Abstract: Large Language Models (LLMs) agents with ability to predict the further world states could avoid security risks. A core capability of the world model is the precise control over the ...