Loc2Repair: A Framework for Evaluating the Impact of File-Level Issue Localization in Repo-Level LLM Repair

2026-06-29 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Expert, quick

Summary

Loc2Repair is a modular evaluation framework designed for repository-grounded automated repair pipelines, specifically to analyze file-level issue localization as an upstream variable. It decouples localization and repair under a shared runtime, artifact schema, and evaluation harness, enabling researchers to combine different localization models and repair backbones. Using three repair backbones on SWE-bench Verified, the framework demonstrated that explicit localization consistently improves resolved rates. Pooled performance increased from 44.7% for baseline repair to 48.9% and 49.1% with predicted localization, and to 52.4% with gold localization. Localization also reduced mean elapsed time by 100.94 s, 52.25 s, and 154.45 s respectively, highlighting its impact on effectiveness and latency.

Key takeaway

For AI Engineers developing repository-grounded code repair systems, you should prioritize integrating explicit file-level issue localization. This approach demonstrably boosts repair success rates from 44.7% to over 48% and reduces mean repair latency by over 50 seconds, as shown by the Loc2Repair framework. Consider evaluating localization models separately to optimize your overall repair pipeline.

Key insights

File-level localization significantly enhances LLM-based code repair effectiveness and efficiency.

Principles

• Repository-grounded repair has distinct failure modes.
• Decoupling localization and repair enables controlled analysis.
• Explicit localization consistently improves repair effectiveness.

Method

Loc2Repair decouples localization and repair using a shared runtime, artifact schema, and evaluation harness to combine various localization models and repair backbones.

In practice

• Integrate explicit file localization into LLM repair pipelines.
• Evaluate localization models independently from repair backbones.

Topics

• Loc2Repair
• LLM Repair
• Code Repair
• Issue Localization
• SWE-bench
• Evaluation Framework

Best for: Research Scientist, AI Scientist, AI Engineer, Machine Learning Engineer

Related on AIssential

• PracRepair: LLM-Empowered Automated Program Repair Inspired by Human-Like Debugging Practices — PracRepair enhances LLM-based automated program repair by integrating dynamic execution and validation feedback, mimicking human debugging.
• Exploration Structure in LLM Agents for Multi-File Change Localization — Non-linear, domain-scoped parallel exploration significantly improves LLM agent performance for multi-file change localization.
• Curvature-Guided Module Localization for Low-Rank Detoxification of Backdoored Large Language Models — Curvature-guided module localization and low-rank repair effectively detoxify backdoored LLMs post hoc, preserving benign behavior.
• To Run or Not to Run: Analyzing the Cost-Effectiveness of Code Execution in LLM-Based Program Repair — Code execution in LLM-based program repair often provides minimal benefit despite significant costs, suggesting indiscriminate application.
• SWE-Explore: Benchmarking How Coding Agents Explore Repositories — SWE-Explore benchmarks coding agents' line-level repository exploration, isolating it from end-to-end repair outcomes.
• Helpful or Harmful? Evaluating LLM-Assisted Vulnerability Patching via a Human Study — LLM-assisted vulnerability patching requires rigorous security validation beyond standard functional checks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.