Supports
Analyses replication of training images and identifies training-data duplication as a contributing factor. Supports the technical possibility of near-copy outputs.[1]
[1] Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion ModelsAcademic Paper
arXiv.org · 12 Dec 2022
Similarity is not a legal infringement test; sampled models and datasets.
Reports reproducible extraction across image-generation systems. Provides additional examples of memorisation beyond a single model study.[2]
[2] A Reproducible Extraction of Training Images from Diffusion ModelsAcademic Paper
arXiv.org
Passage from the sourceCarlini et al. demonstrated the widely used model Stable Diffusion can regurgitate real training samples
Targeted extraction and older models; reproduction frequency in everyday use remains separate.