mlsanityData to 14 Oct 2026

Private Synthetic Text Generation with Diffusion Models

arXiv 2410.22971 ↗First posted 30 Oct 2024cs.CLcited by 10

On the map: Diffusion models → Robustness, privacy and adversarial aspects

Abstract (arXiv)

How capable are diffusion models of generating synthetics texts? Recent research shows their strengths, with performance reaching that of auto-regressive LLMs. But are they also good in generating synthetic data if the training was under differential privacy? Here the evidence is missing, yet the promises from private image generation look strong. In this paper we address this open question by extensive experiments. At the same time, we critically assess (and reimplement) previous works on synthetic private text generation with LLMs and reveal some unmet assumptions that might have led to violating the differential privacy guarantees. Our results partly contradict previous non-private findings and show that fully open-source LLMs outperform diffusion models in the privacy regime. Our complete source codes, datasets, and experimental setup is publicly available to foster future research.

Key facts. Cited by 10 papers in the map (data through 14 Oct 2026). Placed in Robustness, privacy and adversarial aspects, one of 14 directions in Diffusion models. Builds on 28 of its 54 references that are in the map. First posted on arXiv 30 Oct 2024.

Builds on

28 of 54 references.

References · 28

  1. 2017Attention Is All You Needcited by 60,361
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  3. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  4. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,599
  5. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
  6. 2019BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehensioncited by 5,893
  7. 2016Deep Learning with Differential Privacycited by 2,728
  8. 2022BLOOM: A 176B-Parameter Open-Access Multilingual Language Modelcited by 1,984
  9. 2018Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarizationcited by 1,317
  10. 2022Diffusion-LM Improves Controllable Text Generationcited by 1,027
  11. 2018The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networkscited by 864
  12. 2022DiffuSeq: Sequence to Sequence Text Generation with Diffusion Modelscited by 450
16 more
  1. 2023Scalable Extraction of Training Data from (Production) Language Modelscited by 434
  2. 2021Opacus: User-Friendly Differential Privacy Library in PyTorchcited by 383
  3. 2017A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasetscited by 377
  4. 2022Analog Bits: Generating Discrete Data using Diffusion Models with Self-Conditioningcited by 373
  5. 2022Understanding Diffusion Models: A Unified Perspectivecited by 340
  6. 2020Synthetic Data -- Anonymisation Groundhog Daycited by 108
  7. 2022Synthetic Text Generation with Differential Privacy: A Simple and Practical Recipecited by 108
  8. 2022Differentially Private Language Models for Secure Data Sharingcited by 68
  9. 2023DiffuSeq-v2: Bridging Discrete and Continuous Text Spaces for Accelerated Seq2Seq Diffusion Modelscited by 36
  10. 2023The ROOTS Search Tool: Data Transparency for LLMscited by 31
  11. 2020Suicidal Ideation and Mental Disorder Detection with Attentive Relation Networkscited by 28
  12. 2022DP-Rewrite: Towards Reproducibility and Transparency in Differentially Private Text Rewritingcited by 28
  13. 2023Advancing Differential Privacy: Where We Are Now and Future Directions for Real-World Deploymentcited by 27
  14. 2019Skin Lesion Synthesis with Generative Adversarial Networkscited by 22
  15. 2024Novel Interpretable and Robust Web-based AI Platform for Phishing Email Detectioncited by 7
  16. 2023Publishing Wikipedia usage data with strong privacy guaranteescited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 5; 2026: 5.

10 papers.

  1. 2025Synthesizing Privacy-Preserving Text Data via Finetuning without Finetuning Billion-Scale LLMscited by 16
  2. 2025Evaluating Differentially Private Generation of Domain-Specific Textcited by 5
  3. 2025SynBench: A Benchmark for Differentially Private Text Generationcited by 4
  4. 2025ACTG-ARL: Differentially Private Conditional Text Generation with RL-Boosted Controlcited by 3
  5. 2026Diffusion Models Memorize in Training -- and Generalize in Inferencecited by 2
  6. 2025Differentially-private text generation degrades output language qualitycited by 1
  7. 2026Differentially Private Preference Data Synthesis for Large Language Model Alignmentcited by 1
  8. 2026The Conundrum of Trustworthy Research on Attacking Personally Identifiable Information Removal Techniquescited by 0
  9. 2026MAPLE: Metadata Augmented Private Language Evolutioncited by 0
  10. 2026PrivBench: A Holistic and Modular Benchmarking Platform for Evaluating Text-to-Text Privatizationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024DP-RDM: Adapting Diffusion Models to Private Domains Without Fine-Tuning
  2. 2023Differentially Private Diffusion Models Generate Useful Synthetic Images
  3. 2024Leveraging Programmatically Generated Synthetic Data for Differentially Private Diffusion Training
  4. 2025InvisibleInk: High-Utility and Low-Cost Text Generation with Differential Privacy
  5. 2026Private Seeds, Public LLMs: Realistic and Privacy-Preserving Synthetic Data Generation
  6. 2025RAPID: Retrieval Augmented Training of Differentially Private Diffusion Models
  7. 2024Differentially Private Synthetic Data via Foundation Model APIs 2: Text
  8. 2024Evaluating Differentially Private Synthetic Data Generation in High-Stakes Domains