计算机科学
模块化设计
建筑
人工智能
自然语言处理
理解力
阅读(过程)
程序设计语言
语义学(计算机科学)
系统体系结构
自然语言生成
语言学
组分(热力学)
作者
Peng Zhao,John Blake,Evgeny Pyshkin
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-06-01
卷期号:: 1-5
标识
DOI:10.1109/itet70052.2026.11651012
摘要
This paper presents a modular, multi-agent system for automated reading comprehension generation that harnesses specialised, interacting agents. The proposed pipeline separates reading material control from question control, enabling independent generation and validation of passages, taxonomy-driven question types (gist, specific, inference, absence, and lexical), and answer justifications. The system further incorporates learner performance feedback to adapt vocabulary selection and difficulty across generation rounds. Evaluation results demonstrate that the system produces pedagogically suitable reading comprehension materials, with average quality scores exceeding 92% and end-to-end generation times under one minute for 5 questions. These findings suggest that agentic LLM architectures provide an effective and transparent design pattern for scalable reading comprehension assessment generation.
科研通智能强力驱动
Strongly Powered by AbleSci AI