BLAgent: Agentic RAG for File-Level Bug Localization

TL;DR

BLAgent is a novel agentic RAG framework that significantly improves file-level bug localization accuracy and efficiency by integrating code structure-aware encoding, dual-perspective queries, and two-phase reranking.

Contribution

It introduces a new agentic RAG approach for bug localization that balances accuracy and cost, outperforming prior methods on SWE-bench Lite.

Findings

Achieves over 78% Top-1 accuracy with open-source models.

Attains over 86% accuracy with a closed-source model.

Reduces cost by over 18x compared to the strongest baseline.

Abstract

Bug localization remains a key bottleneck in downstream software maintenance tasks, including root cause analysis, triage, and automated program repair (APR), despite recent advances in large language model (LLM)-based repair systems. File-level bug localization is especially critical in hierarchical pipelines, where errors can propagate to downstream stages such as statement-level localization or patch generation. While Retrieval-Augmented Generation (RAG) offers a promising direction for grounding LLMs in repository context, existing RAG pipelines rely on static retrieval and lack the reasoning needed to identify faulty code accurately. In this work, we present BLAgent, a novel agentic RAG framework for file-level bug localization that integrates three key ideas: (i) code structure-aware repository encoding with path-augmented AST-based chunking, (ii) dual-perspective query transformation capturing both structural and behavioral signals, and (iii) two-phase agentic reranking combining symbolic inspection with evidence-grounded reasoning. Unlike prior graph-based or multi-hop agentic approaches, BLAgent performs bounded reasoning over a compact candidate set, balancing accuracy and cost. On SWE-bench Lite, BLAgent attains over 78% Top-1 accuracy with open-source models and over 86% with a closed-source model, while being over 18x cheaper than the strongest baseline using the same model. When integrated into an APR framework, it improves end-to-end repair success by over 20%.