代码作者归属从竞赛到课堂不具泛化性
原标题：Source Code Authorship Attribution Does Not Generalize from Competitions to Classrooms
AI 摘要
一项研究微调预训练Transformer模型CodeBERT，在Google Code Jam竞赛数据上达到92.6%的Top-1准确率（10位作者），但在大学课程作业数据集上表现仅与随机基线相当（Top-1准确率0.2%或更低）。多模型基准测试显示该性能差距普遍存在，表明基于竞赛数据的基准高估了代码作者归属在教育场景中的实际适用性。
正文节选
Computer Science > Software Engineering Title:Source Code Authorship Attribution Does Not Generalize from Competitions to Classrooms View PDF HTML (experimental) Abstract:Source code authorship attribution aims to identify the author of a program fragment from its writing style. We fine-tune the pre-trained transformer CodeBERT on three sources of data: publicly available Google Code Jam (GCJ) submissions from an open Kaggle repository, a curated GCJ archive, and institutional course