面向认知诊断网络的模拟学习数据

UncASSIST数据集是基于ASSIST数据集的修改，利用SSISTMENTS在线辅导系统，去掉重复记录之后，包含包含4151个学生在110个问题上的325673个交互，用以模拟学习证据认知诊断网络对反应偏差的鲁棒性——现实世界中的反应偏差，验证知识诊断算法在现实环境下具有良好的鲁棒性、解释性和推理能力。

The UncASSIST dataset is a modified version based on the ASSIST dataset, which was developed using the SSISTMENTS online tutoring system. After deduplication, it contains 325,673 interactions from 4,151 students across 110 questions. The core goal of this dataset is to simulate the robustness of learning-evidence-based cognitive diagnosis networks against response bias—specifically, real-world response bias—and to verify that knowledge diagnosis algorithms possess strong robustness, interpretability and reasoning capabilities in real-world scenarios.

搜集汇总

背景与挑战

背景概述

该数据集是基于ASSIST数据集的修改版本，包含4151个学生在110个问题上的325673个交互记录，用于验证知识诊断算法在现实环境下的鲁棒性、解释性和推理能力。数据格式为json和docx，总数据量为52.77MB，包含5个文件。