Fantastic-tuning LLMs with user-level differential privateness

Making these algorithms work for LLMs If we run these algorithms “out-of-the-box” for LLMs, issues go…