Elamur Research⌕
☰
home / papers / 2026 / PRISP: Privacy-Safe Few-Shot Personalization via Lightweight Adaptation

PRISP: Privacy-Safe Few-Shot Personalization via Lightweight Adaptation

Paper · 2026 · Junho Park, Dohoon Kim and Taesup Moon

Summary

PRISP addresses LLM personalization as it happens after deployment. In that setting each user supplies only a few examples, adaptation must run on modest (often edge) hardware, and one user's data or adapted weights must never reach another. Earlier methods such as OPPU, Per-Pcs and PriME either depend on plentiful data or reuse LoRA modules trained for other users, which the authors treat as a privacy hazard.

PRISP needs no task data. A pre-trained Text-to-LoRA hypernetwork converts a plain-language task description into an "anchor" LoRA. For each user, the input-side LoRA matrices stay frozen. Only the output-side matrices and a small per-layer bridge matrix are trained, using that user's latest history for one epoch.

The authors built a few-shot version of the LaMP benchmark and tested mainly with Qwen3-0.6B. There, PRISP shows strong overall results against earlier approaches, lowers computational cost and shares nothing across users. This makes it a practical option for private, on-device personalization.

Key points

  • Two-stage method:
    • Stage 1: given nothing but a task description, a Text-to-LoRA hypernetwork produces an anchor LoRA tailored to that task.
    • Stage 2: the A matrices are frozen, and a bridge matrix C (initialized as identity) is trained together with the B matrices. The update takes the form B·C·A, and the design draws on DoMIX.
  • Few-shot LaMP setup:
    • It covers six tasks: LaMP-1, 2, 3, 4, 5 and 7 (three classification, three generation).
    • Baselines receive 50 task samples, on average 0.67% of the full task data. Personalization uses the 10 most recent items per user, about 10.4% of a user's history.
    • PRISP uses zero task data, and training is limited to a single epoch.
  • Models: Qwen3-0.6B is the main backbone, and Llama-3.1-8B-Instruct is reported in an appendix.
  • Transfer to unseen tasks: methods built on sharer LoRAs keep only about 50–60% "adaptability" when moved to a new task. PRISP counts as 100% because it generates task LoRAs zero-shot from descriptions.
  • Robustness and limitations:
    • On LaMP-2, different task-description variants give different scores before personalization, but they converge to similar results after Stage 2.
    • The authors note that continual personalization is not covered and that results depend on how expressive the hypernetwork is.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Seoul National UniversitySeoul · South Korea41 works · 45,671 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

Cite

Cite as Junho Park, Dohoon Kim and Taesup Moon. 2026. PRISP: Privacy-Safe Few-Shot Personalization via Lightweight Adaptation. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1146

@inproceedings{park2026prisp,
  title = {{PRISP: Privacy-Safe Few-Shot Personalization via Lightweight Adaptation}},
  author = {Junho Park, Dohoon Kim and Taesup Moon},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1146},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1146}
}

