This paper discusses a novel system for controlling game agents using natural language commands. The system employs a large language model (LLM) to interpret and convert these commands into a ‘behavior branch’, a newly proposed knowledge expression based on behavior trees. The system was tested in a simulated Pokémon game environment and demonstrated its ability to understand and execute natural language commands. This represents a significant development in real-time language-interactive game agents.
Publication date: 13 Feb 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2402.07442
