Styles + Persona-plug = Customized LLMs
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915719920222208 |
|---|---|
| author | Song, Yutong Wu, Jiang Yuan, Shaofan Shen, Chengze Wang, Jian Rahmani, Amir Dutt, Nikil Wang, Yu |
| author_facet | Song, Yutong Wu, Jiang Yuan, Shaofan Shen, Chengze Wang, Jian Rahmani, Amir Dutt, Nikil Wang, Yu |
| contents | We discover a previously overlooked challenge in personalized text generation: personalization methods are increasingly applied under explicit style instructions, yet their behavior under such constraints remains poorly understood. To balance implicit personalization and explicit style, we formulate personalization as a distributional residual and propose PsPLUG, a lightweight soft-prompt plug-in trained with style-conditioned preference contrasts. Across LaMP benchmark, our framework improves persona alignment, maintains stylistic fidelity, and outperforms retrieval-based and soft-prompt baselines with minimal computation. These results show that residual modeling provides a simple and principled foundation for controllable, style-aware LLM personalization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_06362 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Styles + Persona-plug = Customized LLMs Song, Yutong Wu, Jiang Yuan, Shaofan Shen, Chengze Wang, Jian Rahmani, Amir Dutt, Nikil Wang, Yu Artificial Intelligence We discover a previously overlooked challenge in personalized text generation: personalization methods are increasingly applied under explicit style instructions, yet their behavior under such constraints remains poorly understood. To balance implicit personalization and explicit style, we formulate personalization as a distributional residual and propose PsPLUG, a lightweight soft-prompt plug-in trained with style-conditioned preference contrasts. Across LaMP benchmark, our framework improves persona alignment, maintains stylistic fidelity, and outperforms retrieval-based and soft-prompt baselines with minimal computation. These results show that residual modeling provides a simple and principled foundation for controllable, style-aware LLM personalization. |
| title | Styles + Persona-plug = Customized LLMs |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2601.06362 |