Collapse or Thrive: Perils and Promises of Synthetic Data in a Self-Generating World

Joshua Kazdan, Rylan Schaeffer, Apratim Dey, Matthias Gerstgrasser, Rafael Rafailov +2 more
2/14/2026

No official code found. Best implementation: RylanSchaeffer/KoyejoLab-Collapse-or-Thrive (9 stars, last commit 2025-07-05).

Semantic Scholar

Code Implementations(1)

RylanSchaeffer/KoyejoLab-Collapse-or-ThriveCommunity59%

Code for ICML 2025 Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World

92Shell, PythonSep 12, 20241 years ago

Cite this paper

@article{kazdan2026collapse,
  title  = {Collapse or Thrive: Perils and Promises of Synthetic Data in a Self-Generating World},
  author = {Joshua Kazdan and Rylan Schaeffer and Apratim Dey and Matthias Gerstgrasser and Rafael Rafailov and David L. Donoho and Oluwasanmi Koyejo},
  year   = {2026},
  url    = {https://api.semanticscholar.org/CorpusID:eea13feec764338479e270b92f69b87a36db9215},
  journal = {ICML 2025 2025}
}

Discussion