Tag
The paper theoretically analyzes how repeated paraphrasing can align AI-generated text with human writing distributions, deriving convergence rates and implications for reliable AI-text detection.