A Survey of LLM-based Automated Program Repair: Taxonomies, Design Paradigms, and Applications

2025-06-30

AI Summary

This study systematically reviews 63 LLM-based automated program repair (APR) systems published between January 2022 and June 2025, addressing three core challenges: semantic correctness verification beyond test suites, large-scale repository-level defect repair, and optimization of LLM inference cost. We propose the first comprehensive taxonomy, categorizing APR designs into four paradigms: fine-tuning, prompt engineering, pipeline-based workflows, and agent frameworks. Quantitative analysis demonstrates how retrieval augmentation and static/dynamic code analysis enhance context quality, while revealing fundamental trade-offs among cost, controllability, and scalability across paradigms. Key insights identify lightweight feedback mechanisms, repository-aware retrieval, and cost-aware planning as critical levers for advancement. The work establishes a theoretical framework and practical roadmap to enhance the reliability, scalability, and real-world applicability of LLM-APR systems.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large language models (LLMs) are reshaping automated program repair (APR). We categorize the recent 63 LLM-based APR systems published from January 2022 to June 2025 into four paradigms, and show how retrieval- or analysis-augmented contexts strengthen any of them. This taxonomy clarifies key trade-offs: fine-tuning delivers strong task alignment at high training cost; prompting enables rapid deployment but is limited by prompt design and context windows; procedural pipelines offer reproducible control with moderate overhead; agentic frameworks tackle multi-hunk or cross-file bugs at the price of increased latency and complexity. Persistent challenges include verifying semantic correctness beyond test suites, repairing repository-scale defects, and lowering the costs of LLMs. We outline research directions that combine lightweight human feedback, repository-aware retrieval, code analysis, and cost-aware planning to advance reliable and efficient LLM-based APR.

Problem

• Categorizing LLM-based APR systems into four paradigms
• Addressing challenges in verifying semantic correctness beyond test suites
• Advancing reliable and efficient LLM-based APR with lightweight human feedback

Innovation

• LLM-based APR systems categorized into four paradigms
• Retrieval or analysis-augmented contexts enhance paradigms
• Lightweight human feedback and repository-aware retrieval