Safety

Detecting Training Data of Large Language Models via Expectation Maximization

Novel membership inference attack framework for large language models via an expectation maximization algorithm