Abstract: Despite the widespread deployment of Large Language Models (LLMs) in embodied intelligent agents for their promising human-like reasoning capabilities, they remain hindered by local ...