Preprint · 2021 · Open access
Large Language Models Can Be Strong Differentially Private Learners
arXiv (Cornell University) · Published 12 October 2021 · DOI 10.48550/arxiv.2110.05679
In brief
“Large Language Models Can Be Strong Differentially Private Learners” is a 2021 preprint by Xuechen Li and 3 co-authors, published in arXiv (Cornell University). OpenAlex counts 485 citations as of 19 November 2026. It is open access under a CC BY licence: the full text is free to read at arXiv (Cornell University).
- Citations
- 485
- Most cited in
- 2024 (128)
- References
- 96
- Authors
- 4
Abstract
Differentially Private (DP) learning has seen limited success for building large deep learning models of text, and straightforward attempts at applying Differentially Private Stochastic Gradient Descent (DP-SGD) to NLP tasks have resulted in large performance drops and high computational overhead. We show that this performance drop can be mitigated with (1) the use of large pretrained language models; (2) non-standard hyperparameters that suit DP optimization; and (3) fine-tuning objectives which are aligned with the pretraining procedure. With the above, we obtain NLP models that outperform state-of-the-art DP-trained models under the same privacy budget and strong non-private baselines -- by directly fine-tuning pretrained models with DP optimization on moderately-sized corpora. To address the computational challenge of running DP-SGD with large Transformers, we propose a memory saving technique that allows clipping in DP-SGD to run without instantiating per-example gradients for any linear layer in the model. The technique enables privately training Transformers with almost the same memory cost as non-private training at a modest run-time overhead. Contrary to conventional wisdom that DP optimization fails at learning high-dimensional models (due to noise that scales with dimension) empirical results reveal that private learning with pretrained language models doesn't tend to suffer from dimension-dependent performance degradation. Code to reproduce results can be found at https://github.com/lxuechen/private-transformers.
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Li, X., Tramèr, F., Liang, P., & Hashimoto, T. (2021). Large Language Models Can Be Strong Differentially Private Learners. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2110.05679
MLA 9
Li, Xuechen, et al. “Large Language Models Can Be Strong Differentially Private Learners.” arXiv (Cornell University), 2021. https://doi.org/10.48550/arxiv.2110.05679.
Chicago
Li, Xuechen, Florian Tramèr, Percy Liang, and Tatsunori Hashimoto. 2021. “Large Language Models Can Be Strong Differentially Private Learners.” arXiv (Cornell University). https://doi.org/10.48550/arxiv.2110.05679.
Harvard
Li, X. et al. (2021) ‘Large Language Models Can Be Strong Differentially Private Learners’, arXiv (Cornell University). doi:10.48550/arxiv.2110.05679.
IEEE
X. Li, F. Tramèr, P. Liang, and T. Hashimoto, “Large Language Models Can Be Strong Differentially Private Learners,” arXiv (Cornell University), 2021, doi: 10.48550/arxiv.2110.05679.
BibTeX
@misc{li2021large,
title = {{Large Language Models Can Be Strong Differentially Private Learners}},
author = {Li, Xuechen and Tramèr, Florian and Liang, Percy and Hashimoto, Tatsunori},
howpublished = {arXiv (Cornell University)},
year = {2021},
doi = {10.48550/arxiv.2110.05679},
url = {https://doi.org/10.48550/arxiv.2110.05679}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Citations by year
| Year | Citations | Share of the best year |
|---|---|---|
| 2021 | 5 | |
| 2022 | 43 | |
| 2023 | 82 | |
| 2024 | 128 | |
| 2025 | 122 | |
| 2026 | 105 |
Topic and keywords
Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- differential privacy
- large language models
- DP-SGD
- pre-trained language models
- fine-tuning
- transformer models
- privacy-preserving deep learning
- hyperparameter optimization
Work with this paper
Most-cited papers in Computer Science
- ImageNet classification with deep convolutional neural networksAlex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton · Communications of the ACM · 2017 · 114,726 citations
- Long Short-Term MemorySepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 109,097 citations
- A Mathematical Theory of CommunicationClaude E. Shannon · Bell System Technical Journal · 1948 · 91,736 citations
- Deep learningYann LeCun, Yoshua Bengio, Geoffrey E. Hinton · Nature · 2015 · 90,226 citations
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal NetworksShaoqing Ren, Kaiming He, Ross Girshick and 1 more · IEEE Transactions on Pattern Analysis and Machine Intelligence · 2016 · 74,143 citations
Questions about this paper
How do I cite this paper in APA style?
Li, X., Tramèr, F., Liang, P., & Hashimoto, T. (2021). Large Language Models Can Be Strong Differentially Private Learners. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2110.05679
How many times has this paper been cited?
485 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.
Is this paper free to read?
Yes. It is open access at arXiv (Cornell University): sign in and open the PDF from this page.
Who wrote this paper?
Xuechen Li, Florian Tramèr, Percy Liang, Tatsunori Hashimoto.
What is the DOI of this paper?
10.48550/arxiv.2110.05679, permanently at https://doi.org/10.48550/arxiv.2110.05679. Sign in and open it from this page.
Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.