Differentially Private Synthetic Data via APIs 3: Using Simulators Instead of Foundation Model

arXiv:2502.05505 · cs.LG, cs.CR, cs.CV, stat.ML · Submitted 2025-02-08 · Read on arXiv

cs.LG, cs.CR, cs.CV, stat.ML

Submitted: 2025-02-08

Updated: 2026-10-11

Comments: Published in: (1) TMLR (2026), (2) ICLR 2025 Workshop on Data Problems, (3) ICLR 2025 Workshop on Synthetic Data

Code: https://github.com/microsoft/DPSDA

Project page: https://genesis-embodied-ai.github.io/4https://github.com/blender/blender

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Foundation Model — 5× in this paper · explained in 3 episodes
• Foundation Models — 3× in this paper · explained in 7 episodes
• Differentially Private — 2× in this paper · explained in 3 episodes
• Private Evolution — 2× in this paper
• Accuracy — 1× in this paper · explained in 1 episode

Sources

• Comparative Study of Differentially Private Synthetic Data Algorithms from the NIST PSCR Differential Privacy Synthetic Data Challenge
• Differentially Private Diffusion Models
• Differentially Private Diffusion Models Generate Useful Synthetic Images
• DPImageBench: A Unified Benchmark for Differentially Private Image Synthesis
• Exploring the Limits of Differentially Private Deep Learning with Group-wise Clipping
• PrE-Text: Training Language Models on Private Federated Data in the Age of LLMs
• POPri: Private Federated Learning using Preference-Optimized Synthetic Data
• Large Language Models Can Be Strong Differentially Private Learners
• Distilled Decoding 1: One-step Sampling of Image Auto-regressive Models with Flow Matching
• Is API Access to LLMs Useful for Generating Private Synthetic Tabular Data?
• Benchmarking Differentially Private Synthetic Data Generation Algorithms
• Differentially Private Fine-Tuning of Diffusion Models
• Differentially Private Synthetic Data via Foundation Model APIs 2: Text
• Differentially Private Generative Adversarial Network
• Differentially Private Neural Tangent Kernels for Privacy-Preserving Data Generation
• Selective Pre-training for Private Fine-tuning
• Differentially Private Fine-tuning of Language Models
• Synthetic Text Generation with Differential Privacy: A Simple and Practical Recipe
• Wide Residual Networks
• Contrastive Private Data Synthesis via Weighted Multi-PLM Fusion

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks