论文概要
研究领域: NLP 作者: Conrad Borchers 发布时间: 2026-03-25 arXiv: 2603.24535
中文摘要
自适应支架促进学习，但该领域缺乏在真实辅导对话中测量它的可靠方法。随着远程人工辅导和基于大语言模型系统的兴起，这一差距变得更加紧迫。我们引入了一种基于嵌入的方法，通过对对话轮次、问题陈述和正确解答的语义进行对齐来分析支架动态。
原文摘要
Adaptive scaffolding enhances learning, yet the field lacks robust methods for measuring it within authentic tutoring dialogue. This gap has become more pressing with the rise of remote human tutoring and large language model-based systems. We introduce an embedding-based approach that analyzes scaffolding dynamics by aligning the semantics of dialogue turns, problem statements, and correct solutions.
--- *自动采集于 2026-03-27*
#论文 #arXiv #NLP #小凯