References

36 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, and… find
  2. Nicholas Carlini, Daphne Ippolito, Matthew Jagielski, Katherine Lee, Florian Tramer, and Chiyuan Zhang. 2023. Quantifying memorization across neural language models. In The Eleventh International Conference on Learning Representations (ICLR). find
  3. Nicholas Carlini, Chang Liu, Úlfar Erlingsson, Jernej Kos, and Dawn Song. 2019. The secret sharer: Evaluating and testing unintended memorization in neural networks. In 28th USENIX Security Symposium (USENIX Security 19), pages 267–284, Santa Clara, CA. USENIX Association.
  4. Nicholas Carlini, Florian Tramèr, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Úlfar Erlingsson, Alina Oprea, and Colin Raffel. 2021. Extracting training data from large language models. In 30th USENIX Security Symposium (USENIX Security 21)…
  5. Rujikorn Charakorn, Edoardo Cetin, Yujin Tang, and Robert Tjarko Lange. 2025. Text-to-loRA: Instant transformer adaption. In Forty-second International Conference on Machine Learning (ICML). find
  6. Jin Chen, Zheng Liu, Xu Huang, Chenwang Wu, Qi Liu, Gangwei Jiang, Yuanhao Pu, Yuxuan Lei, Xiaolong Chen, Xingmei Wang, Kai Zheng, Defu Lian, and Enhong Chen. 2024. When large language models meet personalization: perspectives of challenges and opportunities. World Wide Web (WWW), 27(4):42. find
  7. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, and 542 o… — The Llama 3 Herd of Models
  8. Junxian He, Chunting Zhou, Xuezhe Ma, Taylor Berg-Kirkpatrick, and Graham Neubig. 2022. Towards a unified view of parameter-efficient transfer learning. In International Conference on Learning Representations (ICLR).
  9. Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. 2019. Parameter-efficient transfer learning for NLP. In International Conference on Machine Learning (ICML), pages 2790–2799. PMLR. find
  10. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-rank adaptation of large language models. In International Conference on Learning Representations (ICLR).
  11. Huseyin A. Inan, Osman Ramadan, Lukas Wutschitz, Daniel Jones, Victor Rühle, James Withers, and Robert Sim. 2021. Training data leakage analysis in language models. Preprint, arXiv:2101.05405. — [arXiv:2101.05405](https://arxiv.org/abs/2101.05405) find
  12. Gautier Izacard, Patrick Lewis, Maria Lomeli, Lucas Hosseini, Fabio Petroni, Timo Schick, Jane Dwivedi-Yu, Armand Joulin, Sebastian Riedel, and Edouard Grave. 2022. Atlas: Few-shot learning with retrieval augmented language models. Preprint, arXiv:2208.03299.
  13. Dohoon Kim, Donghun Kang, and Taesup Moon. 2025a. DoMIX: An efficient framework for exploiting domain knowledge in fine-tuning. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL) (Volume 1: Long Papers), pages 14581–14602, Vienna, Austria. Association fo… find
  14. Jaehyung Kim and Yiming Yang. 2025. Few-shot personalization of LLMs with mis-aligned responses. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL): Human Language Technologies (Volume 1: Long Papers), pages 11943– 11… find
  15. Kyuyoung Kim, Jinwoo Shin, and Jaehyung Kim. 2025b. Personalized language models via privacy-preserving evolutionary model merging. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 34453–34468, Suzhou, China. Association for Computational Lingu… find
  16. Ishita Kumar, Snigdha Viswanathan, Sushrita Yerra, Alireza Salemi, Ryan A Rossi, Franck Dernoncourt, Hanieh Deilamsalehy, Xiang Chen, Ruiyi Zhang, Shubham Agarwal, and 1 others. 2024. Longlamp: A benchmark for personalized long-form text generation. Preprint, arXiv:2407.11016. — [arXiv:2407.11016](https://arxiv.org/abs/2407.11016) find
  17. Brian Lester, Rami Al-Rfou, and Noah Constant. 2021. The power of scale for parameter-efficient prompt tuning. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 3045–3059, Online and Punta Cana, Dominican Republic. Association for Computational…
  18. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, and 1 others. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. In Advances in neural information processing systems (NeurI…
  19. Xiang Lisa Li and Percy Liang. 2021. Prefix-tuning: Optimizing continuous prompts for generation. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (ACL) (Volume 1: Long Papers), pages… find
