Aug 2026· International journal of computer information systems and industrial management applications· 0 citations
TL;DR
The AgentCodeReview system is presented, a multi-agent system that is able to conduct explainable code review and automated bug repair by leveraging software engineering agents with different code review tasks and its utility and extensibility to the field of explainable AI in software quality assurance are demonstrated.
Abstract
Large Language Models (LLMs) have revolutionized software development, from analyzing code and generating suggestions to detecting bugs and errors, and even creating entire programs. Despite these advances, existing AI-driven code review solutions still provide a one-size-fits-all approach to code review with overall feedback and suggestions, often of a non-specific nature. This restriction promotes modular architectures which would be able to provide specific and direct code quality reports. This paper presents the AgentCodeReview system, a multi-agent system that is able to conduct explainable code review and automated bug repair by leveraging software engineering agents with different code review tasks. There would be five independent entities, each one to be able to review code, analyze security, evaluate performance, document it and be able to automatically fix bugs. They run parallelly under the guidance of a centralized orchestration layer that collects the results from the analytical agents, calculates software quality scores and creates comprehensive HTML and PDF reports. Moreover, a Streamlit-based web interface was created that allows the interactive visualization of the results of the analysis and interactive entry of the input values. A set of twenty python programs was created to test the framework for effectiveness, consisting of a variety of runtime errors, security flaws, performance issues, documentation issues and a mixture of these types of errors. Two metrics, namely execution time and qualitative assessment were used to compare the proposed multi-agent framework with a single-agent framework as baseline. Experimental results demonstrated the benchmark execution success rate was 95%, while the multi-agent architecture provided more structured, explainable and domain specific feedback than the single agent. The extra computational cost of the coordinated analyses was acceptable for software quality assessment tasks because of the resulting interpretability and modularity. Through implementation and experiments, the results demonstrate AgentCodeReview's utility and extensibility to the field of explainable AI in software quality assurance. The proposed architecture can be expanded to other programming languages, integrated into the industrial development flow, and enhanced with the advanced LLMs for scalable intelligent code review.
Modern software development depends heavily on code review and timely bug fixing, yet manual review is slow, inconsistent and hard to scale, while single-model automated approaches based on Large Language Models (LLMs) frequently produce opaque suggestions and rarely close the loop between detecting a defect and repairing it. This paper proposes AgentCodeReview, a multi-agent framework that decomposes explainable code review and automated bug repair into cooperating specialized agents — Retrieval, Review, Explanation, Repair and Verification — coordinated by an Orchestrator over a shared retrieval-augmented context store, with a verification-driven feedback loop that iteratively refines candidate patches until they pass their tests. We describe the architecture, the inter-agent communication protocol and an evaluation design over widely used code-review and program-repair benchmarks, comparing the framework against single-agent and non-agentic baselines using precision, recall, F1, repair success rate and a human-aligned explanation-quality score. The results indicate that role specialization and explicit verification improve review accuracy, repair effectiveness and the transparency of the generated rationales, offering a reproducible pathway toward trustworthy autonomous software maintenance.
Bharath Kumar N, T L Manas· International journal of com...· 0 citations
TestAgent, a multi-agent tool implemented as a VS Code extension that automates the generation of high-quality unit tests for Java projects using repository-level Code Knowledge Graphs, demonstrates its practical utility for regression testing and bug discovery.
Ye Shang, Quanjun Zhang, Zheng Zhan et al.· SIGSOFT FSE Companion· 0 citations
Large Language Models (LLMs) have demonstrated remarkable potential in automated code generation. However, existing test-driven code generation and refinement frameworks are often hindered by the tests' quality: they typically treat self-generated tests as ground truth, leading to ineffective debugging loops where code is modified to satisfy erroneous tests. To address this, we propose BiVCoder, a diagnosis-driven multi-agent framework featuring a novel bidirectional code-test diagnosis mechanism. BiVCoder coordinates three specialized agents—Coding Agent, Test Agent, and Review Agent. Central to this architecture is the Review Agent, which serves as a diagnosis and decision-making hub. By integrating an MCP-based code-test execution tool, the Review Agent rigorously executes programs to diagnose failure root causes, distinguishing between implementation bugs and test case deficiencies, and subsequently triggers targeted repairs (Coding Agent or Test Agent). Furthermore, we introduce BiVCoder-SFT, a role-specific instruction fine-tuning scheme. We construct high-quality datasets to fine-tune the Qwen3-4B base model into specialized agents for coding, testing, and reviewing. Extensive experiments on HumanEval, MBPP, and their rigorous ''ET'' variants demonstrate the superiority of our approach. With GPT-3.5, BiVCoder achieves a Pass@1 of 77.3%. Notably, the specialized BiVCoder-SFT achieves an average score of 79.5% across the four datasets, not only outperforming its base model (69.7%) but also surpassing larger general-purpose models such as Qwen3-8B (77.8%). Additionally, BiVCoder boosts the performance of more powerful models, improving DeepSeek-V3.2 from 87.2% to 92.1% on HumanEval.
Xiaoyang Li, Jinhao Dong, Wenhang Shi et al.· Proceedings of the 32nd ACM...· 0 citations
Large Language Models (LLMs) offer new opportunities for automated code refactoring. However, generated changes must reduce targeted quality problems without introducing new issues or altering behaviour-relevant code structures. We introduce REFINE (Refactoring with Evidence-aware Flow for Integrated ageNtic Execution), a tool-agnostic, evidence-aware multi-agent approach for generating Java file-level refactoring candidates. REFINE combines static-analysis-guided smell identification, smell-informed planning, LLM-based transformation, automated re-analysis, preservation checks, and structured reporting. We evaluate REFINE on 450 Java files from 15 open-source systems, producing 1,350 model-pass outputs using OpenAI GPT-5.5, Google Gemini 3.1 Pro Preview, and Anthropic Claude Opus 4.8. REFINE reduces detected code smells by 68.26%, 72.79%, and 68.49% across the three configurations, respectively, with the strongest reductions observed for major smells. A matched 150-file direct-prompt baseline shows that REFINE achieves a higher median code-smell reduction with smaller edits and fewer public-method removals. However, broader quality improvements are inconsistent, and preservation checks reveal residual risks, including assert/fail-call changes and public-method removal. Therefore, REFINE outputs should be treated as refactoring candidates requiring compilation, testing, dependency analysis, and human review before adoption in repository- or system-level settings.
Muhammad Waseem, Aakash Ahmad, Pekka Abrahamsson· 0 citations
A comprehensive overview of the existing tools and frameworks for implementing MAS in software engineering and a set of lessons learned and challenges that can help researchers and practitioners to select a suitable MAS framework according to their needs are provided.
Mariama Celi Serafim De Oliveira, M. Ibiyo, Marco Gianrusso et al.· 0 citations
The first empirical study focused on agent-reactive (AR) bugs is conducted, constructing a two-axis taxonomy covering observable symptoms and the LLM behaviors that trigger them and highlights challenges specific to LLM agents.
Jingyi Chen, Songqiang Chen, Hengcheng Zhu et al.· 0 citations