DeepTutor: Towards Agentic Personalized Tutoring

DeepTutor, introduced by researchers in a new arXiv paper, is an open-source framework designed to overcome the limitations of conventional AI tutoring systems. Unlike static pre-training or basic RAG-augmented approaches, DeepTutor uses a hybrid personalization engine that couples static knowledge grounding with a dynamic multi-resolution memory. This memory distills interaction history into a continuously evolving learner profile, allowing the system to adapt to each student's knowledge gaps and learning pace in real time. The framework features a closed tutoring loop that bidirectionally couples citation-grounded problem solving with difficulty-calibrated question generation, ensuring both accuracy and appropriate challenge levels.

Beyond reactive interfaces, DeepTutor introduces TutorBot, a proactive multi-agent layer that deploys tutoring capabilities through extensible skills and unified multi-channel access, providing a consistent experience across platforms. The system also supports collaborative writing, multi-agent deep research, and interactive guided learning. To properly evaluate such systems, the authors created TutorBench, a student-centric benchmark with source-grounded learner profiles and a first-person interactive protocol that measures adaptive tutoring from the learner's perspective. Experiments on five authoritative benchmarks show that DeepTutor improves personalized tutoring quality while retaining strong general agentic reasoning abilities, setting a new standard for next-generation AI-powered education tools.