Generating on Generated: An Approach Towards Self-Evolving Diffusion Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866929715395166208 |
|---|---|
| author | Zhang, Xulu Wei, Xiaoyong Wu, Jinlin Wu, Jiaxin Zhang, Zhaoxiang Lei, Zhen Li, Qing |
| author_facet | Zhang, Xulu Wei, Xiaoyong Wu, Jinlin Wu, Jiaxin Zhang, Zhaoxiang Lei, Zhen Li, Qing |
| contents | Recursive Self-Improvement (RSI) enables intelligence systems to autonomously refine their capabilities. This paper explores the application of RSI in text-to-image diffusion models, addressing the challenge of training collapse caused by synthetic data. We identify two key factors contributing to this collapse: the lack of perceptual alignment and the accumulation of generative hallucinations. To mitigate these issues, we propose three strategies: (1) a prompt construction and filtering pipeline designed to facilitate the generation of perceptual aligned data, (2) a preference sampling method to identify human-preferred samples and filter out generative hallucinations, and (3) a distribution-based weighting scheme to penalize selected samples with hallucinatory errors. Our extensive experiments validate the effectiveness of these approaches. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_09963 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Generating on Generated: An Approach Towards Self-Evolving Diffusion Models Zhang, Xulu Wei, Xiaoyong Wu, Jinlin Wu, Jiaxin Zhang, Zhaoxiang Lei, Zhen Li, Qing Computer Vision and Pattern Recognition Recursive Self-Improvement (RSI) enables intelligence systems to autonomously refine their capabilities. This paper explores the application of RSI in text-to-image diffusion models, addressing the challenge of training collapse caused by synthetic data. We identify two key factors contributing to this collapse: the lack of perceptual alignment and the accumulation of generative hallucinations. To mitigate these issues, we propose three strategies: (1) a prompt construction and filtering pipeline designed to facilitate the generation of perceptual aligned data, (2) a preference sampling method to identify human-preferred samples and filter out generative hallucinations, and (3) a distribution-based weighting scheme to penalize selected samples with hallucinatory errors. Our extensive experiments validate the effectiveness of these approaches. |
| title | Generating on Generated: An Approach Towards Self-Evolving Diffusion Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2502.09963 |