MODERN TRENDS AND CHALLENGES IN DEBUGGING SOFTWARE BASED ON LARGE LANGUAGE MODELS
DOI:
https://doi.org/10.34185/1991-7848.itmm.2025.01.039Keywords:
Large Language Models (LLMs), software debugging, automated debugging, interactive assistants, development environments, human-AI interaction, LLM limitations, software developers.Abstract
This report presents the results of research on the application of large language models (LLMs) in the field of software debugging. The current state of research in this area is examined, including both promising directions and existing problems that limit the widespread practical application of such model usage. Various approaches to using LLMs in the debugging process are analyzed. Particular attention is paid to the integration of LLMs into development environments as interactive assistants that work in close collaboration with the developer. It is argued that the effective use of LLMs for debugging requires a comprehensive approach that considers both the development of the models themselves and the improvement of developers' skills to ensure productive human-AI interaction. The work aims to identify optimal ways of applying LLMs in the field of software debugging, considering current technological capabilities and the needs of developers.
References
A survey on large language models for software engineering / Q. Zhang et al. arXiv.org. URL: https://doi.org/10.48550/arXiv.2312.15223 (дата звернення: 27.03.2025).
Large language models for software engineering: survey and open problems / A. Fan et al. 2023 IEEE/ACM international conference on software engineering: future of software engineering (icse-fose), Melbourne, Australia, 14–20 May 2023. 2023.
URL: https://doi.org/10.1109/icse-fose59343.2023.00008 (дата звернення: 25.03.2025).
A comprehensive survey of ai-driven advancements and techniques in automated program repair and code generation / A. Anand et al. arXiv.org e-Print archive.
URL: https://doi.org/10.48550/arXiv.2411.07586 (дата звернення: 25.03.2025).
Zhang Y. Chain of Agents: large language models collaborating on long-context tasks. Google Research - Explore Our Latest Research in Science and AI.
URL: https://research.google/blog/chain-of-agents-large-language-models-collaborating-on-long-context-tasks/ (дата звернення: 25.03.2025).
A deep dive into large language model code generation mistakes: what and why? / Q. Chen et al. arXiv.org. URL: https://doi.org/10.48550/arXiv.2411.01414 (дата звернення: 26.03.2025).
LLMs love python: a study of llms’ bias for programming languages and libraries / L. Twist et al. arXiv.org e-Print archive. URL: https://arxiv.org/html/2503.17181v1#S6 (дата звернення: 27.03.2025).
Teaching large language models to self-debug / X. Chen et al. arXiv.org.
URL: https://doi.org/10.48550/arXiv.2304.05128 (дата звернення: 26.03.2025).
Training LLMs to better self-debug and explain code / N. Jiang et al. Amazon Science. URL: https://www.amazon.science/publications/training-llms-to-better-self-debug-and-explain-code (дата звернення: 26.03.2025).
Zhong L., Wang Z., Shang J. Debug like a human: a large language model debugger via verifying runtime execution step by step. Findings of the association for computational linguistics ACL 2024, Bangkok, Thailand and virtual meeting. Stroudsburg, PA, USA, 2024. P. 851–870. URL: https://doi.org/10.18653/v1/2024.findings-acl.49 (дата звернення: 26.03.2025).
InferFix: End-to-End Program Repair with LLMs / M. Jin et al. ESEC/FSE '23: 31st ACM joint european software engineering conference and symposium on the foundations of software engineering, San Francisco CA USA. New York, NY, USA, 2023. URL: https://doi.org/10.1145/3611643.3613892 (дата звернення: 26.03.2025).
Extending context window of large language models via semantic compression / W. Fei et al. Findings of the association for computational linguistics ACL 2024, Bangkok, Thailand and virtual meeting. Stroudsburg, PA, USA, 2024. P. 5169–5181.
URL: https://doi.org/10.18653/v1/2024.findings-acl.306 (дата звернення: 26.03.2025).
Shiraishi M., Shinagawa T. Context-aware code segmentation for c-to-rust translation using large language models. arXiv.org e-Print archive.
URL: https://arxiv.org/html/2409.10506v1#S8 (дата звернення: 26.03.2025).
Enhancing the code debugging ability of llms via communicative agent based data refinement / W. Yang et al. arXiv.org e-Print archive.
URL: https://arxiv.org/html/2408.05006v1 (дата звернення: 27.03.2025).
KBLaM: knowledge base augmented language model / X. Wang et al. arXiv.org.
URL: https://doi.org/10.48550/arXiv.2410.10450 (дата звернення: 27.03.2025).
Haque M. A. LLMs: a game-changer for software engineers?. arXiv.org.
URL: https://doi.org/10.48550/arXiv.2411.00932 (дата звернення: 27.03.2025).
CodeAgent: enhancing code generation with tool-integrated agent systems for real-world repo-level coding challenges / K. Zhang et al. Proceedings of the 62nd annual meeting of the association for computational linguistics (volume 1: long papers), Bangkok, Thailand. Stroudsburg, PA, USA, 2024. P. 13643–13658. URL: https://doi.org/10.18653/v1/2024.acl-long.737 (дата звернення: 27.03.2025)
Alibaba lingmaagent: improving automated issue resolution via comprehensive repository exploration / Y. Ma et al. arXiv.org.
URL: https://doi.org/10.48550/arXiv.2406.01422 (дата звернення: 27.03.2025).
Tarassow A. The potential of LLMs for coding with low-resource and domain-specific programming languages. arXiv.org. URL: https://doi.org/10.48550/arXiv.2307.13018
(дата звернення: 27.03.2025).
Human-AI experience in integrated development environments: a systematic literature review / A. Sergeyuk та ін. arXiv.org e-Print archive.
URL: https://arxiv.org/html/2503.06195v1#S5 (дата звернення: 01.04.2025).
