Monitoring AI-Modified Content at Scale: A Case Study on the…
We present an approach for estimating the fraction of text in a large corpus which is likely to be substantially modified or produced by a large language model (LLM). Our maximum likelihood model…