Skip to content

llm-policy.md

Latest commit

 

History

History
60 lines (52 loc) · 3.22 KB

llm-policy.md

File metadata and controls

60 lines (52 loc) · 3.22 KB
layout default
title LLM policy
description LLM usage policy in urunc

As a CNCF Sandbox project, urunc adheres to the Linux Foundation Generative AI Policy. All members of the urunc community are therefore expected to read and comply with the above policy. This document enhances the Linux Foundation policy specifically for the urunc project.

Large Language Models (LLMs) are capable of analyzing and generating massive amounts of code in a very short period of time. While this can be beneficial, like many open source projects, urunc has experienced an increase in contributions that exhibit low quality and clear signs of unreviewed LLM usage. For that reason, the urunc project has decided to enforce some rules regarding the use of LLM in its development process:

  • LLM generated code must be treated in the same way as any other code.
  • The use of LLMs for any contributions must be disclosed, including the exact model that was used.
  • The user of LLM takes full responsibility of the generated output.
  • All LLM-generated content must be reviewed by the contributor before opening issues or PRs, ensuring correctness, quality, and relevance.
  • The use of LLMs to respond to comments in issues or PRs is not permitted. Contributors are expected to express their own thoughts and ideas. Acting as a mediator between another user and an LLM is not helpful.
  • LLMs may assist in code review but can not replace human reviews. Human reviewer guidance takes precedence over any LLM comments.

The maintainers and admins of the urunc project reserve the right to close issues or PRs that do not comply with the above rules, with reference to this policy.

The rationale of the above rules

The urunc project can significantly benefit from the responsible use of LLMs and this policy is not intended to prohibit their use. Instead, it aims to protect the project from the misuse of such tools. LLM outputs are probabilistic in nature and may introduce subtle bugs, outdated practices, security risks, or incorrect assumptions. As such, any code or technical contribution produced with the assistance of an LLMs must be carefully reviewed and tested. Responsibility for the correctness and quality of LLM-generated output lies solely with the contributor who uses the tool.

Do not forget that behind the urunc project there are humans who aim to provide meaningful feedback and genuine assistance to the community. However, an influx of low-quality contributions and bug reports significantly reduces their time and energy to properly review and assist other meaningful contributions. Nevertheless, urunc is a unique and novel project with specific design choices and constraints that often require deeper explanation and discussion. As a result, LLMs may not produce helpful or accurate output for many urunc-specific use cases.

With this in mind, it would be way more beneficial to engage in discussions with other people sharing genuine ideas, thoughts and concerns over an issue, PR or in the Slack channel. There is no need to consult LLMs for every interaction, on the contrary do not hesitate to ask other members. Let's help each other learn and share ideas to improve the project.