FlexFL introduces a flexible framework for software bug localization using open-source large language models, eliminating dependency on proprietary AI systems while maintaining effectiveness.

• Leverages multiple bug information sources beyond just test cases, including bug reports
• Achieves comparable performance to proprietary LLM-based fault locators
• Provides practical flexibility for real-world engineering scenarios
• Addresses privacy and cost concerns by using open-source models

This research empowers engineering teams to implement robust fault localization within their security protocols without sacrificing data privacy or incurring high API costs associated with commercial models.

FlexFL: Flexible and Effective Fault Localization with Open-Source Large Language Models