This research demonstrates how Large Language Models can automatically generate high-quality commit messages for code changes without specialized training.

• Employs in-context learning to generate descriptive commit messages from code diffs
• Eliminates the need for time-consuming model training specific to commit message generation
• Improves generalization capabilities compared to traditional approaches
• Offers practical implementation for software development teams

For engineering teams, this approach provides an efficient solution to maintain clear code documentation while reducing developer cognitive load during the commit process.

An Empirical Study on Commit Message Generation using LLMs via In-Context Learning