arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.04439cs.CL
收藏

论未观测序列在大语言模型基于采样的不确定性量化中的作用

On the Role of Unobserved Sequences on Sample-based Uncertainty Quantification for LLMs

National Institute of Advanced Industrial Science and Technology (AIST)(国家先进工业科学与技术研究院)

Lucie Kunitomo-Jacquin, Edison Marrese-Taylor, Ken Fukuda

AI总结:

该研究针对大语言模型基于采样的不确定性量化问题,指出未观测序列的概率具有关键作用,建议未来研究纳入该因素以提升相关方法的性能。

摘要:

大语言模型(LLMs)的不确定性量化对安全关键型应用至关重要,有助于识别被称为幻觉的错误答案。不确定性量化方法的一大主流方向是基于估计LLM潜在输出序列分布的熵。该估计基于通过多次查询LLM获得的一组输出序列及对应概率。本文提出并通过实验证明,未观测序列的概率发挥着关键作用,建议未来研究将其纳入以改进此类LLM不确定性量化方法。

英文摘要:

Quantifying uncertainty in large language models (LLMs) is important for safety-critical applications because it helps spot incorrect answers, known as hallucinations. One major trend of uncertainty quantification methods is based on estimating the entropy of the distribution of the LLM's potential output sequences. This estimation is based on a set of output sequences and associated probabilities obtained by querying the LLM several times. In this paper, we advocate and experimentally show that the probability of unobserved sequences plays a crucial role, and we recommend future research to integrate it to enhance such LLM uncertainty quantification methods.

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。

↑