REVIEW 26 cited by
Differentially Private Fine-tuning of Language Models
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Differentially Private Fine-tuning of Language Models
read the original abstract
We give simpler, sparser, and faster algorithms for differentially private fine-tuning of large-scale pre-trained language models, which achieve the state-of-the-art privacy versus utility tradeoffs on many standard NLP tasks. We propose a meta-framework for this problem, inspired by the recent success of highly parameter-efficient methods for fine-tuning. Our experiments show that differentially private adaptations of these approaches outperform previous private algorithms in three important dimensions: utility, privacy, and the computational and memory cost of private training. On many commonly studied datasets, the utility of private models approaches that of non-private models. For example, on the MNLI dataset we achieve an accuracy of $87.8\%$ using RoBERTa-Large and $83.5\%$ using RoBERTa-Base with a privacy budget of $\epsilon = 6.7$. In comparison, absent privacy constraints, RoBERTa-Large achieves an accuracy of $90.2\%$. Our findings are similar for natural language generation tasks. Privately fine-tuning with DART, GPT-2-Small, GPT-2-Medium, GPT-2-Large, and GPT-2-XL achieve BLEU scores of 38.5, 42.0, 43.1, and 43.8 respectively (privacy budget of $\epsilon = 6.8,\delta=$ 1e-5) whereas the non-private baseline is $48.1$. All our experiments suggest that larger models are better suited for private fine-tuning: while they are well known to achieve superior accuracy non-privately, we find that they also better maintain their accuracy when privacy is introduced.
Forward citations
Cited by 26 Pith papers
-
Private and Stable Test-Time Adaptation with Differential Privacy
Differential privacy versions of TTA methods achieve privacy on ImageNet-C with small accuracy cost and can improve stability via clipping in continual settings.
-
Unveiling Privacy Risks in Multi-modal Large Language Models: Task-specific Vulnerabilities and Mitigation Challenges
Introduces MM-Privacy dataset and evaluations showing MLLMs leak sensitive data from images in various tasks, highlighting task inconsistency effects.
-
Private and Stable Test-Time Adaptation with Differential Privacy
Casting Tent, EATA, SAR, DeYO, and COME into DP-TTA via per-sample clipping and Gaussian noise yields adequate privacy on ImageNet-C at modest accuracy and compute cost, with clipping sometimes improving stability.
-
DP-SelFT: Differentially Private Selective Fine-Tuning for Large Language Models
DP-SelFT improves the privacy-utility trade-off for LLM fine-tuning by selecting robust layer subsets via DP synthetic data and perturbation-matched evaluation.
-
Probing Privacy Leaks in LLM-based Code Generation via Test Generation
A test-driven pipeline with an auto-constructed privacy feature library detects 2.56 times more confirmed privacy leaks in LLM-based code generation than existing baselines.
-
Beyond Individual Mimicry: Constructing Human-Like Social network with Graph-Augmented LLM Agents
GraphMind equips LLM agents with graph awareness to construct human-like social networks, producing botnets that substantially degrade performance of both text-based and graph-based detectors.
-
GroupGPT: A Token-efficient and Privacy-preserving Agentic Framework for Multi-User Chat Assistant
GroupGPT decouples intervention timing from response generation via edge-cloud collaboration for multi-user chats, scoring 4.72/5 on the new MUIR benchmark of 2500 segments while cutting token use by up to 3x and addi...
-
In-Context Probing for Membership Inference in Fine-Tuned Language Models
ICP-MIA infers membership in fine-tuned LLMs by measuring confidence improvement under in-context probes, beating prior black-box attacks at low false-positive rates.
-
Forget What's Sensitive, Remember What Matters: Token-Level Differential Privacy in Memory Sculpting for Continual Learning
PeCL applies token-level dynamic differential privacy and privacy-guided memory sculpting to achieve superior privacy-utility balance in continual learning.
