EduVerse: A User-Defined Multi-Agent Simulation Space for Education Scenario

TL;DR

EduVerse addresses the challenge of simultaneously reproducing cognitive development, group interaction, and longitudinal evolution in virtual classrooms, overcoming limitations of prior short-term or single-agent studies. It introduces a Cognition–Interaction–Evolution (CIE) framework with a PCA-based agent loop and a human-in-the-loop interface, coupled with environment and session customization. Validation in middle-school Chinese language settings across three genres demonstrates realistic IRF discourse ($0.28$–$0.64$) compared to real classrooms ($0.37$–$0.49$), realistic group dynamics (density $0.27$–$0.40$), and a cross-session positive transition rise of $R^{+}=11.7\%$. These findings show EduVerse as a scalable, interpretable platform for educational AI research and cross-disciplinary collaboration, with plans to release the open-source codebase.

Abstract

Reproducing cognitive development, group interaction, and long-term evolution in virtual classrooms remains a core challenge for educational AI, as real classrooms integrate open-ended cognition, dynamic social interaction, affective factors, and multi-session development rarely captured together. Existing approaches mostly focus on short-term or single-agent settings, limiting systematic study of classroom complexity and cross-task reuse. We present EduVerse, the first user-defined multi-agent simulation space that supports environment, agent, and session customization. A distinctive human-in-the-loop interface further allows real users to join the space. Built on a layered CIE (Cognition-Interaction-Evolution) architecture, EduVerse ensures individual consistency, authentic interaction, and longitudinal adaptation in cognition, emotion, and behavior-reproducing realistic classroom dynamics with seamless human-agent integration. We validate EduVerse in middle-school Chinese classes across three text genres, environments, and multiple sessions. Results show: (1) Instructional alignment: simulated IRF rates (0.28-0.64) closely match real classrooms (0.37-0.49), indicating pedagogical realism; (2) Group interaction and role differentiation: network density (0.27-0.40) with about one-third of peer links realized, while human-agent tasks indicate a balance between individual variability and instructional stability; (3) Cross-session evolution: the positive transition rate R+ increase by 11.7% on average, capturing longitudinal shifts in behavior, emotion, and cognition and revealing structured learning trajectories. Overall, EduVerse balances realism, reproducibility, and interpretability, providing a scalable platform for educational AI. The system will be open-sourced to foster cross-disciplinary research.

Figures (9)

• Figure 1: Overview of EduVerse. The framework consists of three main components: (i) user-defined environment configuration (including classroom layouts, seating arrangements, and interaction networks); (ii) CIE-based agent modeling (a three-layer Cognition–Interaction–Evolution architecture for teacher and student agents); (iii) interaction and evolution experiments (covering instructional alignment, group interaction, and cross-session development). Together, these components form a scalable, interpretable, and transferable multi-agent simulation platform for educational AI.
• Figure 2: Visualization of student group interactions across classroom environments. The three panels correspond to: Lecture (left, traditional teacher-centered layout), Collab_Two_Tables (middle, two-group collaborative setting), and Round_Table (right, open discussion layout). This figure highlights EduVerse’s capability to vary classroom environments, while showing that the resulting interaction topologies naturally support subsequent analyses of group dynamics and evolution experiments (see Sec. \ref{['subsec:exp-i']}).
• Figure 3: Overall analysis of student behavior and system design. (a) Distribution of students’ behavioral–emotional–cognitive (BEC) patterns across different classroom environments. (b) Ablation study illustrating the distinct roles of stylization and CIE modules.
• Figure 4: Stable BEC patterns across genres within individual students. Despite genre differences, individual patterns remain stable and align with traits—for instance, high-extraverted students show active engagement and varied cognition, while low-openness or low-conscientiousness students tend toward disengagement, lower-order cognition, and confusion.
• Figure 5: Positive transition trends in cross-session evolution. Rates of positive shifts in behavior, emotion, and cognition increase over time, with behavior improving most rapidly, emotion rising steadily, and cognition progressing gradually.