Auto-document models with GenAI

The LLM (Large Language Model) feature in Vectice enables data scientists to automatically document their work using GenAI. This tool enhances documentation creation by providing an "AskAI" button within the app. Users can generate documentation for iterations, models, or datasets or use the "Ask me anything" feature to prompt AI for custom documentation.

Auto-Generate Documentation with AskAI

Vectice allows you to use "AskAI" to generate documentation directly from your iterations or assets. This feature provides a straightforward way to document your work quickly, ensuring that all necessary details are captured without added effort.

Simply select AskAI and then choose your iterations or assets. AskAI will help you create clear and concise documentation for your projects. You can also use the "Ask Me Anything" feature to produce customized documentation based on your prompts, tailoring the content to meet your needs.

'Ask Me Anything' example use cases

Click the AskAI button, then select Ask Me Anything. Use the following prompt examples to generate asset documentation.

To generate documentation based on your assets and phases, use AskAI's "Ask me anything" feature and use the / character to select the desired assets or phrases.

Summarize model

The following AI prompt will summarize your model's architecture, parameters, and performance metrics.

Generate a summary of the model {{ MDV-XXX }}.

Compare models

The following AI prompt will generate a detailed comparison of multiple models, highlighting differences in technique, training data, and results.

Compare the following models {{ MDV-XXX }} and {{ MDV-YYY }}. Focus on differences in their techniques, training data, and performance outcomes.

Compare datasets

The following AI prompt will generate a comparative analysis of various datasets, focusing on key attributes, data distribution, and suitability for different models.

Compare these datasets {{ DTV-XXX }} and {{ DTV-YYY }} by analyzing their key attributes, data distribution, and suitability for model {{ MDV-XXX }}.

Explain model

The following AI prompt will comprehensively explain your model's functioning, including feature importance and decision-making processes.

Explain the functioning of the model {{ MDV-XXX }} along with its feature importance and decision-making processes.

Customize your own prompt

You can even customize a prompt to help you create specific documentation tailored to your needs.

Document the latest model's performance on the validation set compared to the previous model {{ MDV-XXX }}.

More advanced use cases

To prevent AI from generating incorrect information, use methods that allow you to fine-tune prompts without noise, such as:

Phases:

• {{ PHA-XXX.description }}

• {{ PHA-XXX.documentation }}

Model versions:

• {{ MDV-XXX.description }}

• {{ MDV-XXX.type }}

• {{ MDV-XXX.status }}

• {{ MDV-XXX.technique }}

Dataset versions:

• {{ DTV-XXX.description }}

• {{ DTV-XXX.type }}

This ensures the AI includes accurate descriptions and information and avoids fabricating details.