A theoretical basis for model collapse in recursive training
Fuente:
arXiv
Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914058620370944 |
|---|---|
| author | Borkar, Vivek Shripad |
| author_facet | Borkar, Vivek Shripad |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_09401 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A theoretical basis for model collapse in recursive training Borkar, Vivek Shripad Probability Machine Learning 68T01 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. |
| title | A theoretical basis for model collapse in recursive training |
| topic | Probability Machine Learning 68T01 |
| url | https://arxiv.org/abs/2506.09401 |