#88 · Primary category: Prompt Engineering Tools

tree-of-thought-prompting

large-language-models llm prompt-engineering

Using Tree-of-Thought Prompting to boost ChatGPT's reasoning

Project last updated:12/09/23

GitHub Stars

820

Forks

76

Contributors

4

License

MIT

Why we included this project

Anyone who works with large language models has probably seen a Chain-of-Thought prompt return a confident but wrong answer. This repository shows a different approach: Tree-of-Thought prompting, which pushes the model to explore several reasoning branches, catch its own mistakes, and revise before settling on an answer. The author demonstrates it with a simple spatial puzzle that ChatGPT 3.5 gets wrong but answers correctly once the ToT prompt is applied, and includes a case where the technique brings a 3.5 model up to reasoning that previously needed GPT-4. The README also explains the ideas behind the technique clearly enough that you can adapt the prompt patterns to your own work instead of treating them as a black box. It's a short, practical reference for getting better reasoning out of models you already have.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category