A more flexible training method can make AI text detectors more robust when new models emerge or AI-generated text is disguised to appear human.
Source disclosure
This is a source-linked briefing. Read the original report from Massachusetts Institute Of Technology .