synbench/

synbench-data

	Synthetic NLP Data for Differentially Private Text Generation Evaluation




	
	Dataset Description

This dataset contains original and synthetic text classification data used to evaluate differentially private (DP) synthetic text generation methods across three dimensions: utility, fidelity, and membership inference attack (MIA) resistance.

The original data is sourced from:

🔗 Daniel-ML/sentiment-analysis-for-financial-news

We use this dataset… See the full description on the dataset page: https://huggingface.co/datasets/synbench/synbench-data.