Dataless Weight Disentanglement in Task Arithmetic via Kronecker-Factored Approximate Curvature
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866917518651686912 |
|---|---|
| author | Porrello, Angelo Buzzega, Pietro Dangel, Felix Sommariva, Thomas Salami, Riccardo Bonicelli, Lorenzo Calderara, Simone |
| author_facet | Porrello, Angelo Buzzega, Pietro Dangel, Felix Sommariva, Thomas Salami, Riccardo Bonicelli, Lorenzo Calderara, Simone |
| contents | Task Arithmetic yields a modular, scalable way to adapt foundation models. Combining multiple task vectors, however, can lead to cross-task interference, causing representation drift and degraded performance. Representation drift regularization provides a natural remedy to disentangle task vectors; however, existing approaches typically require external task data, conflicting with modularity and data availability constraints (e.g., privacy requirements). We propose a dataless approach by framing regularization against representation drift as a curvature matrix approximation problem. This allows us to leverage well-established techniques; in particular, we adopt Kronecker-Factored Approximate Curvature and obtain a practical regularizer that achieves state-of-the-art results in task addition and negation. Our method has constant complexity in the number of tasks and promotes robustness to task vector rescaling, eliminating the need for held-out tuning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_17385 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Dataless Weight Disentanglement in Task Arithmetic via Kronecker-Factored Approximate Curvature Porrello, Angelo Buzzega, Pietro Dangel, Felix Sommariva, Thomas Salami, Riccardo Bonicelli, Lorenzo Calderara, Simone Artificial Intelligence Task Arithmetic yields a modular, scalable way to adapt foundation models. Combining multiple task vectors, however, can lead to cross-task interference, causing representation drift and degraded performance. Representation drift regularization provides a natural remedy to disentangle task vectors; however, existing approaches typically require external task data, conflicting with modularity and data availability constraints (e.g., privacy requirements). We propose a dataless approach by framing regularization against representation drift as a curvature matrix approximation problem. This allows us to leverage well-established techniques; in particular, we adopt Kronecker-Factored Approximate Curvature and obtain a practical regularizer that achieves state-of-the-art results in task addition and negation. Our method has constant complexity in the number of tasks and promotes robustness to task vector rescaling, eliminating the need for held-out tuning. |
| title | Dataless Weight Disentanglement in Task Arithmetic via Kronecker-Factored Approximate Curvature |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2602.17385 |