arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2512.10065cs.AIcs.HC
收藏

大语言模型中线性社会人口表示从间接线索中涌现

Linear socio-demographic representations emerge in Large Language Models from indirect cues

Complex Systems Institute of Paris Ile-de-France CNRS(巴黎法兰西大区复杂系统研究所 CNRS);médialab, Sciences Po(媒体实验室,巴黎政治学院);Learning Planet Institute, CY Cergy Paris University(学习星球研究所,CY塞quenes巴黎大学)

Paul Bouchaud, Pedro Ramaciotti

AI总结:

研究揭示大语言模型通过间接线索形成社会人口统计学表示,并展示这些隐性偏见如何影响下游任务如职业推荐。

摘要:

我们研究了LLM如何编码通过姓名和职业等间接线索推断出的人类对话伙伴的社会人口属性。我们显示,LLM在激活空间中发展出用户人口统计学的线性表示,其中刻板印象关联的属性被编码在可解释的几何方向上。我们首先在四个开放的变压器基于LLM(Magistral 24B、Qwen3 14B、GPT-OSS 20B、OLMo2-1B)中探测残差流,这些模型被提示以显式披露人口统计数据。我们显示,相同的探测器可以从隐含线索预测人口统计数据:姓名激活与人口普查对齐的性别和种族表示,而职业触发与现实劳动力统计数据相关的表示。这些线性表示使我们能够解释LLM在对话中隐式形成的种族推断。我们证明这些隐式人口统计学表示积极地塑造下游行为,如职业推荐。我们的研究进一步指出,通过偏见基准测试的模型仍可能持有和利用隐式偏见,这对大规模应用的公平性有影响。

英文摘要:

We investigate how LLMs encode sociodemographic attributes of human conversational partners inferred from indirect cues such as names and occupations. We show that LLMs develop linear representations of user demographics within activation space, wherein stereotypically associated attributes are encoded along interpretable geometric directions. We first probe residual streams across layers of four open transformer-based LLMs (Magistral 24B, Qwen3 14B, GPT-OSS 20B, OLMo2-1B) prompted with explicit demographic disclosure. We show that the same probes predict demographics from implicit cues: names activate census-aligned gender and race representations, while occupations trigger representations correlated with real-world workforce statistics. These linear representations allow us to explain demographic inferences implicitly formed by LLMs during conversation. We demonstrate that these implicit demographic representations actively shape downstream behavior, such as career recommendations. Our study further highlights that models that pass bias benchmark tests may still harbor and leverage implicit biases, with implications for fairness when applied at scale.

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。

↑