Reproduce “Robust Utility-Preserving Text Anonymization Based on Large Language Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tianyu Yang, Xiaodan Zhu, Iryna Gurevych
RUPTA anonymizes text through iterative editing guided by LLM evaluations of re-identification risk and occupation-classification utility. It prioritizes privacy, then seeks to recover utility without reducing the privacy score. On DB-bio, RUPTA (GPT-4) achieves 52.67% attack success and 96.02% accuracy, versus AF’s 52.91% and 92.02%, supporting higher utility at similar attack success in this setting, not a privacy guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.