ContinualFlow: Learning and Unlearning with Neural Flow Matching
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_ | 1866916807354351616 |
|---|---|
| author | Simone, Lorenzo Bacciu, Davide Ma, Shuangge |
| author_facet | Simone, Lorenzo Bacciu, Davide Ma, Shuangge |
| contents | We introduce ContinualFlow, a principled framework for targeted unlearning in generative models via Flow Matching. Our method leverages an energy-based reweighting loss to softly subtract undesired regions of the data distribution without retraining from scratch or requiring direct access to the samples to be unlearned. Instead, it relies on energy-based proxies to guide the unlearning process. We prove that this induces gradients equivalent to Flow Matching toward a soft mass-subtracted target, and validate the framework through experiments on 2D and image domains, supported by interpretable visualizations and quantitative evaluations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_18747 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | ContinualFlow: Learning and Unlearning with Neural Flow Matching Simone, Lorenzo Bacciu, Davide Ma, Shuangge Machine Learning Artificial Intelligence We introduce ContinualFlow, a principled framework for targeted unlearning in generative models via Flow Matching. Our method leverages an energy-based reweighting loss to softly subtract undesired regions of the data distribution without retraining from scratch or requiring direct access to the samples to be unlearned. Instead, it relies on energy-based proxies to guide the unlearning process. We prove that this induces gradients equivalent to Flow Matching toward a soft mass-subtracted target, and validate the framework through experiments on 2D and image domains, supported by interpretable visualizations and quantitative evaluations. |
| title | ContinualFlow: Learning and Unlearning with Neural Flow Matching |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2506.18747 |