Automated authorship scoring hits AUC 0.90 in verifying real manuscript authors

Dr_Atoosa · x · 2026-09-17

An automated scoring system achieved AUC 0.90 predicting whether a participant was the genuine author of a manuscript. User studies found positive experiences, with participants affirming the questions captured meaningful authorship while flagging weak points and refinements needed before real-world deployment.

Related event: greCAPTCHA: Verifying True Authorship by Testing Authors' Understanding of Their Manuscripts(7 posts)→

Original post →

More from Research

Research channel →