This technology is a hybrid large language model (LLM) that connects artificial intelligence (AI) agents into open source 4X and grand strategy games. Vox Deorum enables the observation and comparison of reasoning and decision-making across different AI agents within a game environment, while also facilitating interaction between human players and AI agents. The technology has the potential to enhance human–AI decision making, support the evaluation, training, and optimization of generative AI and agent-based systems, and drive innovation in the design of strategy games. Background: Large Language Models’ capacity to reason in natural language makes them promising for use in 4X (eXploration, eXpansion, eXploitation, and eXtermination) and grand strategy games. LLMs could enable more natural human-AI gameplay interactions such as collaboration and negotiation. However, these games can be challenging due to their complexity and long-horizon nature, which can affect LLMs’ latency and create enormous costs. Vox Deorum addresses this through a hybrid structure that uses LLMs for strategic decision making while specialized subsystems handle tactical execution. Applications:
Advantages: