Language Models Optimized to Fool Detectors Still Have a Distinct Style (And How to Change It)

Open in new window