CoSIL introduces a novel approach that helps developers pinpoint software bugs more efficiently by combining LLMs with code repository graph searching.

• Overcomes context window limitations by creating a structured graph of the codebase
• Uses LLMs to guide intelligent navigation through relevant code paths
• Improves issue localization accuracy, enhancing automated program repair
• Balances search comprehensiveness with relevant context for better results

This research advances software engineering by enabling more efficient debugging workflows and supporting the development of more capable coding assistants, ultimately reducing development time and costs.

CoSIL: Software Issue Localization via LLM-Driven Code Repository Graph Searching