关于LLM使用tool的探索

参考文章：
Toolformer: Language Models Can Teach Themselves to Use Tools

SELF-REFINE: Iterative Refinement with Self-Feedback

HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face
motivation：多模态，复杂任务，性能不如expert
pipeline：task planning, model selection, task execution, response generation
效果一定好吗？
如果是，主要是因为哪一个零件？（哪一部分做得好）