behavioural-signatures

The paper examines whether large language models exhibit stable and interpretable risk preferences in decision-making under uncertainty, using Texas Hold'em to quantify baseline risk dispositions and context-dependent adaptations.