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 2609.00999cs.CLcs.AIcs.LG

正确框架,错误规则:文化线索暴露了它们本应缩小的金融知识差距

Right Frame, Wrong Rule: Cultural Cues Expose the Financial Knowledge Gap They Were Meant to Close

发表机构穆罕默德·本·扎耶德人工智能大学 · 东京大学 · 曼彻斯特大学
另 3 家 · 查看机构详情
  • MBZUAI(穆罕默德·本·扎耶德人工智能大学)
  • The University of Tokyo(东京大学)
  • The University of Manchester(曼彻斯特大学)
  • Georgia Institute of Technology(佐治亚理工学院)
  • Harvard University(哈佛大学)
  • INSAIT, Sofia University “St. Kliment Ohridski”(INSAIT,索菲亚大学“圣克莱门特·奥赫里德斯基”)

机构由 AI 辅助整理,请以论文原文为准。

Rania Elbadry, Ahmed Heakl, Saeed Almheiri, Fan Zhang, Muhra AlMahri, Xueqing Peng, Mohsinul Kabir, Shuyao Wang, Yi Han, Saadeldine Eletter, Duzhen Zhang, Presl… 展开作者

Rania Elbadry, Ahmed Heakl, Saeed Almheiri, Fan Zhang, Muhra AlMahri, Xueqing Peng, Mohsinul Kabir, Shuyao Wang, Yi Han, Saadeldine Eletter, Duzhen Zhang, Preslav Nakov, Yuxia Wang, Fajri Koto, Zhuohan Xie

首次发表
浏览论文内容

中文总结 AI 辅助

该研究针对规范多元主义场景,以伊斯兰金融为案例,发现文化线索会引导模型选择特定框架但导致错误答案,揭示了不同模型在框架选择与准确率上的差异,凸显了模型的框架特定能力差距。

中文摘要 AI 辅助

当一个问题在不同规范框架下都有有效答案时,语言模型必须决定使用哪个框架,以及能否在该框架内正确回答。我们将这种设定称为规范多元主义,并在伊斯兰金融中开展研究,采用四选一分类法,将框架选择与框架内正确性分离开来。这种分离揭示了刻板印象陷阱:文化线索会引导模型选择某一框架,但模型在该框架内会选出错误答案。在12种模型、2种语言和50种人口统计信号的测试中,文化线索改变了框架选择,且暴露出准确率的显著差异,尤其是在非前沿模型中。在最强信号下,大型开放权重模型97%的时间会选择伊斯兰框架。若采用二选一评估,会报告近乎完美的一致性,但其中57%至66%的选择是错误的。这些发现为能力条件路由假说提供了动机,但未直接验证该假说:模型可能更倾向于自身准确率更高的框架,而文化线索可能会暴露特定框架下的能力差距。

英文摘要

When a question has valid answers under different normative frameworks, a language model must decide which framework to use and whether it can answer correctly within it. We call this setting normative pluralism and study it in Islamic finance using a four-choice taxonomy that separates framework selection from within-framework correctness. This separation reveals the stereotype trap: a cultural cue steers a model toward one framework, but the model selects an incorrect answer within that framework. Across twelve models, two languages, and fifty demographic signals, cultural cues change framework selection and reveal substantial differences in accuracy, especially among non-frontier models. Under the strongest signal, large open-weight models select the Islamic framework 97% of the time. A two-choice evaluation would report near-perfect alignment, although 57--66% of those selections are incorrect. These findings motivate, but do not directly test, the competence-conditioned routing hypothesis: models may favor frameworks where they are more accurate, while cultural cues may expose framework-specific competence gaps.

补充信息

↑