JevMade Sign in
← Back to guides

JevMade field notes / Training and evaluation guide

How to train and test a local AI decision tool

This guide explains how to set up and test Kev, a free AI tool that reads text and chooses answers from a list. You will learn how to teach it specific rules for sorting your own information.

Original by Jared PalmerEvaluation

Listen to this guide

JevMade’s plain-English explanation

0:00 /

Our summary

Kev is a set of AI models that act like Jev, an AI tool that chooses from options rather than writing an answer. People use it to automatically sort text, like figuring out which department should handle a customer complaint or deciding if a message is urgent.

The software reads a piece of text and answers specific multiple-choice, rating, or yes-or-no questions about it. The author provides instructions for teaching the tool to follow your own rules using examples of past decisions. You can run this software on your own computer or online.

This guide is useful for people who want to test how well an AI sorts information before trusting it with real tasks. The software struggles with long documents and general knowledge questions. You should always check its accuracy on your own examples before letting it make decisions.

Key takeaways

  1. Build the software using thousands of public examples and generated rules to teach it how to make choices.
  2. Test if asking questions at the same time changes the answers compared to asking them one by one.
  3. Separate the text you want the software to read from the instructions you give it to avoid confusion.

The guide reports the training steps and test results for these local models.

GitHub repository

Read the original guide Opens the author’s site in a new tab.

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.