Data to be translated: Google DeepMind introduces OPRO technology, utilizing LLMs as optimizers for their own prompts. This technique guides LLMs to generate and refine solutions through natural language descriptions of problems, particularly effective for issues where gradients are difficult to obtain. Studies have demonstrated the effectiveness of OPRO in mathematical optimization problems, especially its potential impact on large language models like ChatGPT and PaLM. Although the code has not been released, the concept of OPRO is straightforward and intuitive. LlamaIndex has shared implementation guidelines for using OPRO in retrieval-augmented generation tasks.