Tag
The paper introduces Telescope Perplexity, a metric that measures token repetition probability to detect LLM-generated text in a zero-shot manner, achieving state-of-the-art or competitive performance across diverse datasets.