LLMEE: Enhancing explainability and evaluation of large language models through visual token attribution