XAI for Coding Agent Failures: Transforming Raw Execution Traces into Actionable Insights

Researchers developed an explainable AI (XAI) system that transforms raw execution traces from LLM-based coding agents into structured, human-interpretable explanations. The system enables users to identify failure root causes 2.8 times faster and propose fixes with 73% higher accuracy through domain-specific failure taxonomy, automatic annotation, and hybrid explanation generation.

• →LLM-based coding agents frequently fail in ways that are difficult for developers to understand and debug.
• →The new XAI approach uses three components: failure taxonomy, automatic annotation system, and hybrid explanation generator.
• →User study with 20 participants showed 2.8x faster root cause identification and 73% higher fix accuracy.
• →The structured approach outperforms ad-hoc explanations by providing consistent, domain-specific insights with visualizations.
• →The framework addresses the critical need for interpretable AI systems in software development workflows.