Latent Structure Emergence in Diffusion Models via Confidence-Based Filtering
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911425819049984 |
|---|---|
| author | Wei, Wei Zeng, Yizhou Chen, Kuntian Langer, Sophie Seleznova, Mariia Chou, Hung-Hsu |
| author_facet | Wei, Wei Zeng, Yizhou Chen, Kuntian Langer, Sophie Seleznova, Mariia Chou, Hung-Hsu |
| contents | Diffusion models rely on a high-dimensional latent space of initial noise seeds, yet it remains unclear whether this space contains sufficient structure to predict properties of the generated samples, such as their classes. In this work, we investigate the emergence of latent structure through the lens of confidence scores assigned by a pre-trained classifier to generated samples. We show that while the latent space appears largely unstructured when considering all noise realizations, restricting attention to initial noise seeds that produce high-confidence samples reveals pronounced class separability. By comparing class predictability across noise subsets of varying confidence and examining the class separability of the latent space, we find evidence of class-relevant latent structure that becomes observable only under confidence-based filtering. As a practical implication, we discuss how confidence-based filtering enables conditional generation as an alternative to guidance-based methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_06155 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Latent Structure Emergence in Diffusion Models via Confidence-Based Filtering Wei, Wei Zeng, Yizhou Chen, Kuntian Langer, Sophie Seleznova, Mariia Chou, Hung-Hsu Machine Learning Diffusion models rely on a high-dimensional latent space of initial noise seeds, yet it remains unclear whether this space contains sufficient structure to predict properties of the generated samples, such as their classes. In this work, we investigate the emergence of latent structure through the lens of confidence scores assigned by a pre-trained classifier to generated samples. We show that while the latent space appears largely unstructured when considering all noise realizations, restricting attention to initial noise seeds that produce high-confidence samples reveals pronounced class separability. By comparing class predictability across noise subsets of varying confidence and examining the class separability of the latent space, we find evidence of class-relevant latent structure that becomes observable only under confidence-based filtering. As a practical implication, we discuss how confidence-based filtering enables conditional generation as an alternative to guidance-based methods. |
| title | Latent Structure Emergence in Diffusion Models via Confidence-Based Filtering |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2602.06155 |