mlsanityData to 14 Oct 2026

Large Language Models Suffer From Their Own Output: An Analysis of the Self-Consuming Training Loop

arXiv 2311.16822 ↗First posted 28 Nov 2023cs.LG cs.CL cs.NEcited by 79

Not on the map. Why?

Cited by 33 in Synthetic data and self-improvement for LLMs, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Large Language Models (LLM) are already widely used to generate content for a variety of online platforms. As we are not able to safely distinguish LLM-generated content from human-produced content, LLM-generated content is used to train the next generation of LLMs, giving rise to a self-consuming training loop. From the image generation domain we know that such a self-consuming training loop reduces both quality and diversity of images finally ending in a model collapse. However, it is unclear whether this alarming effect can also be observed for LLMs. Therefore, we present the first study investigating the self-consuming training loop for LLMs. Further, we propose a novel method based on logic expressions that allows us to unambiguously verify the correctness of LLM-generated content, which is difficult for natural language text. We find that the self-consuming training loop produces correct outputs, however, the output declines in its diversity depending on the proportion of the used generated data. Fresh data can slow down this decline, but not stop it. Given these concerning results, we encourage researchers to study methods to negate this process.

Key facts. Cited by 79 papers in the map (data through 14 Oct 2026). Builds on 23 of its 34 references that are in the map. First posted on arXiv 28 Nov 2023.

Builds on

23 of 34 references.

References · 23

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2022Training language models to follow instructions with human feedbackcited by 15,971
  3. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  4. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,512
  5. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  6. 2019Energy and Policy Considerations for Deep Learning in NLPcited by 1,254
  7. 2022Large Language Models Can Self-Improvecited by 664
  8. 2023Can AI-Generated Text be Reliably Detected?cited by 386
  9. 2023The Curse of Recursion: Training on Generated Data Makes Models Forgetcited by 356
  10. 2023Self-Consuming Generative Models Go MADcited by 250
  11. 2023Large Language Model as Attributed Training Data Generator: A Tale of Diversity and Biascited by 232
  12. 2022Will we run out of data? Limits of LLM scaling based on human-generated datacited by 195
11 more
  1. 2023A Categorical Archive of ChatGPT Failurescited by 168
  2. 2022Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Textcited by 155
  3. 2023Exploring the Limits of ChatGPT for Query or Aspect-based Text Summarizationcited by 122
  4. 2023An Analysis of the Automatic Bug Fixing Performance of ChatGPTcited by 110
  5. 2023Artificial Artificial Artificial Intelligence: Crowd Workers Widely Use Large Language Models for Text Production Taskscited by 103
  6. 2023On the Stability of Iterative Retraining of Generative Models on their own Datacited by 100
  7. 2023The Curious Decline of Linguistic Diversity: Training Language Models on Synthetic Textcited by 100
  8. 2021Differentiable Quality Diversitycited by 72
  9. 2023Towards Understanding the Interplay of Generative Artificial Intelligence and the Internetcited by 54
  10. 2021Choose Your Programming Copilot: A Comparison of the Program Synthesis Performance of GitHub Copilot and Genetic Programmingcited by 24
  11. 2023Do You Trust ChatGPT? -- Perceived Credibility of Human and AI-Generated Contentcited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 35; 2025: 24; 2026: 20.

79 papers.

Sort:

By area:

