-
Notifications
You must be signed in to change notification settings - Fork 44.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Rename evaluate_code
to analyze_code
#1371
Conversation
Handles Docker errors separately, and prints a potentially helpful message for users.
@willcallender There are conflicts now |
This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request. |
Sorry, but where can I see the conflicts? Normally I'd use the resolve conflicts button but it's greyed out for me. |
Conflicts have been resolved! 🎉 A maintainer will review the pull request shortly. |
I've tested the changes and they still work as expected. |
This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good to me
Conflicts have been resolved! 🎉 A maintainer will review the pull request shortly. |
ChatGPT is less confused by this phrasing From my own observations and others (ie Significant-Gravitas#101 and Significant-Gravitas#286) ChatGPT seems to think that `evaluate_code` will actually run code, rather than just provide feedback. Since changing the phrasing to `analyze_code` I haven't seen the AI make this mistake. --------- Co-authored-by: Reinier van der Leer <[email protected]>
Background
As seen in #101 and #286, GPT-3.5 misinterprets "evaluate" to mean "execute" rather than "analyze". As such, I changed the name of the function to "analyze" to make the purpose of the function more clear to the AI.
Changes
Renamed the
evaluate_code
command toanalyze_code
everywhere it appears, including the function names and the text as it's given to the AI.Documentation
I tested with several simple prompts, namely hello world programs in various languages, and I haven't seen this error since making the change.
Test Plan
I used the same basic AI personality for these tests, specifically I used Dev-GPT, an AI designed to autonomously develop, run, and test code. I gave it a single goal which was "Write and run a simple hello world program in [language]." I tested Python and Rust, it didn't always succeed but it never got confused by
evaluate_code
.PR Quality Checklist