Skip to content

Welcome to the Prompt Dictionary!

Welcome to The Prompt Dictionary!

Welcome to The Prompt Dictionary! This dictionary aims to be a comprehensive repository of strategies designed to enhance interactions with large language model. I update this repository frequently to keep pace with rapid advancements, though there's often a backlog of new strategies to add. This archive captures each strategy at a specific point in time, acknowledging that some may become less relevant over time as language models evolve.

Simple is best

Many of the techniques listed in the Prompt Dictionary are outlined here for posterity and may not be suitable for everyday use. When it comes to practical prompting simple is best. In his interview on the Cognitive Revolution podcast Sander Schulhoff of LearnPrompting.org recommended a four-step prompt:

  1. Provide context for the task (e.g., if your analyzing a document provide the document).
  2. Give the model examples of the output you want.
  3. Give the model examples of the output you don't want.
  4. Clearly state the task, question, or instruction.

Historical record

The goal of The Prompt Dictionary is to serve as a historical record, documenting the evolution, benefits, and limitations of various prompting techniques. While some practical tutorials and examples are included, the primary function of this dictionary is to provide authoritative, reference-style information on prompt techniques and related practical applications. It functions similarly to a dictionary in the technical domain, offering definitions, usage examples, and specific guidelines, rather than comprehensive theoretical discussions or extended instructional content found in courses or manuals. Each entry provides:

  • Background on the prompting strategy
  • When to use it
  • What to know about its history
  • Best practices
  • What to watch out for
  • A link to the original source
  • Prompting examples
  • Code implementations and small-scale replications