用对比学习让大模型评分更公平，不偏爱啰嗦漂亮但跑题的答案。
Mitigating the Bias of Large Language Model Evaluation
- 对评分模型做校准，削弱表面流畅度的干扰
- 用精心设计的反例样本进行对比训练，提升判断准确性
- 适合关注大模型评估公正性的研究者和开发者
近期流行使用大语言模型作为评分器（LLM-as-a-Judge）来评估其他大模型的输出质量。然而，现有评分模型存在明显偏差：倾向于偏好表达更流畅、更冗长的答案，而忽视指令遵循能力。本文系统研究了这一偏差问题。针对闭源评分模型，提出在概率层和提示层进行校准，降低表面质量的影响；针对开源评分模型，提出通过对比学习，使用精心构建的负样本（偏离指令但表面质量高）进行训练，以缓解偏差。在基准评测中，所提方法显著减轻了评分偏差，同时保持了良好的评估准确率。
原文摘要 · Abstract (English)
Recently, there has been a trend of evaluating the Large Language Model (LLM) quality in the flavor of LLM-as-a-Judge, namely leveraging another LLM to evaluate the current output quality. However, existing judges are proven to be biased, namely they would favor answers which present better superficial quality (such as verbosity, fluency) while ignoring the instruction following ability. In this work, we propose systematic research about the bias of LLM-as-a-Judge. Specifically, for closed-source judge models, we apply calibration to mitigate the significance of superficial quality, both on probability level and prompt level. For open-source judge models, we propose to mitigate the bias by contrastive training, with curated negative samples that deviate from instruction but present better superficial quality. We apply our methods on the bias evaluation benchmark, and experiment results show our methods mitigate the bias by a large margin while maintaining a satisfactory evaluation accuracy.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。