BriefGPT.xyz
Sep, 2022
关于上下文学习中敏感性和准确性之间的关系
On the Relation between Sensitivity and Accuracy in In-context Learning
HTML
PDF
Yanda Chen, Chen Zhao, Zhou Yu, Kathleen McKeown, He He
TL;DR
我们研究了上下文学习的灵敏度，并发现标签偏差使真实灵敏度难以确定。由此，我们提出了一种少样本选择性预测方法—— extsc{SenSel}，此方法避免了敏感的预测，实验结果表明， extsc{SenSel}在十个分类数据集上一致优于两个常用的基于置信度和熵的基准模型。
Abstract
in-context learning
(ICL) suffers from oversensitivity to the prompt, which makes it unreliable in real-world scenarios. We study the
sensitivity
of ICL with respect to multiple types of
→