EUVIMEDEUVIMED entdecken
Crossref

From Parameter to Representation: A Closed-Form Approach for Controllable Model Merging

Jialin Wu, Jian Yang, Handing Wang, Jiajun Wen, Zhiyong Yu

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (32) · 26903-26912

DOI: 10.1609/aaai.v40i32.39902 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Jialin Wu und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Model merging combines expert models for multitask performance but faces challenges from parameter interference. This has sparked recent interest in controllable model merging, giving users the ability to explicitly balance performance trade-offs. Existing approaches employ a compile-then-query paradigm, performing a costly offline multi-objective optimization to enable fast, preference-aware model generation. This offline stage typically involves iterative search or dedicated training, with complexity that grows exponentially with the number of tasks. To overcome these limitations, we shift the perspective from parameter-space optimization to a direct correction of the model's final representation. Our approach models this correction as an optimal linear transformation, yielding a closed-form solution that replaces the entire offline optimization process with a single-step, architecture-agnostic computation. This solution directly incorporates user preferences, allowing a Pareto-optimal model to be generated on-the-fly with complexity that scales linearly with the number of tasks. Experimental results show our method generates a superior Pareto front with more precise preference alignment and drastically reduced computational cost.

Zitieren

APA

Wu, J., Yang, J., Wang, H., Wen, J., & Yu, Z. (2026). From Parameter to Representation: A Closed-Form Approach for Controllable Model Merging. Proceedings of the AAAI Conference on Artificial Intelligence, 40(32), 26903-26912. https://doi.org/10.1609/aaai.v40i32.39902

Vancouver

Wu J, Yang J, Wang H, Wen J, Yu Z. From Parameter to Representation: A Closed-Form Approach for Controllable Model Merging. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(32):26903-26912. doi:10.1609/aaai.v40i32.39902

BibTeX
@article{Wu2026,
  author = {Wu, Jialin and Yang, Jian and Wang, Handing and Wen, Jiajun and Yu, Zhiyong},
  title = {From Parameter to Representation: A Closed-Form Approach for Controllable Model Merging},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {32},
  pages = {26903--26912},
  doi = {10.1609/aaai.v40i32.39902}
}
Kostenlos starten →