Advances and Frontiers of LLM-based Issue Resolution in Software Engineering: A Comprehensive Survey
Caihua Li, Lianghong Guo, Yanlin Wang, Daya Guo, Wei Tao, Zhenyu Shan, Mingwei Liu, Jiachi Chen, Haoyu Song, Duyu Tang, Hongyu Zhang, Zibin Zheng
2026-01-21
Summary
This paper is a broad overview of how artificial intelligence is being used to automatically fix problems in software code, a task known as issue resolution.
What's the problem?
Fixing bugs and issues in software is a really hard task, even for experienced programmers. Recent tests showed that even the most advanced AI language models struggle with this kind of work, meaning they aren't yet capable of independently building and maintaining software.
What's the solution?
The paper breaks down the different ways researchers are trying to tackle this problem. Some approaches don't require any extra training of the AI, instead using pre-built tools and modules. Others involve specifically training the AI using examples of code issues and their fixes, either by showing it correct solutions or letting it learn through trial and error. The paper also looks at how to make sure the data used to train these AIs is good quality and how to understand *why* an AI makes a certain fix.
Why it matters?
Automating issue resolution could dramatically speed up software development and reduce costs. If AI can reliably fix bugs, programmers can focus on more creative tasks. This research is important because it identifies the current state of the field, the challenges that remain, and where future research should be focused, and provides a resource for others to build upon.
Abstract
Issue resolution, a complex Software Engineering (SWE) task integral to real-world development, has emerged as a compelling challenge for artificial intelligence. The establishment of benchmarks like SWE-bench revealed this task as profoundly difficult for large language models, thereby significantly accelerating the evolution of autonomous coding agents. This paper presents a systematic survey of this emerging domain. We begin by examining data construction pipelines, covering automated collection and synthesis approaches. We then provide a comprehensive analysis of methodologies, spanning training-free frameworks with their modular components to training-based techniques, including supervised fine-tuning and reinforcement learning. Subsequently, we discuss critical analyses of data quality and agent behavior, alongside practical applications. Finally, we identify key challenges and outline promising directions for future research. An open-source repository is maintained at https://github.com/DeepSoftwareAnalytics/Awesome-Issue-Resolution to serve as a dynamic resource in this field.