Contrastive Private Data Synthesis via Weighted Multi-PLM Fusion

Tianyuan Zou, Yang Liu, Peng Li, Yufei Xiong, Jianqing Zhang, Jingjing Liu, Xiaozhou Ye, Ye Ouyang, Ya-Qin Zhang
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:80846-80872, 2025.

Abstract

Substantial quantity and high quality are the golden rules of making a good training dataset with sample privacy protection equally important. Generating synthetic samples that resemble high-quality private data while ensuring Differential Privacy (DP), a formal privacy guarantee, promises scalability and practicality. However, existing methods relying on pre-trained models for data synthesis often struggle in data-deficient scenarios, suffering from limited sample size, inevitable generation noise and existing pre-trained model bias. To address these challenges, we propose a novel contrAstive private data Synthesis via Weighted multiple Pre-trained generative models framework, named as WASP. WASP utilizes limited private samples for more accurate private data distribution estimation via a Top-Q voting mechanism, and leverages low-quality synthetic samples for contrastive generation via collaboration among dynamically weighted multiple pre-trained models. Extensive experiments on 6 well-developed datasets with 6 open-source and 3 closed-source PLMs demonstrate the superiority of WASP in improving model performance over diverse downstream tasks. Code is available at https://github.com/LindaLydia/WASP.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-zou25d, title = {Contrastive Private Data Synthesis via Weighted Multi-{PLM} Fusion}, author = {Zou, Tianyuan and Liu, Yang and Li, Peng and Xiong, Yufei and Zhang, Jianqing and Liu, Jingjing and Ye, Xiaozhou and Ouyang, Ye and Zhang, Ya-Qin}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {80846--80872}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/zou25d/zou25d.pdf}, url = {https://proceedings.mlr.press/v267/zou25d.html}, abstract = {Substantial quantity and high quality are the golden rules of making a good training dataset with sample privacy protection equally important. Generating synthetic samples that resemble high-quality private data while ensuring Differential Privacy (DP), a formal privacy guarantee, promises scalability and practicality. However, existing methods relying on pre-trained models for data synthesis often struggle in data-deficient scenarios, suffering from limited sample size, inevitable generation noise and existing pre-trained model bias. To address these challenges, we propose a novel contrAstive private data Synthesis via Weighted multiple Pre-trained generative models framework, named as WASP. WASP utilizes limited private samples for more accurate private data distribution estimation via a Top-Q voting mechanism, and leverages low-quality synthetic samples for contrastive generation via collaboration among dynamically weighted multiple pre-trained models. Extensive experiments on 6 well-developed datasets with 6 open-source and 3 closed-source PLMs demonstrate the superiority of WASP in improving model performance over diverse downstream tasks. Code is available at https://github.com/LindaLydia/WASP.} }
Endnote
%0 Conference Paper %T Contrastive Private Data Synthesis via Weighted Multi-PLM Fusion %A Tianyuan Zou %A Yang Liu %A Peng Li %A Yufei Xiong %A Jianqing Zhang %A Jingjing Liu %A Xiaozhou Ye %A Ye Ouyang %A Ya-Qin Zhang %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-zou25d %I PMLR %P 80846--80872 %U https://proceedings.mlr.press/v267/zou25d.html %V 267 %X Substantial quantity and high quality are the golden rules of making a good training dataset with sample privacy protection equally important. Generating synthetic samples that resemble high-quality private data while ensuring Differential Privacy (DP), a formal privacy guarantee, promises scalability and practicality. However, existing methods relying on pre-trained models for data synthesis often struggle in data-deficient scenarios, suffering from limited sample size, inevitable generation noise and existing pre-trained model bias. To address these challenges, we propose a novel contrAstive private data Synthesis via Weighted multiple Pre-trained generative models framework, named as WASP. WASP utilizes limited private samples for more accurate private data distribution estimation via a Top-Q voting mechanism, and leverages low-quality synthetic samples for contrastive generation via collaboration among dynamically weighted multiple pre-trained models. Extensive experiments on 6 well-developed datasets with 6 open-source and 3 closed-source PLMs demonstrate the superiority of WASP in improving model performance over diverse downstream tasks. Code is available at https://github.com/LindaLydia/WASP.
APA
Zou, T., Liu, Y., Li, P., Xiong, Y., Zhang, J., Liu, J., Ye, X., Ouyang, Y. & Zhang, Y.. (2025). Contrastive Private Data Synthesis via Weighted Multi-PLM Fusion. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:80846-80872 Available from https://proceedings.mlr.press/v267/zou25d.html.

Related Material