@misc{indiciae70ee0233b461,
  title = {When Pretty Isn't Useful: Investigating Why Modern Text-to-Image Models Fail as Reliable Training Data Generators},
  author = {Krzysztof Adamkiewicz and Brian Bernhard Moser and Stanislav Frolov and Tobias Christian Nauen and Federico Raue and Andreas Dengel},
  year = {2026},
  url = {https://arxiv.org/abs/2602.19946},
  note = {Source identifier: 2602.19946}
}