mlsanityData to 14 Oct 2026

PrE-Text: Training Language Models on Private Federated Data in the Age of LLMs

arXiv 2406.02958 ↗First posted 5 Jun 2024cs.LG cs.AI cs.CLcited by 29

Not on the map. Why?

Abstract (arXiv)

On-device training is currently the most common approach for training machine learning (ML) models on private, distributed user data. Despite this, on-device training has several drawbacks: (1) most user devices are too small to train large models on-device, (2) on-device training is communication- and computation-intensive, and (3) on-device training can be difficult to debug and deploy. To address these problems, we propose Private Evolution-Text (PrE-Text), a method for generating differentially private (DP) synthetic textual data. First, we show that across multiple datasets, training small models (models that fit on user devices) with PrE-Text synthetic data outperforms small models trained on-device under practical privacy regimes (ε=1.29, ε=7.58). We achieve these results while using 9× fewer rounds, 6× less client computation per round, and 100× less communication per round. Second, finetuning large models on PrE-Text's DP synthetic data improves large language model (LLM) performance on private data across the same range of privacy budgets. Altogether, these results suggest that training on DP synthetic data can be a better option than training a model on-device on private distributed data. Code is available at https://github.com/houcharlie/PrE-Text.

Key facts. Cited by 29 papers in the map (data through 14 Oct 2026). Builds on 102 of its 130 references that are in the map. First posted on arXiv 5 Jun 2024.

Builds on

102 of 130 references.

References · 102

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  2. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,599
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  4. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
  5. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  6. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,669
  7. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  8. 2019BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehensioncited by 5,893
  9. 2019DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lightercited by 4,307
  10. 2019Advances and Open Problems in Federated Learningcited by 3,046
  11. 2018Federated Optimization in Heterogeneous Networkscited by 2,987
  12. 2016Deep Learning with Differential Privacycited by 2,728
38 more
  1. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  2. 2023Code Llama: Open Foundation Models for Codecited by 2,139
  3. 2020Tackling the Objective Inconsistency Problem in Heterogeneous Federated Optimizationcited by 906
  4. 2020Ensemble Distillation for Robust Model Fusion in Federated Learningcited by 682
  5. 2023Textbooks Are All You Needcited by 569
  6. 2020Dataset Condensation with Gradient Matchingcited by 516
  7. 2022Dataset Distillation by Matching Training Trajectoriescited by 442
  8. 2022Knowledge Unlearning for Mitigating Privacy Risks in Language Modelscited by 437
  9. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 412
  10. 2021Differentially Private Fine-tuning of Language Modelscited by 400
  11. 2022Unnatural Instructions: Tuning Language Models with (Almost) No Human Laborcited by 397
  12. 2021Opacus: User-Friendly Differential Privacy Library in PyTorchcited by 383
  13. 2016Semi-supervised Knowledge Transfer for Deep Learning from Private Training Datacited by 348
  14. 2021A Field Guide to Federated Optimizationcited by 315
  15. 2018Scalable Private Learning with PATEcited by 269
  16. 2020Differentially Private Learning Needs Better Features (or Much More Data)cited by 235
  17. 2023Unlearn What You Want to Forget: Efficient Unlearning for LLMscited by 215
  18. 2022What Does it Mean for a Language Model to Preserve Privacy?cited by 195
  19. 2021Federated Learning with Buffered Asynchronous Aggregationcited by 190
  20. 2021FedKD: Communication Efficient Federated Learning via Knowledge Distillationcited by 187
  21. 2023Towards Building the Federated GPT: Federated Instruction Tuningcited by 185
  22. 2016Practical Secure Aggregation for Federated Learning on User-Held Datacited by 178
  23. 2021Practical and Private (Deep) Learning without Sampling or Shufflingcited by 170
  24. 2018cpSGD: Communication-efficient and differentially-private distributed SGDcited by 158
  25. 2023How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacycited by 153
  26. 2022ProxSkip: Yes! Local Gradient Steps Provably Lead to Communication Acceleration! Finally!cited by 134
  27. 2022PromptFL: Let Federated Participants Cooperatively Learn Prompts Instead of Models -- Federated Learning in Age of Foundation Modelcited by 134
  28. 2021The Distributed Discrete Gaussian Mechanism for Federated Learning with Secure Aggregationcited by 128
  29. 2022Differentially Private Diffusion Modelscited by 119
  30. 2024Machine Unlearning of Pre-trained Large Language Modelscited by 119
  31. 2020DP-CGAN: Differentially Private Synthetic Data and Label Generationcited by 118
  32. 2021Large-Scale Differentially Private BERTcited by 118
  33. 2022Synthetic Text Generation with Differential Privacy: A Simple and Practical Recipecited by 108
  34. 2023Stochastic Controlled Averaging for Federated Learning with Communication Compressioncited by 101
  35. 2023Privacy-Preserving In-Context Learning with Differentially Private Few-Shot Generationcited by 96
  36. 2020Bypassing the Ambient Dimension: Private SGD with Gradient Subspace Identificationcited by 90
  37. 2021Large Scale Private Learning via Low-rank Reparametrizationcited by 88
  38. 2022Automatic Clipping: Differentially Private Deep Learning Made Easier and Strongercited by 88

