arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2404.08710cs.GTcs.AI
收藏

大型语言模型会学习类人的战略偏好吗?

Do Large Language Models Learn Human-Like Strategic Preferences?

Tennessee Technological University(田纳西理工大学);Vanderbilt University(范德堡大学)

Jesse Roberts, Kyle Moore, Doug Fisher

AI总结:

该研究考察LLM是否形成类人战略偏好,发现Solar与Mistral在囚徒困境和旅行者困境中具备稳定的价值导向合作偏好,并揭示模型规模、表面性及滑动窗口注意力的关联。

摘要:

本文评估LLM在战略情境中是否会像已知实证结果所显示的那样,学会作出类人的偏好判断。结果表明,Solar和Mistral表现出与人类一致的、稳定的基于价值的偏好,并在囚徒困境(包括赌注规模效应)和旅行者困境(包括惩罚规模效应)中表现出类人的合作偏好。我们建立了模型规模、基于价值的偏好与表面性之间的关系。最后,本文结果表明,倾向于不那么脆弱的模型依赖滑动窗口注意力,这暗示二者之间可能存在联系。此外,我们提出了一种从任意LLM构建偏好关系的新方法,并为关于人类在旅行者困境中行为的一项假设提供了支持。

英文摘要:

In this paper, we evaluate whether LLMs learn to make human-like preference judgements in strategic scenarios as compared with known empirical results. Solar and Mistral are shown to exhibit stable value-based preference consistent with humans and exhibit human-like preference for cooperation in the prisoner's dilemma (including stake-size effect) and traveler's dilemma (including penalty-size effect). We establish a relationship between model size, value-based preference, and superficiality. Finally, results here show that models tending to be less brittle have relied on sliding window attention suggesting a potential link. Additionally, we contribute a novel method for constructing preference relations from arbitrary LLMs and support for a hypothesis regarding human behavior in the traveler's dilemma.

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。

↑