DYL

Researchers have developed new testbeds to evaluate lie detectors for language models, finding that existing detection methods often fail when models are trained to hold false beliefs. The study teste…