We have compiled this new guide to prompt engineering, introducing relevant papers, learning guides, models, lectures, references, language model capabilities, and other tools related to prompt engineering.
This guide introduces the basics of prompt words, helps users understand how to interact with large language models through prompt words, and provides relevant guidance suggestions.
Unless otherwise specified, all examples in this south korea mobile database guide are tested on OpenAI's large language model text-davinci-003, using the model's default configuration, such as temperature=0.7 and top-p=1.
Model setup
When using prompt words, you can interact with the large language model through the API or directly. You can get different prompt results by configuring some parameters.
Temperature : In simple terms, temperaturethe smaller the parameter value, the more certain the results returned by the model. If this parameter is increased, large language models may return more random results, that is, more diverse or more creative outputs. We can also achieve similar effects by increasing the weights of other possible tokens. In practical applications, for tasks such as quality assurance, we can set a lower temperaturevalue to make the model return more realistic and concise results based on facts. For creative tasks such as poetry generation, appropriately increase temperaturethe parameter value.