State of Prompt Engineering

Explore the state of prompt engineering and its groundbreaking impact on AI and task performance. The sophistication of prompt engineering has a significant impact on task performance, marking a breakthrough in the world of artificial intelligence and large language models (LLMs). My name is Fede Nolasco, and in this article we are diving deep into the State of Prompt Engineering (State of AI Report 2023, page 35), unraveling the intricacies and marvels it holds.

State of Prompt Engineering

Chain of Thought and Tree of Thought

Chain of Thought (CoT) prompting requires a language model to output intermediate reasoning steps, a technique that has been proven to enhance performance significantly. Furthermore, the Tree of Thought (ToT) methodology improves on CoT by sampling multiple times and representing thoughts as nodes in a tree structure. In ToT, various search algorithms can be applied to explore the tree structure, with the LLM assigning values to each node, such as “sure”, “likely”, or “impossible”.

Graph of Thought

Additionally, we see the emergence of the Graph of Thought (GoT), which transforms the reasoning tree into a graph by combining similar nodes. This intricate web of thoughts provides a comprehensive perspective on the problem-solving process. The most exciting part is that LLMs are proving to be adept prompt engineers themselves. For instance, Auto-CoT achieves performance comparable to or even exceeding CoT on ten reasoning tasks.

The Automatic Prompt Engineer

The Automatic Prompt Engineer (APE) is a remarkable development in the field of artificial intelligence, representing a significant leap forward in the efficiency and effectiveness of machine learning models. With its ability to outperform on 19 out of 24 tasks, APE has proven itself to be an invaluable tool for those looking to obtain more accurate and informative responses from their models. The importance of APE lies in its capacity to guide models towards providing answers that are not only correct but also rich in information, ensuring that users receive the most comprehensive response possible.

Optimization by Prompting

Optimization by Prompting (OPRO) represents another groundbreaking advancement in the world of artificial intelligence. With its innovative approach to prompt design, OPRO has demonstrated that optimized prompts can significantly outperform human-designed prompts. This is particularly evident in challenging benchmarks such as GSM8K and Big-Bench Hard, where optimized prompts have been shown to improve performance by over 50%. The success of OPRO highlights the potential for artificial intelligence to revolutionize the way we approach problem-solving, providing more accurate and efficient solutions to even the most complex challenges.

The Future of Problem Solving

Let’s take a moment to consider the implications of above advancements. Have you ever wondered how these sophisticated techniques could revolutionize the way we approach problem-solving in various domains? Whether you are a business professional or a budding data scientist, the implications are profound. This is not just about machines getting smarter; it’s about enhancing human capabilities and opening new horizons for innovation and creativity.


As we conclude this article of the State of Prompt Engineering, let us take a moment to reflect on the potential of these methodologies to revolutionize our approach to problem-solving and innovation. I encourage you all to share your thoughts and ideas on this topic. Let’s engage in a meaningful conversation and explore the infinite possibilities that these technologies hold. And remember, I am always open to your suggestions for new articles and topics for my blog “datatunnel”.

I would like to share a humorous yet insightful quote by Albert Einstein: “The only reason for time is so that everything doesn’t happen at once.” Let’s embrace the changes and advancements in artificial intelligence and prompt engineering, and let’s work together to shape a future that is more innovative, efficient, and creative.

And, if you enjoyed this article, feel free to follow me on LinkedIn or Twitter for more insightful content on artificial intelligence, data management, and much more. Let’s connect and explore the infinite possibilities that the world of data has to offer.


  1. Graph of Thoughts: Solving Elaborate Problems with Large Language Models [arXiv]
  2. Automatic Chain of Thought Prompting in Large Language Models [arXiv]
  3. Large Language Models Are Human-Level Prompt Engineers [arXiv]
  4. Large Language Models as Optimizers [arXiv]
  5. How is ChatGPT’s Behaviour Changing over Time? [arXiv]
  6. Does Context Length Matters?

Similar Posts