This research explores using Large Language Models to evaluate software quality, addressing the limitations of traditional metrics and expensive human reviews.

• LLMs can effectively judge code quality, readability, and usefulness
• Provides a cost-effective alternative to human evaluation
• Outperforms traditional metrics like BLEU that require reference solutions
• Enables scalable, consistent assessment of software artifacts

For engineering teams, this approach offers a promising path to integrate automated quality assessment into development workflows, potentially improving code review processes and software reliability.

From Code to Courtroom: LLMs as the New Software Judges