KG-AST Prioritizes Risky Commits for Budgeted Testing and AI Review

Summary

A study introduces KG-AST, which scores commit risk from git history and code-graph features to allocate testing and AI review budgets to the riskiest commits first. Across 155 open-source projects in 11 languages, testing the top 20% riskiest commits covers 49–64% of bug-introducing commits, outperforming random selection (20%) and biggest-change-first (40–55%). In 20 Java projects, the gain at a 20% budget is 15.1 points (64.6% vs 49.5%). An AI reviewer with graph context finds 32% of later-fixed defects exactly, and a static call-graph selector runs about 8% of a test suite while including a fix-related test in 85% of 67 fix commits.

This is a reviewed summary. Read the complete article at the original source.

Why it matters by industry

• Artificial intelligence

  Relevant to AI and software engineering: demonstrates a practical method to optimize CI/CD budgets using machine learning for commit risk scoring and AI-assisted bug review, with measurable improvements over baselines.