Tag
This paper examines how persona prompts influence strategic behavior of large language model agents in an iterated Split or Steal game, finding that mutual Split outcomes dominate and that model choice and persona type significantly affect cooperation and exploitation.