Analysis of Research Status in the Field of Automated Program Repair
DOI:
https://doi.org/10.61173/3k7v9734Keywords:
Automated Program Repair (APR), Large Language Model (LLM), Code RepairAbstract
As software systems keep developing and becoming more widely used, it’s impossible to avoid the program bugs that come with them. To reduce the pressure on developers when modifying programs and make bug fixes more efficient, Automated Program Repair (APR) plays a key role. This paper summarizes the main research progress in the field of APR, analyzes in detail the features, development process, and performance of various methods applied in this field. By comparing the advantages and disadvantages of these methods, it also discusses the challenges faced in the APR field and the possible directions for its future development. After discussion and comparison, it can be found that APR technology has made noticeable progress—especially the methods based on Large Language Models (LLMs) and hybrid agent methods. However, core challenges still exist, such as limitations of datasets and insufficient ability of models to understand code. In the future, it will be necessary to build more representative datasets, improve models’ ability to understand complex code logic, optimize the cooperation mechanism between multiple models, and reduce reliance on specific tools.
References
[1] Dai Z, Chen B, Zhao Z, Tang X, Wu S, Yao C, ... & Chen J. Less is More: Adaptive Program Repair with Bug Localization and Preference Learning. In Proceedings of the AAAI Conference on Artificial Intelligence. 2025. (Vol. 39, No. 1, pp. 128-136).
[2] Dikici S, Bilgin T T. Advancements in automated program repair: a comprehensive review. Knowledge and Information Systems, 2025: 1-47.
[3] Ji T, Qi Y H & Mao X G. Tool based on software automatic repair evaluation defect localization technology: GenProg-FL. Computer Science, 2014, 41(09), 88-91 124.
[4] Jiang J J, Xiong Y F & Xia X. A manual inspection of Defects4J bugs and its implications for automatic program repair. Science China (Information Sciences), 2019, 62(10), 31- 46.
[5] Yu B X, Zhu Y X, He P J, and Kang D. UTBoost: Rigorous Evaluation of Coding Agents on SWE-Bench. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3762–3774, Vienna, Austria. Association for Computational Linguistics. 2025.
[6] Ahmed E. AI Copilots: Boosting Software Engineers’ Productivity or Replacing Them? SIGSOFT Softw. Eng. Notes 50, 3 2025, 12–13. https://doi.org/10.1145/3743095.3743098
[7] Tian Y, Yan W, Yang Q, Zhao X, Chen Q, Wang W, Luo Z, Ma L, & Song D. CodeHalu: Investigating Code Hallucinations in LLMs via Execution-based Verification. Proceedings of the AAAI Conference on Artificial Intelligence, 39(24), 25300- 25308. 2025.
[8] Liu Y, Zhang L, Zhang Z. A Survey of Test Based Automatic Program Repair. J. Softw., 2018, 13(8): 437-452.
[9] Winter E, Nowack V, Bowes D, et al. Let’s talk with developers, not about developers: A review of automatic program repair research. IEEE Transactions on Software Engineering, 2022, 49(1): 419-436.
[10] Puri R, Kung D S, Janssen G, Zhang W, Domeniconi G, Zolotov V, ... & Reiss F. Codenet: A large-scale ai for code dataset for learning a diversity of coding tasks. arxiv preprint arxiv:2105.12655. 2021.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 by the authors.

This work is licensed under a Creative Commons Attribution 4.0 International License.
