The content discusses the challenges of manually generating Socratic questions for students and introduces a method to automate this process using large language models. By augmenting datasets with invalid questions and optimizing preferences, the proposed approach outperforms existing methods. The study showcases experiments on a dataset for student code debugging, demonstrating the effectiveness of the proposed method in avoiding invalid questions and enhancing learning outcomes.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Nischal Asho... at arxiv.org 03-04-2024
https://arxiv.org/pdf/2403.00199.pdfDeeper Inquiries