一句话：提出区分贡献与认可的开发者贡献度量框架，结合预测建模、强化学习与异常检测以提升公平性和抗操纵性。
现有软件贡献评估常将贡献与认可混为一谈，在 AI 辅助开发兴起后更易受偏见和操纵影响。本文提出一个结构化框架，明确区分贡献与认可，将预测建模与强化学习结合，并引入异常检测，以提升评估的公平性与鲁棒性。该框架随时间动态精炼贡献分数，能够适应延迟出现的认可信号，同时抑制人为夸大。摘要未报告实证结果，未来工作将聚焦于实证验证、抗操纵能力以及为开源和企业软件开发改进评估透明度。
Evaluating software contributions remains a challenge, especially with the rise of AI-assisted development. Existing methods often conflate contributions with recognition, making assessments susceptible to bias and manipulation. This paper proposes a structured framework that distinguishes contributions from recognition, integrates predictive modeling with reinforcement learning, and incorporates anomaly detection to ensure fairness and robustness. By dynamically refining contribution scores over time, the framework adapts to delayed recognition signals while mitigating artificial inflation. Future work will focus on empirical validation, resistance to manipulation, and improving assessment transparency for both open-source and enterprise software development.
DOI 原文 ·
@inproceedings{paperbot9010,
title = {Beyond Stars and Commits: A Fair and Manipulation-Resistant Framework for Measuring Developer Contributions},
author = {Yuqiang Sun and Sun Ding and Liwei Tan and Yue Xue and Bo Li and Charles Huang and Siyang Xiong},
booktitle = {FSE},
year = {2025},
doi = {10.1145/3696630.3730570}
}