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 2403.15412cs.CYcs.AIcs.CL
收藏

迈向衡量与建模大语言模型中的“文化”:一项综述

Towards Measuring and Modeling "Culture" in LLMs: A Survey

MBZUAI(穆罕默德·本·扎耶德人工智能大学);Indian Institute of Technology Kanpur(印度理工学院坎普尔分校);Microsoft Research Africa(微软研究院非洲分部)

Muhammad Farid Adilazuarda, Sagnik Mukherjee, Pradhyumna Lavania, Siddhant Singh, Alham Fikri Aji, Jacki O'Neill, Ashutosh Modi, Monojit Choudhury

AI总结:

该综述分析了90余篇研究大语言模型中文化表征的论文,发现已有工作缺乏对文化的明确定义,仅通过人口统计学和语义代理探测了价值观等有限方面,并指出探测鲁棒性不足及文化误表征影响研究缺失等关键空白。

摘要:

我们呈现了一项对90余篇近期论文的综述,这些论文旨在研究大语言模型(LLMs)中的文化表征与包容性。我们观察到,没有一项研究明确地定义“文化”,这是一个复杂、多面向的概念;相反,它们在一些专门设计的数据集上探测模型,这些数据集代表了“文化”的某些方面。我们将这些方面称为文化的代理(proxies),并沿着人口统计学代理和语义代理两个维度对其进行组织。我们还对所使用的探测方法进行了分类。我们的分析表明,只有“文化”的某些方面,如价值观和目标,得到了研究,而其他几个有趣且重要的方面,尤其是众多的语义领域(Thompson et al., 2020)和aboutness(Hershcovich et al., 2022),尚未被探索。另外两个关键空白是探测技术缺乏鲁棒性,以及缺乏关于基于LLM的应用中文化误表征和表征不足所产生影响的情境化研究。

英文摘要:

We present a survey of more than 90 recent papers that aim to study cultural representation and inclusion in large language models (LLMs). We observe that none of the studies explicitly define "culture, which is a complex, multifaceted concept; instead, they probe the models on some specially designed datasets which represent certain aspects of "culture". We call these aspects the proxies of culture, and organize them across two dimensions of demographic and semantic proxies. We also categorize the probing methods employed. Our analysis indicates that only certain aspects of ``culture,'' such as values and objectives, have been studied, leaving several other interesting and important facets, especially the multitude of semantic domains (Thompson et al., 2020) and aboutness (Hershcovich et al., 2022), unexplored. Two other crucial gaps are the lack of robustness of probing techniques and situated studies on the impact of cultural mis- and under-representation in LLM-based applications.

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。

↑