Group selection to evolve cooperative prompts in populations of LLM agents

Current approaches to instill cooperation in large language model (LLM) agents often rely on humans specifying desired behaviors at the individual level, which does not guarantee cooperation within LLM populations. As frontier training shifts toward individual rewards for verifiable tasks, such as mathematics and coding, this outcome-based focus may further undermine cooperation in multi-agent settings. Large-scale cooperation in human populations emerged via unguided evolutionary mechanisms, not a central architect. Group selection, under which cooperative groups within a population can outcompete less cooperative ones, has been argued to be essential for the evolution of cooperation. In this study, we explore whether group selection can promote cooperation in populations of LLM agents. We introduce a multi-agent simulation framework in which LLM agents play a repeated social dilemma game and transmit their natural-language prompts across generations under either individual- or group-level selection. Under group selection, prompts from high-performing groups are transmitted, thereby promoting cooperation. Under individual selection, self-interested prompts dominate, causing populations to collapse into collective defection. This gap is robust across prompt ablations, alternative game framings, and model swaps. We theoretically reproduce key results using a replicator-mutator model, whose empirical transmission kernel predicts a phase transition at a critical threshold. Finally, we find that frontier models as GPT 5.4, 5.6 and Fable 5 are anticipating the best strategy for a certain selection regime side-stepping any evolutionary process involved.