Method for optimizing prompt engineering and related products

A device for optimizing a prompt engineering, the device comprising:

• a processor;
• and
• a memory having stored thereon computer instructions for optimizing a prompt engineering that, when executed by the processor, cause implementation of:
• graphically presenting a target manipulation interface to a user, wherein different creation modes for the prompt engineering are displayed on the target manipulation interface, and each creation mode supports a structured creation of the prompt engineering;
• in response to the user's selection operation of the different creation modes, implementing the structured creation of the prompt engineering with the creation mode selected by the user;
• performing quality assessment on a structured prompt engineering based on a pre-established training platform, wherein the performing of the quality assessment comprises:
• creating a test task regarding the structured prompt engineering and test data for the test task on the training platform, wherein the creating of the test task comprises:
• presenting a task creation interface graphically in the training platform, wherein the task creation interface includes a task setting column and a test data setting column;
• and
• in response to the user's setting operations on the task setting column and the test data setting column, determining the test task and test data based on the settings of the task setting column and the test data setting column;
• invoking a second large language model to output a test result based on the test data and the structured prompt engineering;
• and
• conducting the quality assessment of the structured prompt engineering based on the test result;
• wherein the structured prompt engineering includes multiple prompt engineerings to be assessed, and the invoking a second large language model to output a test result based on the test data and the structured prompt engineering, and
• conducting the quality assessment of the structured prompt engineering based on the test result comprises:
• obtaining multiple experimental groups regarding the test task and the test data, wherein each experimental group is configured with a corresponding prompt engineering to be assessed;
• invoking the second large language model to output test results for each experimental group based on its corresponding prompt engineering to be assessed;
• obtaining scores for the test results of each experimental group;
• and
• determining the quality of the prompt engineerings to be assessed corresponding to each experimental group based on the scores for the test results of each experimental group.

A non-transitory machine readable medium having stored thereon computer program instructions for optimizing a prompt engineering, which when executed by one or more processors, cause implementing of:

• graphically presenting a target manipulation interface to a user, wherein different creation modes for the prompt engineering are displayed on the target manipulation interface, and each creation mode supports a structured creation of the prompt engineering;
• in response to the user's selection operation of the different creation modes, implementing the structured creation of the prompt engineering with the creation mode selected by the user;
• performing quality assessment on a structured prompt engineering based on a pre-established training platform, wherein the performing of the quality assessment comprises:
• creating a test task regarding the structured prompt engineering and test data for the test task on the training platform, wherein the creating of the test task comprises:
• presenting a task creation interface graphically in the training platform, wherein the task creation interface includes a task setting column and a test data setting column;
• and
• in response to the user's setting operations on the task setting column and the test data setting column, determining the test task and test data based on the settings of the task setting column and the test data setting column;
• invoking a second large language model to output a test result based on the test data and the structured prompt engineering;
• and
• conducting the quality assessment of the structured prompt engineering based on the test result;
• wherein the structured prompt engineering includes multiple prompt engineerings to be assessed, and the invoking a second large language model to output a test result based on the test data and the structured prompt engineering, and
• conducting the quality assessment of the structured prompt engineering based on the test result comprises:
• obtaining multiple experimental groups regarding the test task and the test data, wherein each experimental group is configured with a corresponding prompt engineering to be assessed;
• invoking the second large language model to output test results for each experimental group based on its corresponding prompt engineering to be assessed;
• obtaining scores for the test results of each experimental group;
• and
• determining the quality of the prompt engineerings to be assessed corresponding to each experimental group based on the scores for the test results of each experimental group.