Explore the Reasoning Capability of LLMs in the Chess Testbed