Fig. 17
Comparison of − log10(p-values) for multiple statistical tests on training-set features, illustrating that all evaluated text attributes significantly differ (p < 0.05) between AI-generated and human-written texts.
Fig. 17 | Scientific Reports
