BriefGPT.xyz
Jul, 2023
大型语言模型执行诊断推理
Large Language Models Perform Diagnostic Reasoning
HTML
PDF
Cheng-Kuang Wu, Wei-Lin Chen, Hsin-Hsi Chen
TL;DR
通过将思路链式提示(CoT)拓展到医学推理领域,我们提出了诊断推理 CoT(DR-CoT)。实证结果表明,通过仅向仅训练于一般文本语料库的大型语言模型提供两个 DR-CoT 实例来指导,诊断准确性提高了 15%。此外,在领域外情景中,该差距达到显著的 18%。我们的研究结果表明,大型语言模型中可以通过适当的提示引出专家知识推理。
Abstract
We explore the extension of chain-of-thought (CoT) prompting to
medical reasoning
for the task of
automatic diagnosis
. Motivated by doctors' underlying reasoning process, we present
→