for Cross-hospital Radiology Data Sharing
The UPDP pipeline filters privacy terms and generates de-identified radiology images that preserve diagnostic pathology information, enabling models with competitive disease detection accuracy but reduced identity leakage and improved cross-hospital performance.