JevMade field notes / Pipeline guide

Sort training data and measure changes in an AI assistant

This guide explains how to check text files for errors, use an AI tool to sort the information, and measure how the software changes after learning from the approved text.

Original by RenaGaoEvaluation

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Credits and license

“Screen data, train a model, and evaluate the change” by RenaGao. Read the original source.

This expanded guide is an AI-narrated adaptation of the source’s essential explanation, examples and caveats, not a word-for-word reading. The synthetic voice does not imitate the author or imply their endorsement.

License: MIT

MIT License

Copyright (c) 2026 Rena Gao

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Our summary

This project provides a step-by-step process for preparing information to teach an AI assistant, which is software that carries out tasks. People use this to sort their text files into groups before training begins, separating useful examples from text that needs human review.

The software first checks the text for basic errors and exact duplicates. Then, it sends the remaining text to Jev, an AI tool that chooses from options rather than writing an answer. Jev sorts the information into piles to keep, review, or reject based on specific rules.

This setup helps researchers measure changes in the software's text prediction scores, known as loss and perplexity, which does not prove it performs tasks better. The public website only runs a limited test; real training requires your own computer hardware and a private access code.

Key takeaways

Remove exact duplicates and basic errors on your own computer before asking Jev to sort the text.
Jev is an AI tool that evaluates the information, not the software that you are actually training.
Measure text prediction scores before and after training to see how the software changed.

More guides →See the experiments →