50 most cited of 102 All 102 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 2; 2025: 16; 2026: 11.

29 papers.

Sort:

By area:

  1. 2025Synthesizing Privacy-Preserving Text Data via Finetuning without Finetuning Billion-Scale LLMscited by 16
  2. 2025POPri: Private Federated Learning using Preference-Optimized Synthetic Datacited by 14
  3. 2025Differentially Private Synthetic Data via APIs 3: Using Simulators Instead of Foundation Modelcited by 13
  4. 2025Struct-Bench: A Benchmark for Differentially Private Structured Text Generationcited by 7
  5. 2025Is API Access to LLMs Useful for Generating Private Synthetic Tabular Data?cited by 5
  6. 2025FLoRIST: Singular Value Thresholding for Efficient and Accurate Federated Fine-Tuning of Large Language Modelscited by 4
  7. 2024FedSpaLLM: Federated Pruning of Large Language Modelscited by 3
  8. 2025RewardDS: Privacy-Preserving Fine-Tuning for Large Language Models via Reward Driven Data Synthesiscited by 3
  9. 2025MaskSQL: Safeguarding Privacy for LLM-Based Text-to-SQL via Abstractioncited by 3
  10. 2025Research in Collaborative Learning Does Not Serve Cross-Silo Federated Learning in Practicecited by 3
  11. 2025ACTG-ARL: Differentially Private Conditional Text Generation with RL-Boosted Controlcited by 3
  12. 2025Synthesizing and Adapting Error Correction Data for Mobile Large Language Model Applicationscited by 2
17 more
  1. 2025Private Evolution Convergescited by 2
  2. 2026Private Seeds, Public LLMs: Realistic and Privacy-Preserving Synthetic Data Generationcited by 2
  3. 2026Differentially Private Synthetic Data via APIs 4: Tabular Datacited by 2
  4. 2024LanFL: Differentially Private Federated Learning with Large Language Models using Synthetic Samplescited by 1
  5. 2025FedRW: Efficient Privacy-Preserving Data Reweighting for Enhancing Federated Learning of Language Modelscited by 1
  6. 2026Differentially Private Preference Data Synthesis for Large Language Model Alignmentcited by 1
  7. 2025FedTDP: A Privacy-Preserving and Unified Framework for Trajectory Data Preparation via Federated Learningcited by 0
  8. 2025LLM-QFL: Distilling Large Language Model for Quantum Federated Learningcited by 0
  9. 2025Private Training & Data Generation by Clustering Embeddingscited by 0
  10. 2026Resource-Adaptive Federated Text Generation with Differential Privacycited by 0
  11. 2026MAPLE: Metadata Augmented Private Language Evolutioncited by 0
  12. 2026Jacobian-Guided Anisotropic Noise Reshaping for Enhancing Representation Utility under Local Differential Privacycited by 0
  13. 2026ContinuousBench: Can Differentially Private Synthetic Text Improve Capabilities?cited by 0
  14. 2026Collaborative Synthetic Data Generation for Knowledge Transfer in Federated Learningcited by 0
  15. 2026Canaries in the Bank: Auditing User-Level Privacy in Private Evolutioncited by 0
  16. 2026QuanText: Protecting Dataset-Level Secrets in Textual Data Sharingcited by 0
  17. 2026Only Pay What You Must Spend: On-Demand Privacy Budget Payment for Differentially Private RAGcited by 0