SGAgent: Suggestion-Guided LLM-Based Multi-Agent Framework for Repository-Level Software Repair

2026-02-27

AI Summary

This work proposes SGAgent, a novel agent-based approach to automated program repair that addresses the limitations of conventional two-stage “locate-then-repair” paradigms by introducing an intermediate “suggestion” phase, thereby enhancing reasoning coherence between fault localization and patch generation. SGAgent integrates a knowledge graph to improve repository-level contextual understanding and comprises three collaborative sub-agents—Locator, Suggestor, and Repairer—that jointly leverage large language models and knowledge graph retrieval for end-to-end repair. Evaluated on SWE-Bench, SGAgent achieves a 51.3% repair accuracy with file- and function-level localization accuracies of 81.2% and 52.4%, respectively, at a cost of $1.48 per instance. It also attains 48% accuracy on both VUL4J and VJBench, demonstrating strong cross-task and cross-language generalization capabilities.

Technology Category

• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Teamwork

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

The rapid advancement of Large Language Models (LLMs) has led to the emergence of intelligent agents capable of autonomously interacting with environments and invoking external tools. Recently, agent-based software repair approaches have received widespread attention, as repair agents can automatically analyze and localize bugs, generate patches, and achieve state-of-the-art performance on repository-level benchmarks. However, existing approaches usually adopt a localize-then-fix paradigm, jumping directly from"where the bug is"to"how to fix it", leaving a fundamental reasoning gap. To this end, we propose SGAgent, a Suggestion-Guided multi-Agent framework for repository-level software repair, which follows a localize-suggest-fix paradigm. SGAgent introduces a suggestion phase to strengthen the transition from localization to repair. The suggester starts from the buggy locations and incrementally retrieves relevant context until it fully understands the bug, and then provides actionable repair suggestions. Moreover, we construct a Knowledge Graph from the target repository and develop a KG-based toolkit to enhance SGAgent's global contextual awareness and repository-level reasoning. Three specialized sub-agents (i.e., localizer, suggester, and fixer) collaborate to achieve automated end-to-end software repair. Experimental results on SWE-Bench show that SGAgent with Claude-3.5 achieves 51.3% repair accuracy, 81.2% file-level and 52.4% function-level localization accuracy with an average cost of $1.48 per instance, outperforming all baselines using the same base model. Furthermore, SGAgent attains 48% accuracy on VUL4J and VJBench for vulnerability repair, demonstrating strong generalization across tasks and programming languages.

Problem

• software repair
• reasoning gap
• repository-level
• bug localization
• LLM-based agents

Innovation

• Suggestion-Guided Repair
• Multi-Agent Framework
• Knowledge Graph
• Repository-Level Reasoning
• Automated Program Repair