Evaluating deep generative models on cognitive tasks: a case study by Zhisheng Tang, Mayank Kejriwal

“…Similarly, in evaluating ChatGPT on the rationality axioms developed under the classical Von Neumann-Morgenstern utility theorem, we find that, although it demonstrates some level of rational decision-making, many of its decisions violate at least one of the axioms even under reasonable constructions of preferences, bets, and decision-making prompts. …”
Get full text