# Re-examining Low Rank adaptation for private LLM fine-tuning

arXiv: https://arxiv.org/abs/2510.01137 · HTML: https://mlsanity.com/paper/2510.01137/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 39 references that are in the map.
- First posted on arXiv 1 Oct 2025.

## Builds on (most cited first, 25 of 25)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://mlsanity.com/paper/1804.07461/) (2018)
- [Deep Learning with Differential Privacy](https://mlsanity.com/paper/1607.00133/) (2016)
- [Neural Tangent Kernel: Convergence and Generalization in Neural Networks](https://mlsanity.com/paper/1806.07572/) (2018)
- [Extracting Training Data from Large Language Models](https://mlsanity.com/paper/2012.07805/) (2020)
- [Membership Inference Attacks From First Principles](https://mlsanity.com/paper/2112.03570/) (2021)
- [Wide Neural Networks of Any Depth Evolve as Linear Models Under Gradient Descent](https://mlsanity.com/paper/1902.06720/) (2019)
- [A Mean Field View of the Landscape of Two-Layers Neural Networks](https://mlsanity.com/paper/1804.06561/) (2018)
- [GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection](https://mlsanity.com/paper/2403.03507/) (2024)
- [Large Language Models Can Be Strong Differentially Private Learners](https://mlsanity.com/paper/2110.05679/) (2021)
- [Differentially Private Fine-tuning of Language Models](https://mlsanity.com/paper/2110.06500/) (2021)
- [The E2E Dataset: New Challenges For End-to-End Generation](https://mlsanity.com/paper/1706.09254/) (2017)
- [Improving the Gaussian Mechanism for Differential Privacy: Analytical Calibration and Optimal Denoising](https://mlsanity.com/paper/1805.06530/) (2018)
- [Adversary Instantiation: Lower Bounds for Differentially Private Machine Learning](https://mlsanity.com/paper/2101.04535/) (2021)
- [Numerical Composition of Differential Privacy](https://mlsanity.com/paper/2106.02848/) (2021)
- [DART: Open-Domain Structured Data Record to Text Generation](https://mlsanity.com/paper/2007.02871/) (2020)
- [Differentially Private Optimization on Large Model at Small Cost](https://mlsanity.com/paper/2210.00038/) (2022)
- [Scalable and Efficient Training of Large Convolutional Neural Networks with Differential Privacy](https://mlsanity.com/paper/2205.10683/) (2022)
- [When Does Differentially Private Learning Not Suffer in High Dimensions?](https://mlsanity.com/paper/2207.00160/) (2022)
- [Differentially Private Bias-Term Fine-tuning of Foundation Models](https://mlsanity.com/paper/2210.00036/) (2022)
- [On the Convergence and Calibration of Deep Learning with Differential Privacy](https://mlsanity.com/paper/2106.07830/) (2021)
- [u-μP: The Unit-Scaled Maximal Update Parametrization](https://mlsanity.com/paper/2407.17465/) (2024)
- [DiSK: Differentially Private Optimizer with Simplified Kalman Filter for Noise Reduction](https://mlsanity.com/paper/2410.03883/) (2024)
- [DOPPLER: Differentially Private Optimizers with Low-pass Filter for Privacy Noise Reduction](https://mlsanity.com/paper/2408.13460/) (2024)

## Abstract (arXiv)

Privacy is a central concern when fine-tuning large language models (LLMs) on sensitive data, and differentially private stochastic gradient descent (DP-SGD) -- which clips per-sample gradients and adds calibrated Gaussian noise -- is the standard tool for formal privacy guarantees. Both theory and practice show that lower-rank models are better suited to DP training, a property especially relevant for LLMs, whose fine-tuning gradients exhibit a strong low-rank structure. Methods such as DP-LoRA exploit this by restricting updates to a low-rank subspace, i.e., retaining only a few non-zero components in the SVD of each layer's gradient. However, we argue that while having few non-zero components is important, the isotropic noise injected by DP-SGD inflates the singular values of the gradient matrix, disrupting their naturally fast decay. In this work, we investigate whether this noise-induced eigenvalue blow-up reduces performance, and show that partially restoring the original singular-value profile significantly improves the sample efficiency of DP-SGD. Experiments on language classification (GLUE benchmark with RoBERTa) and text generation (E2E and DART table-to-text benchmarks with Qwen and Llama models up to 4B parameters) showcase that restoring the fast decay of singular values is a viable strategy for speeding up the DP optimization process, without compromising privacy guarantees.