一句话：通过适配与微调让大语言模型无缝融入 IDE，并按项目特征与开发者风格个性化代码生成。
本文关注大语言模型在软件开发中日益重要、但其与开发工具及集成开发环境（IDE）的集成仍碎片化且不理想的问题。研究目标是通过适配与微调现有大语言模型，而非从头构建新模型，使其能无缝地与现有开发工具和 IDE 交互。作者与 JetBrains 合作，重点研究代码生成的个性化：根据项目的独特特征以及每位开发者的个人编码风格调整模型。该方法有望把 LLM 能力更自然地嵌入日常开发流程，提升代码生成的相关性与实用性。摘要未报告具体实验设置、基准或量化结果，因此评估细节尚不明确。
Large language models have emerged as one of the most impactful applications in software engineering. However, their integration into the software development ecosystem remains fragmented and suboptimal. This study aims to bridge this gap by adapting and fine-tuning existing large language models for real-world software development scenarios. Rather than building new models from scratch, our approach leverages advanced techniques that allow Large Language Models to interact seamlessly with existing development tools and Integrated Development Environments (IDEs). This research, conducted in collaboration with JetBrains, focuses on personalizing code generation by adjusting the model to the unique features of a project and the individual coding style of each developer.