LLM Agent for Issue-to-Commit Link Recovery
A new autonomous LLM-based agent, LinkAnchor, addresses the challenge of recovering links between issues and commits in software repositories. Only 42.2% of GitHub issues are correctly linked to commits. Existing ML/DL and LLM approaches face limitations: restricted context windows prevent processing all data sources (commit histories, issue discussions, code repositories), and pairwise scoring fails to capture complex associativity of software fixes. LinkAnchor overcomes these by operating as an agent that iteratively gathers and synthesizes information from multiple sources. The system is designed to improve software traceability and project management by autonomously recovering issue-to-commit links with higher accuracy. The paper is available on arXiv (2508.12232).
Key facts
- LinkAnchor is an autonomous LLM-based agent for issue-to-commit link recovery.
- Only 42.2% of GitHub issues are correctly linked to commits.
- Existing LLM approaches are limited by context windows and pairwise scoring.
- LinkAnchor processes multiple data sources including commit histories and issue discussions.
- The agent iteratively gathers and synthesizes information.
- The goal is to improve software traceability and project management.
- The paper is published on arXiv with ID 2508.12232.
Entities
Institutions
- arXiv
- GitHub