Differentially Private In-Context Learning
Differentially Private In-Context Learning
Ashwinee Panda,Tong Wu,Jiachen T. Wang,Prateek Mittal
2023 · DOI: 10.48550/arXiv.2305.01639
arXiv.org · 26 Citations
TLDR
Differentially Private In-context Learning (DP-ICL) is proposed to enable LLMs to adapt to new tasks while maintaining privacy guarantees, and achieves comparable performance (< 2% degradation) with non-private ICL.