Methods and Quality of Synthetic Chinese Corpora Aligned with the Proficiency Grading Standards -A Teacher Evaluation Study-
Year: 2026
Type: article
Abstract: The groundbreaking advancements in the field of artificial intelligence have introduced new possibilities for addressing challenges in the construction of educational resources.This paper explores the mechanism by which prompt design, based on the Grammar Learning Manual, influences the quality of teaching resources generated by large language models (LLMs).By evaluating model outputs under various prompt configurations, the study examines the impact of different types of prompt seeds on generation quality.The results indicate that, under the optimal prompt configuration, 86.5% of generated example sentences passed the assessment team's rigorous multi-dimensional review.Using a single seed significantly increases the risk of grammatical errors, while the combination of two seeds can markedly offset the negative effects observed with a single seed.However, employing three seeds simultaneously significantly reduces the pass rate across multiple dimensions.Moderation in providing prompt information is thus key to enhancing the generation quality of LLMs.From the perspectives of foreign language education and international Chinese education, this study investigates the notion of "moderation" in prompt engineering.The findings provide valuable references for program design optimization and prompt engineering, contributing to the high-quality generation of teaching resources and the promotion of the Proficiency Standards.
Author: Wang, Dong
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7152334026