PredicateFix introduces a novel approach to enhance LLMs in fixing security-critical code errors by leveraging bridging predicates between similar code examples.

• Addresses LLM hallucination problems when fixing complex static analysis alerts
• Improves upon standard retrieval-augmented generation (RAG) with a specialized bridging technique
• Significantly enhances fix accuracy for security vulnerabilities and code defects
• Offers a practical solution for integrating LLMs into secure software development workflows

This research matters for security teams by providing more reliable automated remediation of potential vulnerabilities identified through static analysis, reducing the security burden on developers.

PredicateFix: Repairing Static Analysis Alerts with Bridging Predicates