We seek to elevate the planning capabilities of large language models (LLMs)investigating four main directions. First, we construct a comprehensive benchmark suite encompassing both classical planning domains and