Generative agent-based modeling with actions grounded in physical, social, or digital space using Concordia

https://arxiv.org/abs/2312.03664

Agents state intended actions in natural language; a Game Master agent, inspired by tabletop RPGs, decides outcomes, checks physical plausibility, and can call digital tools. Behavior comes from components mediating LLM calls and associative memory retrieval.

Many-agent LLM societies, LLM game master and narrative director, LLM-adjudicated state transitions (LLM as game mechanic)