Challenge: Training models with differential privacy has received a lot of attention since it provides theoretical guarantee of privacy preservation.
Approach: They propose methods that fine-tune large-scale pre-trained models with freezing unimportant parameters for downstream tasks while satisfying differential privacy.
Outcome: The proposed methods fine-tune large pre-trained models with freezing unimportant parameters while satisfying differential privacy while preserving their utility.

Similar Papers

One size does not fit all: Investigating strategies for differentially-private learning across NLP tasks (2022.emnlp-main)

Copied to clipboard

Challenge: Existing research on the efficiency of differentially-private stochastic gradient descent (DP-SGD) in NLP is inconclusive or even counter-intuitive.
Approach: They propose to use differentially-private stochastic gradient descent (DP-SGD) to preserve privacy in NLP by using modern neural models based on BERT and XtremeDistil architectures to conduct extensive experiments.
Outcome: The proposed models and training strategies provide the best trade-off between privacy and performance on different NLP tasks.
Fine-Tuning Language Models with Differential Privacy through Adaptive Noise Allocation (2024.findings-emnlp)

Copied to clipboard

Challenge: Language models can memorize detailed information and patterns, but raise privacy concerns . ANADP reduces the performance gap between regular and DP fine-tuning while maintaining the privacy constraints.
Approach: They propose an algorithm that allocates additive noise based on the importance of model parameters to reduce the performance gap between regular fine-tuning and traditional DP fine- tuning.
Outcome: The proposed algorithm narrows the performance gap between regular fine-tuning and traditional DP fine- tuning while maintaining privacy constraints.
Training Text-to-Text Transformers with Privacy Guarantees (2022.findings-acl)

Copied to clipboard

Challenge: Recent advances in NLP often stem from large transformer-based pre-trained models.
Approach: They propose differentially private (DP) training as a potential mitigation for models that can memorize parts of training data.
Outcome: The proposed model can memorize parts of training data and mitigate memorization concerns.
Just Fine-tune Twice: Selective Differential Privacy for Large Language Models (2022.emnlp-main)

Copied to clipboard

Challenge: Existing approaches to protect language models from privacy leakage suffer from limited user control and low utility . et al., 2018: a novel framework that achieves SDP for state-of-the-art large transformer-based models.
Approach: They propose a framework that applies differential privacy to large language models . they use redacted in-domain data to fine-tune the model with original in- domain data .
Outcome: The proposed framework achieves strong utility compared to baselines.
Evaluating Parameter-Efficient Finetuning Approaches for Pre-trained Models on the Financial Domain (2023.findings-emnlp)

Copied to clipboard

Challenge: Large-scale language models with millions, billions, or trillions of trainable parameters are becoming increasingly popular.
Approach: They compare performance of financial BERT-like models to their fully fine-tuned counterparts by using parameter-efficient tuning methods.
Outcome: The proposed approaches match full fine-tuning performance on common NLP tasks, but are less studied in finance.
Differentially Private Learning Needs Better Model Initialization and Self-Distillation (2025.naacl-long)

Copied to clipboard

Challenge: Differentially private SGD (DPSGD) enables privacy-preserving training of language models, but often reduces utility, diversity, and linguistic quality.
Approach: They propose a method that initializes a model using data synthesis from a small pre-trained LM with rigorous filtering, applies DP finetuning on private data, and performs self-distillation to refine outputs.
Outcome: The proposed method outperforms vanilla DPSGD with significant improvements in lexical diversity and grammar errors.
Not All Adapters Matter: Selective Adapter Freezing for Memory-Efficient Fine-Tuning of Language Models (2025.naacl-long)

Copied to clipboard

Challenge: Existing methods for fine-tuning pre-trained models impose substantial resource usage.
Approach: They propose a parameter-efficient fine-tuning method that freezes adapters early to reduce resource usage while maintaining performance.
Outcome: The proposed method reduces memory usage, computation amount, and training time by 42.85%, 34.59%, and 11.82% while maintaining performance.
How reparametrization trick broke differentially-private text representation learning (2022.acl-short)

Copied to clipboard

Challenge: Differential privacy (DP) is a formal mathematical treatment of privacy protection . it guarantees how much privacy can be lost in the worst case . adapting DP mechanisms to NLP properly is largely non-trivial task .
Approach: They propose to use differential privacy to learn text representations using DPText to quantify and guarantee how much privacy can be lost in the worst case.
Outcome: The proposed methods are falsely claimed to be differentially private and violate privacy loss guarantees.
Differentially Private Knowledge Distillation via Synthetic Text Generation (2024.findings-acl)

Copied to clipboard

Challenge: Large Language models (LLMs) are achieving state-of-the-art performance in many downstream tasks, but data privacy is a major challenge for practitioners.
Approach: They propose a differentially private knowledge distillation algorithm that exploits the knowledge of a teacher LLM and a student's output distribution.
Outcome: The proposed algorithm significantly improves the utility over baselines on the Big Patent dataset, with strong privacy parameters, =2.
Synthetic Text Generation with Differential Privacy: A Simple and Practical Recipe (2023.acl-long)

Copied to clipboard

Challenge: Privacy concerns have increased in data-driven products due to the tendency of machine learning models to memorize sensitive training data.
Approach: They propose a method for generating useful synthetic text with a formal privacy guarantee by fine-tuning a pretrained generative language model with DP.
Outcome: The proposed method produces synthetic text competitive in terms of utility with its non-private counterpart, while providing strong protection against potential privacy leakages.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations