LATIDIA · Ciberseguridad

¿Humanizado pero aún cierto? Auditoría de hechos y detección en humanizadores de texto basados ​​en IA

arXiv:2611.00359v1 Anuncio Tipo: cross Resumen: Los "humanizadores" de IA reescriben texto generado por máquina para que los detectores lo clasifiquen como escrito por humanos. Las tablas de clasificación públicas puntúan la evasión y la similitud de incrustación, pero no si

La noticia

arXiv:2611.00359v1 Announce Type: cross Abstract: AI "humanizers" rewrite machine-generated text so that detectors classify it as human-written. Public leaderboards score evasion and embedding similarity, but not whether checkable details survive. We check whether numbers, dates, person and organization names, and quoted spans survive each rewrite, using a rule-based checker validated by two AI annotators (precision 0.84-1.00, recall 0.93-0.95; no human annotation). In the October 2026 cycle of a public humanizer leaderboard, all 11 commercial tools changed or dropped at least one of these units in 35-65% of the documents that contained one. Most changes were to quoted spans; counting numbers and names alone, 10 of the 11 tools still altered