Specialist LLMs Drive Belief Consensus in Multi-Agent Social Networks

Using the CoevolveSim framework, research shows that introducing fine-tuned specialist LLMs into multi-agent networks significantly impacts population-level belief consensus. While persona assignment and network structure alter individual revision, only domain specialization effectively reshapes group-level outcomes.