偏好泄露：作为判断者的LLM中的污染问题
Preference Leakage: A Contamination Problem in LLM-as-a-judge
AI总结：
本文提出偏好泄露问题，揭示LLM作为判断者时因数据生成器与评估者相关性导致的偏见，并通过实验验证其普遍存在性和检测难度。
摘要：
大型语言模型（LLM）作为判断者和基于LLM的数据合成已成为模型开发中两种基本的LLM驱动的数据标注方法。尽管它们的结合显著提高了模型训练和评估的效率，但很少有人关注这种新模型开发范式带来的潜在污染。在本工作中，我们揭示了偏好泄露，这是由数据生成器LLM与基于LLM的评估者之间的相关性导致的LLM作为判断者中的污染问题。为了研究这一问题，我们首先定义了数据生成器LLM与判断者LLM之间的三种常见相关性：同为模型、具有继承关系、以及同属同一模型家族。通过广泛的实验，我们实证地确认了由于偏好泄露，判断者对相关学生模型的偏见在多个LLM基线和基准测试中普遍存在。进一步分析表明，偏好泄露是一个普遍存在且现实世界中的问题，比之前识别的LLM作为判断者场景中的偏见更难检测。所有这些发现表明，偏好泄露是LLM作为判断者领域中的广泛而具有挑战性的问题。我们发布了所有代码和数据：https://github.com/David-Li0406/Preference-Leakage。
英文摘要：
Large Language Models (LLMs) as judges and LLM-based data synthesis have emerged as two fundamental LLM-driven data annotation methods in model development. While their combination significantly enhances the efficiency of model training and evaluation, little attention has been given to the potential contamination brought by this new model development paradigm. In this work, we expose preference leakage, a contamination problem in LLM-as-a-judge caused by the relatedness between the synthetic data generators and LLM-based evaluators. To study this issue, we first define three common relatednesses between the data generator LLM and the judge LLM: being the same model, having an inheritance relationship, and belonging to the same model family. Through extensive experiments, we empirically confirm the bias of judges towards their related student models caused by preference leakage across multiple LLM baselines and benchmarks. Further analysis suggests that preference leakage is a pervasive and real-world problem that is harder to detect compared to previously identified biases in LLM-as-a-judge scenarios. All of these findings imply that preference leakage is a widespread and challenging problem in the area of LLM-as-a-judge. We release all codes and data at: https://github.com/David-Li0406/Preference-Leakage.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
