<p>Cognitive diagnosis (CD) is a core task in intelligent education, which models students’ abilities and assesses their knowledge state by analyzing the interaction data between students and exercises. Although deep learning provides new perspectives for cognitive diagnosis, single-mode assessment methods may lead to biased results when applied to learning modes that involve both theoretical and experimental paths. To address this issue, we propose the Dual-Path Awareness Neural Cognitive Diagnosis Framework (DNCD) , which aims to explore the impact of simultaneously offering theoretical and experimental components within the same course on students’ knowledge state. The model first introduces a shared student proficiency module to capture the knowledge mastery of students in both the theoretical and experimental phases. Subsequently, leveraging a task-weighting strategy, the model interacts with the connections between theoretical and experimental problems while progressively aggregating their spatiotemporal relationships, enabling more efficient information integration. Finally, through the joint-separation training mechanism, we achieve effective integration of theoretical and experimental questions while maintaining the ability for independent diagnosis, all based on ensuring the consistency of knowledge representations. This approach improves the accuracy of student interaction representation and the quality of knowledge state inference, demonstrating excellent performance in student performance prediction. Our research provides a potential future direction for the field of cognitive diagnosis, exploring integrated assessment methods in multi-mode learning environments. To support further research and innovation in this field, we will be sharing our data and source code at <a href="https://github.com/xinjiesun-ustc/DNCD">https://github.com/xinjiesun-ustc/DNCD</a>.</p>