mlsanityData to 7 Oct 2026

Collapse of Self-trained Language Models

arXiv 2404.02305 ↗First posted 2 Apr 2024cs.CL cs.AIcited by 5

On the map: Synthetic data and self-improvement for LLMs → Self-rewarding alignment and self-training

Abstract (arXiv)

In various fields of knowledge creation, including science, new ideas often build on pre-existing information. In this work, we explore this concept within the context of language models. Specifically, we explore the potential of self-training models on their own outputs, akin to how humans learn and build on their previous thoughts and actions. While this approach is intuitively appealing, our research reveals its practical limitations. We find that extended self-training of the GPT-2 model leads to a significant degradation in performance, resulting in repetitive and collapsed token output.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in Self-rewarding alignment and self-training, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 3 of its 8 references that are in the map. First posted on arXiv 2 Apr 2024.

Builds on

3 of 8 references.

References · 3

  1. 2016Pointer Sentinel Mixture Modelscited by 2,992
  2. 2017Dynamic Evaluation of Neural Sequence Modelscited by 53
  3. 2020Gradient Amplification: An efficient way to train deep neural networkscited by 15

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 4.

5 papers.

  1. 2025The Best Instruction-Tuning Data are Those That Fitcited by 48
  2. 2026Formal Disco: Scalable Open-Ended Generation of Formally Verified Programscited by 1
  3. 2026Data Selection Through Iterative Self-Filtering for Vision-Language Settingscited by 0
  4. 2026Preventing Model Collapse: A Fisher-Rao Perspective on the Dynamics of Training with Synthetic Datacited by 0
  5. 2026Scope Before You Persist: Preventing Cross-Family Interference in Agent Memorycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructions
  2. 2026Self-Improvement of Large Language Models: A Technical Overview and Future Outlook
  3. 2026Continually self-improving AI
  4. 2024Mind the Gap: Examining the Self-Improvement Capabilities of Large Language Models
  5. 2024Self-Improvement in Language Models: The Sharpening Mechanism
  6. 2026Linger and Lose: Knowledge Collapse in Low-Bit Language Models
  7. 2025The Path of Self-Evolving Large Language Models: Achieving Data-Efficient Learning via Intrinsic Feedback
  8. 2024Preserving Knowledge in Large Language Model with Model-Agnostic Self-Decompression