LLM Log Rules
You may use language models such as GPT, Claude, Mixtral, LLaMA, etc. for projects in this class, subject to the following rules:
- You may not attempt to obtain the entire project solution in a single query.
- Any code within your project that is sourced from a language model must be clearly marked as such with attribution comments.
- All queries and responses made to a language model for the purposes of the project must be submitted alongside the program. This applies even if code/info generated by the language model is not used in the final submission.
This document contains instructions for what to include in the language model logs.
If you did not use a language model #
Include with your project submission a file called lm_log.txt that has the
following text:
I did not use a language model for this project.
If you did use a language model #
Download the language model in whatever plaintext format you can. Many language model aggregators such as Poe should support export of conversations. If you are using one of the major language models like GPT or Claude, there are typically browser plugins or standalone applications that allow you to export conversations.
Acceptable plaintext formats include:
- Text files (
.txt) - Markdown files (
.md) - HTML sources (
.html)
Do not use PDF or Word formats. Place this in a file named lm_log with the
appropriate file extension.
In addition, if you are using an OpenAI model like GPT, place a shareable link to the conversation at the top of your log file. Make sure you keep the shared conversation available until grades are submitted.
Submit the lm_log file in your project ZIP.