基于评分标准的大型语言模型评估中的自我偏好偏差
Self-Preference Bias in Rubric-Based Evaluation of Large Language Models
AI总结：
研究探讨了基于评分标准的评估中自我偏好偏差对模型评估的影响，发现即使评分标准客观，自我偏好仍会导致评分偏差，且通过多评委ensemble可部分缓解。
摘要：
LLM-as-a-judge已成为评估LLM输出的常用方法。然而，评委存在自我偏好偏差（SPB）：他们倾向于偏爱自己或同一家族模型的输出。这种偏差扭曲了评估，从而阻碍了模型发展，尤其是在递归自我改进的场景中。本文首次研究了基于评分标准的评估中的SPB，这是一种日益流行的基准测试范式，其中评委对个体评估标准进行二元裁定，而非分配总体分数或排名。使用具有可编程验证评分标准的IFEval基准，我们发现即使评分标准完全客观，在生成器失败的评分标准中，评委可能将50%以上的输出错误标记为满意，当输出是他们自己的时候。我们还发现，与其它评估范式类似，多个评委的ensemble有助于缓解SPB，但无法完全消除。在具有主观评分标准的HealthBench医疗聊天基准中，我们观察到SPB会使模型得分偏差高达10分，这在排序前沿模型时可能具有决定性影响。我们分析了此场景中驱动SPB的因素，发现负性评分标准、极端评分标准长度和主观主题如紧急转诊特别容易受到SPB的影响。
英文摘要：
LLM-as-a-judge has become the de facto approach for evaluating LLM outputs. However, judges are known to exhibit self-preference bias (SPB): they tend to favor outputs produced by themselves or by models from their own family. This skews evaluations and, thus, hinders model development, especially in settings of recursive self-improvement. We present the first study of SPB in rubric-based evaluation, an increasingly popular benchmarking paradigm where judges issue binary verdicts on individual evaluation criteria, instead of assigning holistic scores or rankings. Using IFEval and LiveCodeBench, benchmarks with programmatically verifiable rubrics, we show that SPB persists even when evaluation criteria are entirely objective: among rubrics where generators fail, judges can be more than 50% more likely to incorrectly mark them as satisfied when the output is their own. We also find that, similarly to other evaluation paradigms, ensembling multiple judges helps mitigate SPB, but without fully eliminating it. On HealthBench, a medical chat benchmark with subjective rubrics, we observe that SPB skews model scores by up to 10 points, a potentially decisive margin when ranking frontier models. We analyze the factors that drive SPB in this setting, finding that negative rubrics and subjective topics like communication and emergency referrals are particularly susceptible.
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。
