paper-with-me

홈 › Papers

A theoretical basis for model collapse in recursive training

2025-06-11 · Vivek Shripad Borkar

It is known that recursive training from generative models can lead to the so called `collapse' of the simulated probability distribution. This note shows that one in fact gets two different asymptotic behaviours depending on whether an external source, howsoever minor, is also contributing samples.

📄 PDF Abstract BibTeX arXiv:2506.09401

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Knowledge Collapse in LLMs: When Fluency Survives but Facts Fail under Recursive Synthetic Training

2025-09-05 · Figarri Keisha, Zekun Wu, Ze Wang, Adriano Koshiyama 외 arxiv

Large language models increasingly rely on synthetic data due to human-written content scarcity, yet recursive training on model-generated outputs leads to model collapse, a degenerative process threatening factual relia…

Computational Efficiency

Rate of Model Collapse in Recursive Training

2024-12-23 · Ananda Theertha Suresh, Andrew Thangaraj, Aditya Nanda Kishore Khandavally

Given the ease of creating synthetic data from machine learning models, new models can be potentially trained on synthetic data generated by previous models. This recursive training process raises concerns about the long…

model

How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapse

2024-04-07 · Mohamed El Amine Seddik, Suei-Wen Chen, Soufiane Hayou, Pierre Youssef 외

The phenomenon of model collapse, introduced in (Shumailov et al., 2023), refers to the deterioration in performance that occurs when new models are trained on synthetic data generated from previously trained models. Thi…

Language ModelingLanguage Modelling

Curated Synthetic Data Doesn't Have to Collapse: A Theoretical Study of Generative Retraining with Pluralistic Preferences

2026-05-08 · Ali Falahati, Mohammad Mohammadi Amiri, Kate Larson, Lukasz Golab arxiv

Recursive retraining of generative models poses a critical representation challenge: when synthetic outputs are curated based on a fixed reward signal, the model tends to collapse onto a narrow set of outputs that over-o…

A Probabilistic Perspective on Model Collapse

2025-05-20 · SHIRONG XU, Hengzhi He, Guang Cheng

In recent years, model collapse has become a critical issue in language model training, making it essential to understand the underlying mechanisms driving this phenomenon. In this paper, we investigate recursive paramet…

model