arXiv 2403.046962024-06-06cs.CLcs.AIcs.LG

通过词元级不确定性量化对大型语言模型输出进行事实核查

Fact-Checking the Output of Large Language Models via Token-Level Uncertainty Quantification

Center for Artificial Intelligence Technology（人工智能技术中心）；MBZUAI（穆罕默德·本·扎耶德人工智能大学）；AIRI（人工智能研究院）；FRC CSC RAS（俄罗斯科学院计算机科学与控制研究中心联邦研究中心）；The University of Melbourne（墨尔本大学）；QCRI（卡塔尔计算研究所）；HSE University（高等经济大学）

Ekaterina Fadeeva, Aleksandr Rubashevskii, Artem Shelmanov, Sergey Petrakov, Haonan Li, Hamdy Mubarak, Evgenii Tsymbalov, Gleb Kuzmin, Alexander Panchenko, Timo… 展开作者

Ekaterina Fadeeva, Aleksandr Rubashevskii, Artem Shelmanov, Sergey Petrakov, Haonan Li, Hamdy Mubarak, Evgenii Tsymbalov, Gleb Kuzmin, Alexander Panchenko, Timothy Baldwin, Preslav Nakov, Maxim Panov

备注：: Accepted to ACL-2024 (Findings). Ekaterina Fadeeva, Aleksandr Rubashevskii, and Artem Shelmanov contributed equally

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出基于词元级不确定性量化的事实核查方法CCP，通过仅衡量特定主张值的不确定性来检测LLM幻觉，在传记生成任务上对七个LLMs和四种语言均优于基线，并与外部知识事实核查工具表现相当。

摘要：

大型语言模型（LLMs）因产生幻觉而臭名昭著，即在其输出中产生错误的主张。这种幻觉可能是危险的，因为生成文本中偶尔出现的事实性错误可能被其余总体上事实正确的输出所掩盖，使用户极难发现它们。当前利用LLMs的服务通常不提供任何检测不可靠生成内容的手段。在此，我们旨在弥合这一差距。特别是，我们提出了一种基于词元级不确定性量化的新型事实核查与幻觉检测流水线。不确定性分数利用封装在神经网络或其各层输出中的信息来检测不可靠的预测，我们表明这些分数可用于对LLM输出中的原子主张进行事实核查。此外，我们提出了一种新颖的词元级不确定性量化方法，该方法消除了关于在当前步骤生成什么主张以及使用什么表面形式的不确定性的影响。我们的方法Claim Conditioned Probability（CCP）仅衡量模型所表达的特定主张值的不确定性。在传记生成任务上的实验表明，与基线相比，CCP在七个LLMs和四种语言上均取得了显著改进。人工评估表明，基于不确定性量化的事实核查流水线与利用外部知识的事实核查工具相比具有竞争力。

英文摘要：

Large language models (LLMs) are notorious for hallucinating, i.e., producing erroneous claims in their output. Such hallucinations can be dangerous, as occasional factual inaccuracies in the generated text might be obscured by the rest of the output being generally factually correct, making it extremely hard for the users to spot them. Current services that leverage LLMs usually do not provide any means for detecting unreliable generations. Here, we aim to bridge this gap. In particular, we propose a novel fact-checking and hallucination detection pipeline based on token-level uncertainty quantification. Uncertainty scores leverage information encapsulated in the output of a neural network or its layers to detect unreliable predictions, and we show that they can be used to fact-check the atomic claims in the LLM output. Moreover, we present a novel token-level uncertainty quantification method that removes the impact of uncertainty about what claim to generate on the current step and what surface form to use. Our method Claim Conditioned Probability (CCP) measures only the uncertainty of a particular claim value expressed by the model. Experiments on the task of biography generation demonstrate strong improvements for CCP compared to the baselines for seven LLMs and four languages. Human evaluation reveals that the fact-checking pipeline based on uncertainty quantification is competitive with a fact-checking tool that leverages external knowledge.

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。