  1. 2025R-Zero: Self-Evolving Reasoning LLM from Zero Datacited by 228
  2. 2024Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Datacited by 151
  3. 2024A Tale of Tails: Model Collapse as a Change of Scaling Lawscited by 130
  4. 2024How to Train Data-Efficient LLMscited by 93
  5. 2024Understanding Hallucinations in Diffusion Models through Mode Interpolationcited by 91
  6. 2024Model Collapse Demystified: The Case of Regressioncited by 82
  7. 2024How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapsecited by 79
  8. 2024Strong Model Collapsecited by 65
  9. 2024DenseFormer: Enhancing Information Flow in Transformers via Depth Weighted Averagingcited by 53
  10. 2024Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating Worldcited by 52
  11. 2025The Best Instruction-Tuning Data are Those That Fitcited by 52
  12. 2024Beyond Model Collapse: Scaling Up with Synthesized Data Requires Verificationcited by 45
38 more
  1. 2024Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferencescited by 39
  2. 2024Are We in the AI-Generated Text World Already? Quantifying and Monitoring AIGT on Social Mediacited by 35
  3. 2025Position: Model Collapse Does Not Mean What You Thinkcited by 33
  4. 2025Self-Improving Transformers Overcome Easy-to-Hard and Length Generalization Challengescited by 31
  5. 2024Towards Theoretical Understandings of Self-Consuming Generative Modelscited by 29
  6. 2024How to Synthesize Text Data without Model Collapse?cited by 25
  7. 2024Under the Surface: Tracking the Artifactuality of LLM-Generated Datacited by 24
  8. 2024Organic or Diffused: Can We Distinguish Human Art from AI-generated Images?cited by 21
  9. 2025BeyondWeb: Lessons from Scaling Synthetic Data for Trillion-scale Pretrainingcited by 21
  10. 2024Predicting vs. Acting: A Trade-off Between World Modeling & Agent Modelingcited by 20
  11. 2025Generation from Noisy Examplescited by 20
  12. 2024ATM: Adversarial Tuning Multi-agent System Makes a Robust Retrieval-Augmented Generatorcited by 19
  13. 2026TTCS: Test-Time Curriculum Synthesis for Self-Evolvingcited by 19
  14. 2024The Impact of Large Language Models in Academia: from Writing to Speakingcited by 18
  15. 2024Mind the Gap: Examining the Self-Improvement Capabilities of Large Language Modelscited by 18
  16. 2025A Theoretical Perspective: How to Prevent Model Collapse in Self-consuming Training Loopscited by 16
  17. 2024Step-On-Feet Tuning: Scaling Self-Alignment of LLMs via Bootstrappingcited by 14
  18. 2025Machine-generated text detection prevents language model collapsecited by 14
  19. 2025When Models Don't Collapse: On the Consistency of Iterative MLEcited by 13
  20. 2024Model Collapse in the Self-Consuming Chain of Diffusion Finetuning: A Novel Perspective from Quantitative Trait Modelingcited by 12
  21. 2024Beware of Words: Evaluating the Lexical Diversity of Conversational LLMs using ChatGPT as Case Studycited by 10
  22. 2025Preventing Model Collapse Under Overparametrization: Optimal Mixing Ratios for Interpolation Learning and Ridge Regressioncited by 9
  23. 2024Why Does ChatGPT "Delve" So Much? Exploring the Sources of Lexical Overrepresentation in Large Language Modelscited by 8
  24. 2025Why Less is More (Sometimes): A Theory of Data Curationcited by 8
  25. 2024MONAL: Model Autophagy Analysis for Modeling Human-AI Interactionscited by 7
  26. 2024Multilingual Arbitrage: Optimizing Data Pools to Accelerate Multilingual Progresscited by 7
  27. 2024Leveraging Large Language Models to Detect npm Malicious Packagescited by 6
  28. 2024LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectivescited by 6
  29. 2024A Little Human Data Goes A Long Waycited by 6
  30. 2024Exploring the Escalation of Source Bias in User, Data, and Recommender System Feedback Loopcited by 5
  31. 2026Language Generation with Replay: A Learning-Theoretic View of Model Collapsecited by 5
  32. 2026Self-Improvement of Large Language Models: A Technical Overview and Future Outlookcited by 5
  33. 2024Improving Fairness and Mitigating MADness in Generative Modelscited by 4
  34. 2025Model Misalignment and Language Change: Traces of AI-Associated Language in Unscripted Spoken Englishcited by 4
  35. 2024Recursive InPainting (RIP): how much information is lost under recursive inferences?cited by 2
  36. 2024Learning by Surprise: Adaptive Mitigation of Model Collapse in Large Language Modelscited by 2
  37. 2024The Superalignment of Superhuman Intelligence with Large Language Modelscited by 2
  38. 2025Swallowing the Poison Pills: Insights from Vulnerability Disparity Among LLMscited by 2

50 most cited of 79 All 79 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Demystifying Synthetic Data in LLM Pre-training: A Systematic Study of Scaling Laws, Benefits, and Pitfalls
  2. 2023Automatically Correcting Large Language Models: Surveying the landscape of diverse self-correction strategies
  3. 2025Self-Improvement in Multimodal Large Language Models: A Survey
  4. 2024Self-training Large Language Models through Knowledge Detection
  5. 2023Large Language Models Cannot Self-Correct Reasoning Yet
  6. 2026Experimental evidence of progressive ChatGPT models self-convergence
  7. 2024Regurgitative Training: The Value of Real Data in Training Large Language Models
  8. 2023Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes