The field of agent-based modeling and large language models is moving towards increased complexity and realism, with a focus on simulating dynamic systems and making decisions in complex environments. Recent research has demonstrated the potential of these models to capture nuanced interactions and behaviors, such as those found in urban development and financial markets. The integration of large language models with agent-based modeling has enabled the creation of more sophisticated and interpretable models, capable of reproducing empirical patterns and making predictions about future outcomes. Notably, the use of prompt-guided agents has shown promise in emulating the strategies of expert investors and making profitable trades in real-world markets.

Some noteworthy papers in this area include: GuruAgents, which demonstrates the ability of prompt-guided AI agents to emulate the strategies of legendary investment gurus and achieve high performance in backtests. StockBench, which introduces a contamination-free benchmark for evaluating large language models in realistic stock trading environments and highlights the challenges and opportunities in developing LLM-powered financial agents.