  20. Jiahong Liu, Zexuan Qiu, Zhongyang Li, Quanyu Dai, Wenhao Yu, Jieming Zhu, Minda Hu, Menglin Yang, Tat-Seng Chua, and Irwin King. 2025. A survey of personalized large language models: Progress and future directions. Preprint, arXiv:2502.11528. — [arXiv:2502.11528](https://arxiv.org/abs/2502.11528) find
  21. Fanxu Meng, Zhaohui Wang, and Muhan Zhang. 2024. Pissa: Principal singular values and singular vectors adaptation of large language models. In Advances in Neural Information Processing Systems (NeurIPS), volume 37, pages 121038–121072. Curran Associates, Inc. find
  22. Sheshera Mysore, Zhuoran Lu, Mengting Wan, Longqi Yang, Bahareh Sarrafzadeh, Steve Menezes, Tina Baghaee, Emmanuel Barajas Gonzalez, Jennifer Neville, and Tara Safavi. 2024. Pearl: Personalizing large language model writing assistants with generation-calibrated retrievers. In Proceedings of the 1st… find
  23. Milad Nasr, Javier Rando, Nicholas Carlini, Jonathan Hayase, Matthew Jagielski, A. Feder Cooper, Daphne Ippolito, Christopher A. Choquette-Choo, Florian Tramèr, and Katherine Lee. 2025. Scalable extraction of training data from aligned, production language models. In The Thirteenth International Con… find
  24. Jonas Pfeiffer, Ivan Vulić, Iryna Gurevych, and Sebastian Ruder. 2020. MAD-X: An Adapter-Based Framework for Multi-Task Cross-Lingual Transfer. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7654–7673, Online. Association for Computational L…
  25. Chris Richardson, Yao Zhang, Kellen Gillespie, Sudipta Kar, Arshdeep Singh, Zeynab Raeesy, Omar Zia Khan, and Abhinav Sethy. 2023. Integrating summarization and retrieval for enhanced personalization via large language models. Preprint, arXiv:2310.20081. — [arXiv:2310.20081](https://arxiv.org/abs/2310.20081) find
  26. Alireza Salemi, Sheshera Mysore, Michael Bendersky, and Hamed Zamani. 2024. LaMP: When large language models meet personalization. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (ACL) (Volume 1: Long Papers), pages 7370–7392, Bangkok, Thailand. Association…
  27. Alireza Salemi and Hamed Zamani. 2025. LaMP-QA: A benchmark for personalized long-form question answering. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1139–1159, Suzhou, China. Association for Computational Linguistics.
  28. Zhaoxuan Tan, Zheyuan Liu, and Meng Jiang. 2024a. Personalized pieces: Efficient personalized large language models through collaborative efforts. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6459–6475, Miami, Florida, USA. Association for…
  29. Zhaoxuan Tan, Qingkai Zeng, Yijun Tian, Zheyuan Liu, Bing Yin, and Meng Jiang. 2024b. Democratizing large language models via personalized parameter-efficient fine-tuning. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6476–6491, Miami, Flori…
  30. Gemma Team, Thomas Mesnard, Cassidy Hardin, Robert Dadashi, Surya Bhupatiraju, Shreya Pathak, Laurent Sifre, Morgane Rivière, Mihir Sanjay Kale, Juliette Love, and 1 others. 2024. Gemma: Open models based on gemini research and technology. Preprint, arXiv:2403.08295. find
  31. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, and 1 others. 2023. LLaMA: Open and efficient foundation language models. Preprint, arXiv:2302.13971.
  32. Andrew Trotman, Antti Puurula, and Blake Burgess. 2014. Improvements to bm25 and language models examined. In Proceedings of the 19th Australasian Document Computing Symposium (ADCS), page 58–65, New York, NY, USA. Association for Computing Machinery. find
  33. Shaowen Wang, Linxi Yu, and Jian Li. 2024. Lora-ga: Low-rank adaptation with gradient approximation. In Advances in Neural Information Processing Systems (NeurIPS), volume 37, pages 54905–54931. Curran Associates, Inc. find
  34. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. Preprint, arXiv:2505.09388.
  35. Linhai Zhang, Jialong Wu, Deyu Zhou, and Yulan He. 2025a. PROPER: A progressive learning framework for personalized large language models with group-level adaptation. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL) (Volume 1: Long Papers), pages 16399…
  36. Zhehao Zhang, Ryan A. Rossi, Branislav Kveton, Yijia Shao, Diyi Yang, Hamed Zamani, Franck Dernoncourt, Joe Barrow, Tong Yu, Sungchul Kim, Ruiyi Zhang, Jiuxiang Gu, Tyler Derr, Hongjie Chen, Junda Wu, Xiang Chen, Zichao Wang, Subrata Mitra, Nedim Lipka, and 2 others. 2025b. Personalization of large… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/park-2026-prisp-privacy-safe-few-shot/