一句话：面向代码变更理由的细粒度、个性化、任务定制化自动生成，支撑代码评审、调试与新人上手等维护任务。
理解代码变更背后的理由（rationale）对软件维护至关重要，但文档碎片化、缺少清晰可执行的解释，使开发者难以弄清“为何改”。本博士研究旨在提供细粒度、个性化、任务定制的变更理由解释，包含三项工作：（1）抽取并生成跨多文档的细粒度理由摘要；（2）建模开发者画像，依据其专业水平与熟悉度裁剪解释；（3）面向代码评审、调试、新人上手等维护活动生成任务特定的理由。预期贡献包括新的实证洞见、一个跨工件且细粒度的理由基准，以及一个支持个性化、任务感知理由生成的自动化工具，以增进开发者理解并支撑软件维护。
Understanding the reasons(rationale) behind code changes is essential for effective software maintenance, yet developers often struggle with fragmented documentation and the absence of clear, actionable rationale explanations. This research aims to bridge that gap by enabling fine-grained, personalized, and task-specific explanations of why code changes occur. We (1) develop techniques to extract and generate multi-document, fine-grained rationale summaries, (2) model developer profiles to tailor explanations to their expertise and familiarity, and (3) generate task-specific rationales for activities such as code review, debugging, and onboarding. Together, this work contributes new empirical insights, a cross-artifact and fine-grained rationale benchmark, and an automated tool for personalized, task-aware rationale generation to advance developer understanding and support software maintenance.