持一粒 SALT：LLM 在社会维度上是否公平？
With a Grain of SALT: Are LLMs Fair Across Social Dimensions?
AI总结：
本文提出 SALT 基准，从性别、宗教和种族维度评估 Llama、Gemma 等开源 LLM 的社会偏见，并结合 DeepSeek-R1 自动评估与人工验证揭示不同群体遭受系统性差别对待。
摘要：
本文系统分析了开源大型语言模型（LLM）在性别、宗教和种族方面的偏见。研究使用 SALT（Social Appropriateness in LLM-Generated Text，即 LLM 生成文本中的社会适当性）数据集评估较小规模 Llama 和 Gemma 模型的偏见，该数据集包含五种不同的偏见触发场景：General Debate（一般辩论）、Positioned Debate（立场辩论）、Career Advice（职业建议）、Problem Solving（问题解决）和 CV Generation（简历生成）。为量化偏见，我们测量 General Debate 中的胜率，以及 Positioned Debate 中负面角色的分配情况。对于 Career Advice、Problem Solving 和 CV Generation 等真实应用场景，我们对输出进行匿名化处理以移除明确的人口统计标识，并使用 DeepSeek-R1 作为自动评估器。我们还处理了基于 LLM 的评估中固有的偏见，包括评估偏见、位置偏见和长度偏见，并通过人工评估验证结果。研究发现各模型之间存在持续的两极分化，某些人口统计群体受到系统性的优待或苛待。通过引入 SALT，我们为偏见分析提供了全面基准，并强调在公平 AI 系统开发中需要稳健的偏见缓解策略。
英文摘要：
This paper presents a systematic analysis of biases in open-source Large Language Models (LLMs), across gender, religion, and race. Our study evaluates bias in smaller-scale Llama and Gemma models using the SALT ($\textbf{S}$ocial $\textbf{A}$ppropriateness in $\textbf{L}$LM-Generated $\textbf{T}$ext) dataset, which incorporates five distinct bias triggers: General Debate, Positioned Debate, Career Advice, Problem Solving, and CV Generation. To quantify bias, we measure win rates in General Debate and the assignment of negative roles in Positioned Debate. For real-world use cases, such as Career Advice, Problem Solving, and CV Generation, we anonymize the outputs to remove explicit demographic identifiers and use DeepSeek-R1 as an automated evaluator. We also address inherent biases in LLM-based evaluation, including evaluation bias, positional bias, and length bias, and validate our results through human evaluations. Our findings reveal consistent polarization across models, with certain demographic groups receiving systematically favorable or unfavorable treatment. By introducing SALT, we provide a comprehensive benchmark for bias analysis and underscore the need for robust bias mitigation strategies in the development of equitable AI systems.
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。
