Don’t Start Over: A Cost-Effective Framework for Migrating Personalized Prompts Between LLMs
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (41) · 35003-35011
DOI: 10.1609/aaai.v40i41.40805 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Ziyi Zhao und 7 weiteren Autor:innen.
Abstract (Originalsprache)
Personalization in Large Language Models (LLMs) often relies on user-specific soft prompts. However, these prompts become obsolete when the foundation model is upgraded, necessitating costly, full-scale retraining. To overcome this limitation, we propose the Prompt-level User Migration Adapter (PUMA), a lightweight framework to efficiently migrate personalized prompts across incompatible models. PUMA utilizes a parameter-efficient adapter to bridge the semantic gap, combined with a group-based user selection strategy to significantly reduce training costs. Experiments on three large-scale datasets show our method matches or even surpasses the performance of retraining from scratch, reducing computational cost by up to 98%. The framework demonstrates strong generalization across diverse model architectures and robustness in advanced scenarios like chained and aggregated migrations, offering a practical path for the sustainable evolution of personalized AI by decoupling user assets from the underlying models.
Zitieren
APA
Zhao, Z., Gao, C., Zhang, Y., Liu, H., Gan, W., Guo, H., Liu, Y., & Feng, F. (2026). Don’t Start Over: A Cost-Effective Framework for Migrating Personalized Prompts Between LLMs. Proceedings of the AAAI Conference on Artificial Intelligence, 40(41), 35003-35011. https://doi.org/10.1609/aaai.v40i41.40805
Vancouver
Zhao Z, Gao C, Zhang Y, Liu H, Gan W, Guo H, et al. Don’t Start Over: A Cost-Effective Framework for Migrating Personalized Prompts Between LLMs. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(41):35003-35011. doi:10.1609/aaai.v40i41.40805
BibTeX
@article{Zhao2026,
author = {Zhao, Ziyi and Gao, Chongming and Zhang, Yang and Liu, Haoyan and Gan, Weinan and Guo, Huifeng and Liu, Yong and Feng, Fuli},
title = {Don’t Start Over: A Cost-Effective Framework for Migrating Personalized Prompts Between LLMs},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {41},
pages = {35003--35011},
doi = {10.1609/aaai.v40i41.40805}
}
EUVIMED