What makes a commit message precise, informative, and consistent with a project's language and style? CoRaCMG explores how contextual retrieval of similar historical code changes and their human-written commit messages can help LLMs generate higher-quality commit messages across different models.
