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 2602.04064cs.CY
收藏

公民查询基准:一种新的数据集和评估流程,用于衡量LLM在公民查询任务中的性能

The CitizenQuery Benchmark: A Novel Dataset and Evaluation Pipeline for Measuring LLM Performance in Citizen Query Tasks

Open Data Institute (ODI)(开放数据研究所);NASA MSFC Office of Data Science and Informatics, University of Alabama in Huntsville(美国国家航空航天局马歇尔太空飞行中心数据科学与信息办公室,阿拉巴马大学亨茨维尔分校);National Aeronautics and Space Administration(美国国家航空航天局);Department of Data Science, Georgia Institute of Technology(佐治亚理工学院数据科学系);College of Computing, Georgia Institute of Technology(佐治亚理工学院计算机学院);Department for Informatics, King’s College London(伦敦国王学院信息学系);Department of Computer Science, University of Oxford(牛津大学计算机科学系)

Neil Majithia, Rajat Shinde, Zo Chapman, Prajun Trital, Jordan Decker, Manil Maskey, Elena Simperl, Nigel Shadbolt

AI总结:

本文提出CitizenQuery-UK数据集和评估流程,用于评估LLM在公民查询任务中的可信度,发现模型在事实性、回避率和冗余性方面存在差异,强调了可信赖AI的必要性。

摘要:

公民查询是指个人针对政府政策、指导和服务提出的问题,涵盖福利、税收、移民、就业、公共卫生等广泛主题。这为大型语言模型(LLM)提供了一个有吸引力的用例,即能够根据用户的背景提供信息,并根据他们的需求进行沟通。然而,在这种用例中,任何错误信息都可能对依赖模型响应的个人造成严重、负面且可能难以察觉的后果。为此,我们引入了CitizenQuery-UK,这是一个包含2.2万个公民查询和响应配对的数据集,这些数据是从gov.uk上大量关于英国政府的公共信息中合成生成的。我们介绍了CitizenQuery-UK的编纂方法及其内容概述。我们还提出了一种使用FActScore的改编方法来评估LLM的基准方法,用于评估11个模型的事实性、回避频率和冗余性。我们记录了这些结果,并在公共部门的背景下进行了解释,发现:(i)不同模型家族表现出不同的性能特征,但每个模型都具有竞争力;(ii)高方差削弱了实用性;(iii)回避率低而冗余率高,对可靠性有影响;(iv)更可信赖的AI需要在与用户互动时承认“不完美”。我们的研究贡献在于评估LLM在公民查询任务中的可信度;随着AI越来越多地融入日常生活中,我们的基准,完全基于公开数据,为在AI和公共部门做出更基于证据的决策奠定了基础。

英文摘要:

"Citizen queries" are questions asked by an individual about government policies, guidance, and services that are relevant to their circumstances, encompassing a range of topics including benefits, taxes, immigration, employment, public health, and more. This represents a compelling use case for Large Language Models (LLMs) that respond to citizen queries with information that is adapted to a user's context and communicated according to their needs. However, in this use case, any misinformation could have severe, negative, likely invisible ramifications for an individual placing their trust in a model's response. To this effect, we introduce CitizenQuery-UK, a benchmark dataset of 22 thousand pairs of citizen queries and responses that have been synthetically generated from the swathes of public information on $gov.uk$ about government in the UK. We present the curation methodology behind CitizenQuery-UK and an overview of its contents. We also introduce a methodology for the benchmarking of LLMs with the dataset, using an adaptation of FActScore to benchmark 11 models for factuality, abstention frequency, and verbosity. We document these results, and interpret them in the context of the public sector, finding that: (i) there are distinct performance profiles across model families, but each is competitive; (ii) high variance undermines utility; (iii) abstention is low and verbosity is high, with implications on reliability; and (iv) more trustworthy AI requires acknowledged "fallibility" in the way it interacts with users. The contribution of our research lies in assessing the trustworthiness of LLMs in citizen query tasks; as we see a world of increasing AI integration into day-to-day life, our benchmark, built entirely on open data, lays the foundations for better evidenced decision-making regarding AI and the public sector.

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。

↑