Join us for a thought-provoking session led by Chulin Xie as she presents her latest research, “Differentially Private Synthetic Data via Foundation Model APIs 2: Text.” In her work, Chulin addresses the critical challenge of privacy in AI by introducing a method for generating high-quality, synthetic text data that ensures differential privacy—essential for sensitive applications like healthcare, finance, and legal fields.
Using an innovative approach, Chulin’s method bypasses traditional, resource-heavy finetuning and instead leverages only API access to large language models. This breakthrough allows organizations to produce privacy-preserving synthetic data with significant efficiency and adaptability, expanding possibilities for privacy-compliant AI development.
Don’t miss this session, ideal for AI professionals, data scientists, and researchers exploring advanced privacy techniques in AI.
Chulin Xie is a fifth-year Ph.D. candidate in Computer Science at the University of Illinois Urbana-Champaign. Her research interests primarily center on trustworthy machine learning and optimization. She studies the risks in ML systems (e.g., LLMs, LLM agents, federated learning), including safety vulnerabilities, privacy breaches, failures in reasoning and generalization, and designs tailored mitigation solutions to make ML systems more reliable. Her work received a NeurIPS 2023 outstanding paper award at Datasets and Benchmark track, and a VLDB 2024 best research paper nomination. She gained industry experience during research internships at Nvidia, Microsoft, and Google.
©2024 SSI Club. All Rights Reserved