Abstract: Large Language Models (LLMs) have shown impressive capabilities in code generation. However, they often struggle in complex programming scenarios due to incomplete semantic understanding and ...