Agenttutor: Empowering Personalized Learning With Multi-Turn Interactive Teaching In Intelligent Education Systems

Yuxin Liu, Shanghai Jiao Tong University

Zeqing Song, Shanghai Jiao Tong University

Jiong Lou, Shanghai Jiao Tong University

Chentao Wu

Jie Li

Date

12/2025

AI Overview

Why it matters

Most AI education systems rely on single-turn question-answering, which cannot assess learners' cognitive levels or dynamically adapt teaching strategies based on real-time feedback.

Study at a glance

Question: Can a multi-agent system dynamically adapt multi-turn interactive tutoring to match a learner's cognitive level?
Setting and data: Benchmark coding tasks (164 HumanEval tasks, 974 MBPP tasks) evaluated with simulated learners and three human teaching experts.
Method: System development and automated benchmark evaluation with simulated learners, complemented by expert human ratings.
AI tools: AgentTutor, GPT-3.5-turbo, GPT-4, Qwen2.5-VL-7B-Instruct, Llama-3.1-8B-instruct, Prometheus 2

Key findings

AgentTutor enabled simulated learners to achieve higher first-attempt pass rates on coding tasks than baseline methods, reaching 92.7% on HumanEval and 89.4% on MBPP.
Automated model evaluations rated AgentTutor higher across conversational adaptability, feedback quality, and question difficulty compared to baseline prompting approaches.
Three human teaching experts rated AgentTutor highest among evaluated methods across accuracy, conversational quality, helpfulness, and question set quality.

Summary drafted with AI from the full paper and may contain errors. See the full paper for details.

The rapid advancement of large-scale language models (LLMs) has shown their potential to transform intelligent education systems (IESs) through automated teaching and learning support applications. However, current IESs often rely on single-turn static question-answering, which fails to assess learners' cognitive levels, cannot adjust teaching strategies based on real-time feedback, and is limited to providing simple one-off responses. To address these issues, we introduce AgentTutor, a multi-turn interactive intelligent education system to empower personalized learning. It features an LLM-powered generative multi-agent system and a learner-specific personalized learning profile environment that dynamically optimizes and delivers teaching strategies based on learners' learning status, personalized goals, learning preferences, and multimodal study materials. It includes five key modules: curriculum decomposition, learner assessment, dynamic strategy, teaching reflection, and knowledge & experience memory. We conducted extensive experiments on multiple benchmark datasets, AgentTutor significantly enhances learners' performance while demonstrating strong effectiveness in multi-turn interactions and competitiveness in teaching quality among other baselines.