@misc{indiciaee8c3e24012e5,
  title = {Recursive Training Loops in LLMs: How training data properties modulate distribution shift in generated data?},
  author = {Grgur Kovač and Jérémy Perez and Rémy Portelas and Peter Ford Dominey and Pierre-Yves Oudeyer},
  year = {2025},
  url = {https://arxiv.org/abs/2504.03814},
  note = {Source identifier: 2504.03814}
}