For evals, what is OpenAI thoughts for different sets of evals for source code, e.g. comprehensible , optimized, etc.?

[EricGT]

This request is directed towards the esteemed individuals at OpenAI.

---

When creating source code there may be different priorities for creating the code

• class assignment - comprehensible and idomatic
• production - safe and fast
• pure

In considering creating evals for Prolog was wondering if OpenAI would accept different sets?

For example the N-queens puzzle is the 8-queens puzzle but for any size rectangualar board. While there are many ways to solve this some are comprehensible, some are very fast, and some are pure. Many have more than one triat but AFAIK no solution has all of them.

References:
https://www.metalevel.at/queens/
https://groups.google.com/g/comp.lang.prolog/c/qiyibDALhTE/m/uk6f6AQzOCAJ
https://stackoverflow.com/a/53412988/1243762

By submitting various sets, the model can potentially learn to customize the generated code to match the desired characteristic.

Thank you for your consideration.