-
Memory-Efficient Differentially Private Training with Gradient Random Projection
DP-GRAPE reduces memory in differentially private neural network training by using random Gaussian projections on gradients instead of SVD, achieving comparable privacy-utility tradeoffs to DP-SGD and scaling to 6.7B ...
-
ConfusionPrompt: Practical Private Inference for Online Large Language Models
ConfusionPrompt enables private black-box LLM inference via prompt decomposition and pseudo-prompt mixing, claiming better privacy-utility trade-off than perturbation methods and lower memory use than open-source loca...
-
The False Promise of Imitating Proprietary LLMs
Finetuning open LMs on ChatGPT outputs creates models that mimic style and fool human raters but fail to close the performance gap to proprietary systems on tasks not well-represented in the imitation data.
-
PeerCheck: Enhancing LLM-Generated Academic Reviews Towards Human-Level Quality
PeerCheck finds that chain-of-thought prompting improves LLM academic reviews while retrieval-augmented generation sometimes lowers quality, and that LLMs and humans emphasize different aspects of papers.
-
SharedRequest: Privacy-Preserving Model-Agnostic Inference for Large Language Models
SharedRequest is a model-agnostic batch-level framework that mixes prompts with noise and groups equivalent instructions to achieve higher utility and lower query cost than individual differential privacy methods for ...
-
Membership Inference Attacks on Tokenizers of Large Language Models
Tokenizers leak dataset membership: using distinctive tokens that appear in a dataset's text, an attacker can detect with AUC up to 0.77 whether that dataset was part of tokenizer training.
-
Re-examining Low Rank adaptation for private LLM fine-tuning
Denoising DP-SGD gradients with random-matrix-theory singular-value shrinkage speeds up private RoBERTa fine-tuning on GLUE, cutting steps to target accuracy by 20–100%.
-
Differentially-private text generation degrades output language quality
DP fine-tuning systematically degrades LLM output length, grammatical correctness, and lexical diversity, and this degradation grows as the privacy budget shrinks.
-
Balancing Utility and Privacy: Dynamically Private SGD with Random Projection
D2P2-SGD combines time-decreasing privacy noise with random projection to improve the accuracy of differentially private SGD, with convergence rates matching ordinary SGD.
-
FedShield-LLM: A Secure and Scalable Federated Fine-Tuned Large Language Model
FedShield-LLM integrates pruning and FHE on LoRA parameters to support secure, scalable federated fine-tuning of LLMs such as Llama-2.
-
Public Data Assisted Differentially Private In-Context Learning
A private ICL algorithm that aggregates LLM responses with DPM clustering and uses public data representatives achieves near-non-private utility at epsilon=1.
-
Towards the Anonymization of the Language Modeling
Authors introduce MLM and CLM specialization methods that avoid memorizing identifiers in sensitive training data while aiming for a privacy-utility tradeoff on medical datasets.
-
Industry Practitioners Perspectives on AI Model Quality: Perceptions, Challenges, and Solutions
Industry AI practitioners view model quality through nine attributes with context-dependent priorities, where data imbalance is a key challenge addressed by strategies like active learning, as confirmed by interviews ...
-
$\pi$-RAG: Oblivious Retrieval via Semantic Quantization and Transcendental Addressing for Large Language Models
π-RAG uses π digits for transcendental addressing and projects queries onto Canonical Intent Centroids mapped via cryptographic salt to produce oblivious π-keys that point to data without exposing embeddings.
-
Low-Rank Adaptation Redux for Large Models
An overview revisits LoRA variants by categorizing advances in architectural design, efficient optimization, and applications while linking them to classical signal processing tools for principled fine-tuning.
-
SLM Finetuning for Natural Language to Domain Specific Code Generation in Production
Fine-tuned small language models outperform larger models in natural language to domain-specific code generation with improved performance, latency, and the ability to adapt to customer-specific scenarios without losi...
-
How to DP-fy Your Data: A Practical Guide to Generating Synthetic Data With Differential Privacy
A practical, extremely thorough survey of differentially private synthetic data generation: methods, privacy units, evaluation metrics, and end-to-end system components across four data modalities.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.